Add files using upload-large-folder tool

benchmark/IOL/ioling_hf/reports/dataset_report.json

@@ -0,0 +1,217 @@
+{
+  "built_at": "2026-07-02T22:27:24.208740+00:00",
+  "year_count": 22,
+  "years": [
+    2003,
+    2004,
+    2005,
+    2006,
+    2007,
+    2008,
+    2009,
+    2010,
+    2011,
+    2012,
+    2013,
+    2014,
+    2015,
+    2016,
+    2017,
+    2018,
+    2019,
+    2021,
+    2022,
+    2023,
+    2024,
+    2025
+  ],
+  "problem_source_count": 130,
+  "record_count": 555,
+  "text_strict_record_count": 478,
+  "text_strict_problem_source_count": 111,
+  "text_strict_years": [
+    2006,
+    2007,
+    2008,
+    2009,
+    2010,
+    2011,
+    2012,
+    2013,
+    2014,
+    2015,
+    2016,
+    2017,
+    2018,
+    2019,
+    2021,
+    2022,
+    2023,
+    2024,
+    2025
+  ],
+  "problem_only_gap_count": 4,
+  "records_by_type": {
+    "full_problem": 130,
+    "subproblem": 259,
+    "partial_variant": 166
+  },
+  "records_by_year": {
+    "2003": 36,
+    "2004": 16,
+    "2005": 24,
+    "2006": 25,
+    "2007": 6,
+    "2008": 37,
+    "2009": 15,
+    "2010": 24,
+    "2011": 28,
+    "2012": 33,
+    "2013": 18,
+    "2014": 22,
+    "2015": 28,
+    "2016": 22,
+    "2017": 22,
+    "2018": 25,
+    "2019": 26,
+    "2021": 35,
+    "2022": 32,
+    "2023": 17,
+    "2024": 24,
+    "2025": 40
+  },
+  "text_strict_records_by_type": {
+    "full_problem": 111,
+    "subproblem": 223,
+    "partial_variant": 144
+  },
+  "text_strict_records_by_year": {
+    "2006": 25,
+    "2007": 5,
+    "2008": 37,
+    "2009": 15,
+    "2010": 24,
+    "2011": 28,
+    "2012": 33,
+    "2013": 18,
+    "2014": 22,
+    "2015": 28,
+    "2016": 22,
+    "2017": 22,
+    "2018": 25,
+    "2019": 26,
+    "2021": 35,
+    "2022": 32,
+    "2023": 17,
+    "2024": 24,
+    "2025": 40
+  },
+  "selected_missing": [
+    "2004-team-solution",
+    "2005-team-solution",
+    "2006-team-problem",
+    "2006-team-solution",
+    "2013-team-solution"
+  ],
+  "download_errors": [],
+  "low_confidence_source_problem_ids": [
+    "2003-individual-1",
+    "2003-individual-2",
+    "2003-individual-3",
+    "2003-individual-4",
+    "2003-individual-5",
+    "2003-team-1",
+    "2003-team-2",
+    "2003-team-3",
+    "2004-individual-1",
+    "2004-individual-2",
+    "2004-individual-3",
+    "2004-individual-4",
+    "2004-individual-5",
+    "2005-individual-1",
+    "2005-individual-2",
+    "2005-individual-3",
+    "2005-individual-4",
+    "2005-individual-5",
+    "2007-individual-1",
+    "2007-team-1",
+    "2008-team-1",
+    "2009-team-1",
+    "2010-team-1",
+    "2011-team-1",
+    "2012-team-1",
+    "2014-team-1",
+    "2015-team-1",
+    "2016-team-1",
+    "2017-team-1",
+    "2018-team-1",
+    "2019-team-1",
+    "2021-team-1",
+    "2022-team-1",
+    "2023-team-1",
+    "2024-team-1",
+    "2025-team-1"
+  ],
+  "source_problem_ids_without_detected_subparts": [
+    "2003-individual-5",
+    "2004-individual-2",
+    "2004-individual-5",
+    "2006-individual-5",
+    "2007-individual-1",
+    "2007-individual-2",
+    "2007-individual-3",
+    "2007-individual-4",
+    "2007-individual-5",
+    "2007-team-1",
+    "2008-individual-4",
+    "2009-individual-1",
+    "2009-individual-3",
+    "2009-team-1",
+    "2010-individual-1",
+    "2013-individual-3",
+    "2013-individual-5",
+    "2014-individual-2",
+    "2014-team-1",
+    "2016-individual-1",
+    "2017-individual-4",
+    "2018-individual-1",
+    "2019-individual-4",
+    "2019-team-1",
+    "2021-individual-5",
+    "2022-individual-3",
+    "2023-individual-3",
+    "2023-individual-4",
+    "2024-individual-1"
+  ],
+  "requires_images_count": 189,
+  "manual_verification": {
+    "expected_source_problem_count": 130,
+    "covered_source_problem_count": 130,
+    "verdicts": {
+      "fix_needed": 130
+    },
+    "issue_severity": {
+      "critical": 61,
+      "high": 101,
+      "medium": 83,
+      "low": 27
+    },
+    "issue_fields": {
+      "prompt": 65,
+      "answer_units": 128,
+      "canonical": 45,
+      "provenance": 11,
+      "subparts": 20,
+      "other": 2,
+      "metadata": 1
+    },
+    "worst_source_severity": {
+      "critical": 44,
+      "high": 54,
+      "medium": 32
+    }
+  },
+  "records_by_training_readiness": {
+    "not_ready_manual_fix_needed": 555
+  }
+}

benchmark/IOL/ioling_hf/reports/eval_summary.md

@@ -0,0 +1,423 @@
+# Eval Summary
+
+Generated: 2026-06-30 23:41:54 UTC
+
+Inputs:
+- Scores: 17 file(s) from `evals/results`
+- Generations: 17 file(s) from `evals/generations`
+- Dataset metadata: `data/processed/ioling_dataset_text_strict.jsonl`
+- Startup attempts without scores: 2 from `evals/results/model_attempts.jsonl`
+
+## Run Summary
+
+| run | model | n | score | max_score | fraction | source_problems | generated_tokens | verdicts |
+| --- | --- | ---: | ---: | ---: | ---: | ---: | --- | --- |
+| HuggingFaceTB__SmolLM3-3B__diverse5 | HuggingFaceTB/SmolLM3-3B | 5 | 0 | 24 | 0.000 | 5 | 19842 total / 3968.4 avg / 472-8192 range | incorrect=5 |
+| Qwen__Qwen2.5-3B-Instruct__diverse5 | Qwen/Qwen2.5-3B-Instruct | 5 | 0 | 24 | 0.000 | 5 | 18402 total / 3680.4 avg / 229-8192 range | incorrect=5 |
+| Qwen__Qwen3-4B-Instruct-2507 | Qwen/Qwen3-4B-Instruct-2507 | 3 | 0 | 14 | 0.000 | 1 | 18526 total / 6175.3 avg / 2142-8192 range | incorrect=3 |
+| Qwen__Qwen3-4B-Instruct-2507__diverse5 | Qwen/Qwen3-4B-Instruct-2507 | 5 | 1 | 24 | 0.042 | 5 | 38530 total / 7706.0 avg / 5762-8192 range | incorrect=4, partial=1 |
+| Qwen__Qwen3-4B-Instruct-2507__strict30 | Qwen/Qwen3-4B-Instruct-2507 | 30 | 2 | 139 | 0.014 | 30 | 235730 total / 7857.7 avg / 2698-8192 range | incorrect=28, partial=2 |
+| Qwen__Qwen3-4B-Thinking-2507__diverse5 | Qwen/Qwen3-4B-Thinking-2507 | 5 | 1 | 24 | 0.042 | 5 | 40960 total / 8192.0 avg / 8192-8192 range | incorrect=4, partial=1 |
+| Qwen__Qwen3-4B-Thinking-2507__strict30 | Qwen/Qwen3-4B-Thinking-2507 | 30 | 7 | 139 | 0.050 | 30 | 245760 total / 8192.0 avg / 8192-8192 range | correct=1, incorrect=27, partial=2 |
+| google__gemma-3-12b-it__diverse5 | google/gemma-3-12b-it | 5 | 0 | 24 | 0.000 | 5 | 4271 total / 854.2 avg / 511-1395 range | incorrect=5 |
+| google__gemma-3-12b-it__strict30 | google/gemma-3-12b-it | 30 | 0 | 139 | 0.000 | 30 | 34755 total / 1158.5 avg / 384-4685 range | incorrect=30 |
+| google__gemma-3-4b-it__diverse5 | google/gemma-3-4b-it | 5 | 0 | 24 | 0.000 | 5 | 17208 total / 3441.6 avg / 195-8192 range | incorrect=5 |
+| meta-llama__Llama-3.2-3B-Instruct__diverse5 | meta-llama/Llama-3.2-3B-Instruct | 5 | 1 | 24 | 0.042 | 5 | 14943 total / 2988.6 avg / 451-8192 range | incorrect=4, partial=1 |
+| meta-llama__Llama-3.2-3B-Instruct__strict30 | meta-llama/Llama-3.2-3B-Instruct | 30 | 17 | 139 | 0.122 | 30 | 99132 total / 3304.4 avg / 241-8192 range | correct=3, incorrect=27 |
+| microsoft__Phi-4-mini-instruct | microsoft/Phi-4-mini-instruct | 3 | 0 | 14 | 0.000 | 1 | 16537 total / 5512.3 avg / 153-8192 range | incorrect=3 |
+| microsoft__Phi-4-mini-instruct__diverse5 | microsoft/Phi-4-mini-instruct | 5 | 0 | 24 | 0.000 | 5 | 16821 total / 3364.2 avg / 1-8192 range | incorrect=5 |
+| microsoft__Phi-4-mini-instruct__strict30 | microsoft/Phi-4-mini-instruct | 30 | 2 | 139 | 0.014 | 30 | 150345 total / 5011.5 avg / 1-8192 range | correct=1, incorrect=28, partial=1 |
+| mistralai__Mistral-7B-Instruct-v0.3__diverse5 | mistralai/Mistral-7B-Instruct-v0.3 | 5 | 0 | 24 | 0.000 | 5 | 2020 total / 404.0 avg / 115-759 range | incorrect=5 |
+| mistralai__Mistral-7B-Instruct-v0.3__strict30 | mistralai/Mistral-7B-Instruct-v0.3 | 30 | 5 | 139 | 0.036 | 30 | 90125 total / 3004.2 avg / 99-8192 range | correct=1, incorrect=28, partial=1 |
+
+## Sample Coverage
+
+Unique evaluated records across all runs: 35
+Unique source problems across all runs: 33
+Aggregate sources: `2006-individual-1` (3), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2015-individual-5` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-2` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1)
+
+## Attempted Runs Without Scores
+
+| run | model | status | requested | max_tokens | max_model_len | notes |
+| --- | --- | --- | ---: | ---: | ---: | --- |
+| Qwen__Qwen3-30B-A3B-Instruct-2507__diverse2 | Qwen/Qwen3-30B-A3B-Instruct-2507 | interrupted_during_startup | 2 | 8192 | 16384 | Loaded the 56.87 GiB checkpoint and completed torch.compile, then remained in vLLM profile_run while FlashInfer CUTLASS MoE kernels were built during startup; interrupted after several quiet minutes with no generations written. |
+| ibm-granite__granite-3.3-8b-instruct__diverse5 | ibm-granite/granite-3.3-8b-instruct | interrupted_during_startup | 5 | 8192 | 16384 | vLLM accepted GraniteForCausalLM and allocated about 16.8 GiB on the H200, then remained in engine initialization for several quiet minutes with no score file or generations written; interrupted cleanly. |
+
+## Per-run Coverage
+
+| run | records | source_problems | sources |
+| --- | ---: | ---: | --- |
+| HuggingFaceTB__SmolLM3-3B__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| Qwen__Qwen2.5-3B-Instruct__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| Qwen__Qwen3-4B-Instruct-2507 | 3 | 1 | `2006-individual-1` (3) |
+| Qwen__Qwen3-4B-Instruct-2507__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| Qwen__Qwen3-4B-Instruct-2507__strict30 | 30 | 30 | `2006-individual-1` (1), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1) |
+| Qwen__Qwen3-4B-Thinking-2507__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| Qwen__Qwen3-4B-Thinking-2507__strict30 | 30 | 30 | `2006-individual-1` (1), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1) |
+| google__gemma-3-12b-it__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| google__gemma-3-12b-it__strict30 | 30 | 30 | `2006-individual-1` (1), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1) |
+| google__gemma-3-4b-it__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| meta-llama__Llama-3.2-3B-Instruct__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| meta-llama__Llama-3.2-3B-Instruct__strict30 | 30 | 30 | `2006-individual-1` (1), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1) |
+| microsoft__Phi-4-mini-instruct | 3 | 1 | `2006-individual-1` (3) |
+| microsoft__Phi-4-mini-instruct__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| microsoft__Phi-4-mini-instruct__strict30 | 30 | 30 | `2006-individual-1` (1), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1) |
+| mistralai__Mistral-7B-Instruct-v0.3__diverse5 | 5 | 5 | `2006-individual-1` (1), `2011-individual-1` (1), `2015-individual-5` (1), `2021-individual-2` (1), `2025-team-1` (1) |
+| mistralai__Mistral-7B-Instruct-v0.3__strict30 | 30 | 30 | `2006-individual-1` (1), `2006-individual-5` (1), `2007-individual-5` (1), `2008-individual-2` (1), `2008-team-1` (1), `2009-individual-4` (1), `2010-individual-2` (1), `2010-team-1` (1), `2011-individual-3` (1), `2012-individual-1` (1), `2012-individual-5` (1), `2013-individual-3` (1), `2014-individual-2` (1), `2014-individual-5` (1), `2015-individual-3` (1), `2016-individual-1` (1), `2016-individual-5` (1), `2017-individual-2` (1), `2017-team-1` (1), `2018-individual-4` (1), `2019-individual-2` (1), `2019-team-1` (1), `2021-individual-3` (1), `2022-individual-1` (1), `2022-individual-5` (1), `2023-individual-3` (1), `2024-individual-1` (1), `2024-individual-4` (1), `2025-individual-2` (1), `2025-team-1` (1) |
+
+Warnings:
+- `Qwen__Qwen3-4B-Instruct-2507` is too narrow: 3 rows all map to `2006-individual-1`.
+- `microsoft__Phi-4-mini-instruct` is too narrow: 3 rows all map to `2006-individual-1`.
+Recommendation: use `python3 scripts/run_vllm_eval.py --sample-strategy diverse --limit <N> --run-name <name>` for comparison runs; keep `head` samples for smoke tests only.
+
+## Per-record Verdicts
+
+### HuggingFaceTB__SmolLM3-3B__diverse5
+
+Model: `HuggingFaceTB/SmolLM3-3B`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 8192 | 474 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 472 | 496 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 852 | 1325 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 2134 | 550 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2291 | a.1 |
+
+### Qwen__Qwen2.5-3B-Instruct__diverse5
+
+Model: `Qwen/Qwen2.5-3B-Instruct`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 450 | 480 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 229 | 498 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 1379 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 1339 | 571 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2366 | a.1 |
+
+### Qwen__Qwen3-4B-Instruct-2507
+
+Model: `Qwen/Qwen3-4B-Instruct-2507`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1 | 2006-individual-1 | full_problem | 0 | 9 | 0.000 | incorrect | 8192 | 560 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4, assignment_2.1, assignment_3.1, assignment_3.2, assignment_3.3, assignment_3.4 |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 8192 | 480 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p1-sub-assignment_2 | 2006-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 2142 | 372 | assignment_2.1 |
+
+### Qwen__Qwen3-4B-Instruct-2507__diverse5
+
+Model: `Qwen/Qwen3-4B-Instruct-2507`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 1 | 4 | 0.250 | partial | 8192 | 480 | assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 5762 | 498 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 1379 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 571 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2366 | a.1 |
+
+### Qwen__Qwen3-4B-Instruct-2507__strict30
+
+Model: `Qwen/Qwen3-4B-Instruct-2507`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 8192 | 480 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p5 | 2006-individual-5 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 598 | full.1 |
+| iol-2007-individual-p5 | 2007-individual-5 | full_problem | 0 | 4 | 0.000 | incorrect | 8192 | 525 | full.1, full.2, full.3, full.4 |
+| iol-2008-individual-p2-sub-b | 2008-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 1517 | b.1 |
+| iol-2008-team-p1-sub-a | 2008-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 646 | a.1 |
+| iol-2009-individual-p4-sub-a | 2009-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 2698 | 391 | a.1 |
+| iol-2010-individual-p2-sub-a | 2010-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 213 | a.1 |
+| iol-2010-team-p1 | 2010-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1542 | full.1 |
+| iol-2011-individual-p3-sub-a | 2011-individual-3 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 426 | a.1 |
+| iol-2012-individual-p1-sub-a | 2012-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 869 | a.1 |
+| iol-2012-individual-p5-sub-a | 2012-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 474 | a.1 |
+| iol-2013-individual-p3 | 2013-individual-3 | full_problem | 0 | 5 | 0.000 | incorrect | 8192 | 898 | full.1, full.2, full.3, full.4, full.5 |
+| iol-2014-individual-p2 | 2014-individual-2 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 492 | full.1 |
+| iol-2014-individual-p5-sub-a | 2014-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 280 | a.1 |
+| iol-2015-individual-p3-sub-a | 2015-individual-3 | subproblem | 0 | 5 | 0.000 | incorrect | 8192 | 404 | a.1, a.2, a.3, a.4, a.5 |
+| iol-2016-individual-p1 | 2016-individual-1 | full_problem | 0 | 15 | 0.000 | incorrect | 8192 | 910 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15 |
+| iol-2016-individual-p5-sub-a | 2016-individual-5 | subproblem | 0 | 2 | 0.000 | incorrect | 8192 | 312 | a.1, a.2 |
+| iol-2017-individual-p2-sub-a | 2017-individual-2 | subproblem | 0 | 16 | 0.000 | incorrect | 8192 | 393 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16 |
+| iol-2017-team-p1 | 2017-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1539 | full.1 |
+| iol-2018-individual-p4-sub-a | 2018-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 501 | a.1 |
+| iol-2019-individual-p2-sub-a | 2019-individual-2 | subproblem | 0 | 17 | 0.000 | incorrect | 8192 | 422 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17 |
+| iol-2019-team-p1 | 2019-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 2835 | full.1 |
+| iol-2021-individual-p3-sub-a | 2021-individual-3 | subproblem | 1 | 4 | 0.250 | partial | 8192 | 641 | a.2, a.3, a.4 |
+| iol-2022-individual-p1-sub-a | 2022-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 3656 | 311 | a.1 |
+| iol-2022-individual-p5-sub-a | 2022-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 728 | a.1 |
+| iol-2023-individual-p3 | 2023-individual-3 | full_problem | 0 | 13 | 0.000 | incorrect | 8192 | 624 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13 |
+| iol-2024-individual-p1 | 2024-individual-1 | full_problem | 1 | 17 | 0.059 | partial | 8192 | 728 | full.1, full.2, full.3, full.4, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15, full.16, full.17 |
+| iol-2024-individual-p4-sub-a | 2024-individual-4 | subproblem | 0 | 19 | 0.000 | incorrect | 8192 | 439 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17, a.18, a.19 |
+| iol-2025-individual-p2-sub-a | 2025-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 541 | a.1 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2366 | a.1 |
+
+### Qwen__Qwen3-4B-Thinking-2507__diverse5
+
+Model: `Qwen/Qwen3-4B-Thinking-2507`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 8192 | 480 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 498 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 1379 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 1 | 9 | 0.111 | partial | 8192 | 571 | a.1, a.2, a.3, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2366 | a.1 |
+
+### Qwen__Qwen3-4B-Thinking-2507__strict30
+
+Model: `Qwen/Qwen3-4B-Thinking-2507`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 8192 | 480 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p5 | 2006-individual-5 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 598 | full.1 |
+| iol-2007-individual-p5 | 2007-individual-5 | full_problem | 0 | 4 | 0.000 | incorrect | 8192 | 525 | full.1, full.2, full.3, full.4 |
+| iol-2008-individual-p2-sub-b | 2008-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 1517 | b.1 |
+| iol-2008-team-p1-sub-a | 2008-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 646 | a.1 |
+| iol-2009-individual-p4-sub-a | 2009-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 391 | a.1 |
+| iol-2010-individual-p2-sub-a | 2010-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 213 | a.1 |
+| iol-2010-team-p1 | 2010-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1542 | full.1 |
+| iol-2011-individual-p3-sub-a | 2011-individual-3 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 426 | a.1 |
+| iol-2012-individual-p1-sub-a | 2012-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 869 | a.1 |
+| iol-2012-individual-p5-sub-a | 2012-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 474 | a.1 |
+| iol-2013-individual-p3 | 2013-individual-3 | full_problem | 0 | 5 | 0.000 | incorrect | 8192 | 898 | full.1, full.2, full.3, full.4, full.5 |
+| iol-2014-individual-p2 | 2014-individual-2 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 492 | full.1 |
+| iol-2014-individual-p5-sub-a | 2014-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 280 | a.1 |
+| iol-2015-individual-p3-sub-a | 2015-individual-3 | subproblem | 0 | 5 | 0.000 | incorrect | 8192 | 404 | a.1, a.2, a.3, a.4, a.5 |
+| iol-2016-individual-p1 | 2016-individual-1 | full_problem | 0 | 15 | 0.000 | incorrect | 8192 | 910 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15 |
+| iol-2016-individual-p5-sub-a | 2016-individual-5 | subproblem | 0 | 2 | 0.000 | incorrect | 8192 | 312 | a.1, a.2 |
+| iol-2017-individual-p2-sub-a | 2017-individual-2 | subproblem | 2 | 16 | 0.125 | partial | 8192 | 393 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.12, a.13, a.14, a.16 |
+| iol-2017-team-p1 | 2017-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1539 | full.1 |
+| iol-2018-individual-p4-sub-a | 2018-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 501 | a.1 |
+| iol-2019-individual-p2-sub-a | 2019-individual-2 | subproblem | 0 | 17 | 0.000 | incorrect | 8192 | 422 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17 |
+| iol-2019-team-p1 | 2019-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 2835 | full.1 |
+| iol-2021-individual-p3-sub-a | 2021-individual-3 | subproblem | 4 | 4 | 1.000 | correct | 8192 | 641 |  |
+| iol-2022-individual-p1-sub-a | 2022-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 311 | a.1 |
+| iol-2022-individual-p5-sub-a | 2022-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 728 | a.1 |
+| iol-2023-individual-p3 | 2023-individual-3 | full_problem | 0 | 13 | 0.000 | incorrect | 8192 | 624 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13 |
+| iol-2024-individual-p1 | 2024-individual-1 | full_problem | 1 | 17 | 0.059 | partial | 8192 | 728 | full.1, full.2, full.3, full.4, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15, full.16, full.17 |
+| iol-2024-individual-p4-sub-a | 2024-individual-4 | subproblem | 0 | 19 | 0.000 | incorrect | 8192 | 439 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17, a.18, a.19 |
+| iol-2025-individual-p2-sub-a | 2025-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 541 | a.1 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2366 | a.1 |
+
+### google__gemma-3-12b-it__diverse5
+
+Model: `google/gemma-3-12b-it`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 511 | 485 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 786 | 495 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 823 | 1276 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 1395 | 561 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 756 | 2364 | a.1 |
+
+### google__gemma-3-12b-it__strict30
+
+Model: `google/gemma-3-12b-it`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 568 | 485 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p5 | 2006-individual-5 | full_problem | 0 | 1 | 0.000 | incorrect | 1326 | 596 | full.1 |
+| iol-2007-individual-p5 | 2007-individual-5 | full_problem | 0 | 4 | 0.000 | incorrect | 1121 | 528 | full.1, full.2, full.3, full.4 |
+| iol-2008-individual-p2-sub-b | 2008-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 742 | 1476 | b.1 |
+| iol-2008-team-p1-sub-a | 2008-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 486 | 606 | a.1 |
+| iol-2009-individual-p4-sub-a | 2009-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 384 | 421 | a.1 |
+| iol-2010-individual-p2-sub-a | 2010-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 1059 | 202 | a.1 |
+| iol-2010-team-p1 | 2010-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 2435 | 1440 | full.1 |
+| iol-2011-individual-p3-sub-a | 2011-individual-3 | subproblem | 0 | 1 | 0.000 | incorrect | 891 | 441 | a.1 |
+| iol-2012-individual-p1-sub-a | 2012-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 384 | 893 | a.1 |
+| iol-2012-individual-p5-sub-a | 2012-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 477 | 473 | a.1 |
+| iol-2013-individual-p3 | 2013-individual-3 | full_problem | 0 | 5 | 0.000 | incorrect | 1128 | 886 | full.1, full.2, full.3, full.4, full.5 |
+| iol-2014-individual-p2 | 2014-individual-2 | full_problem | 0 | 1 | 0.000 | incorrect | 1516 | 487 | full.1 |
+| iol-2014-individual-p5-sub-a | 2014-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 1524 | 285 | a.1 |
+| iol-2015-individual-p3-sub-a | 2015-individual-3 | subproblem | 0 | 5 | 0.000 | incorrect | 453 | 402 | a.1, a.2, a.3, a.4, a.5 |
+| iol-2016-individual-p1 | 2016-individual-1 | full_problem | 0 | 15 | 0.000 | incorrect | 1204 | 901 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15 |
+| iol-2016-individual-p5-sub-a | 2016-individual-5 | subproblem | 0 | 2 | 0.000 | incorrect | 706 | 310 | a.1, a.2 |
+| iol-2017-individual-p2-sub-a | 2017-individual-2 | subproblem | 0 | 16 | 0.000 | incorrect | 1177 | 398 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16 |
+| iol-2017-team-p1 | 2017-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 4685 | 1347 | full.1 |
+| iol-2018-individual-p4-sub-a | 2018-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 813 | 461 | a.1 |
+| iol-2019-individual-p2-sub-a | 2019-individual-2 | subproblem | 0 | 17 | 0.000 | incorrect | 1535 | 400 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17 |
+| iol-2019-team-p1 | 2019-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 1449 | 2809 | full.1 |
+| iol-2021-individual-p3-sub-a | 2021-individual-3 | subproblem | 0 | 4 | 0.000 | incorrect | 1010 | 621 | a.1, a.2, a.3, a.4 |
+| iol-2022-individual-p1-sub-a | 2022-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 588 | 305 | a.1 |
+| iol-2022-individual-p5-sub-a | 2022-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 1005 | 722 | a.1 |
+| iol-2023-individual-p3 | 2023-individual-3 | full_problem | 0 | 13 | 0.000 | incorrect | 1122 | 620 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13 |
+| iol-2024-individual-p1 | 2024-individual-1 | full_problem | 0 | 17 | 0.000 | incorrect | 918 | 702 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15, full.16, full.17 |
+| iol-2024-individual-p4-sub-a | 2024-individual-4 | subproblem | 0 | 19 | 0.000 | incorrect | 1511 | 423 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17, a.18, a.19 |
+| iol-2025-individual-p2-sub-a | 2025-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 1759 | 474 | a.1 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 779 | 2364 | a.1 |
+
+### google__gemma-3-4b-it__diverse5
+
+Model: `google/gemma-3-4b-it`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 396 | 485 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 233 | 495 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 1276 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 561 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 195 | 2364 | a.1 |
+
+### meta-llama__Llama-3.2-3B-Instruct__diverse5
+
+Model: `meta-llama/Llama-3.2-3B-Instruct`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 1 | 4 | 0.250 | partial | 513 | 475 | assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 726 | 497 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 451 | 1326 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 551 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 5061 | 2292 | a.1 |
+
+### meta-llama__Llama-3.2-3B-Instruct__strict30
+
+Model: `meta-llama/Llama-3.2-3B-Instruct`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 566 | 475 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p5 | 2006-individual-5 | full_problem | 0 | 1 | 0.000 | incorrect | 850 | 586 | full.1 |
+| iol-2007-individual-p5 | 2007-individual-5 | full_problem | 0 | 4 | 0.000 | incorrect | 809 | 524 | full.1, full.2, full.3, full.4 |
+| iol-2008-individual-p2-sub-b | 2008-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 1521 | b.1 |
+| iol-2008-team-p1-sub-a | 2008-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 434 | 606 | a.1 |
+| iol-2009-individual-p4-sub-a | 2009-individual-4 | subproblem | 1 | 1 | 1.000 | correct | 597 | 392 |  |
+| iol-2010-individual-p2-sub-a | 2010-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 203 | a.1 |
+| iol-2010-team-p1 | 2010-team-1 | full_problem | 1 | 1 | 1.000 | correct | 1864 | 1499 |  |
+| iol-2011-individual-p3-sub-a | 2011-individual-3 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 445 | a.1 |
+| iol-2012-individual-p1-sub-a | 2012-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 879 | a.1 |
+| iol-2012-individual-p5-sub-a | 2012-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 664 | 471 | a.1 |
+| iol-2013-individual-p3 | 2013-individual-3 | full_problem | 0 | 5 | 0.000 | incorrect | 288 | 896 | full.1, full.2, full.3, full.4, full.5 |
+| iol-2014-individual-p2 | 2014-individual-2 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 490 | full.1 |
+| iol-2014-individual-p5-sub-a | 2014-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 280 | a.1 |
+| iol-2015-individual-p3-sub-a | 2015-individual-3 | subproblem | 0 | 5 | 0.000 | incorrect | 351 | 379 | a.1, a.2, a.3, a.4, a.5 |
+| iol-2016-individual-p1 | 2016-individual-1 | full_problem | 15 | 15 | 1.000 | correct | 2676 | 892 |  |
+| iol-2016-individual-p5-sub-a | 2016-individual-5 | subproblem | 0 | 2 | 0.000 | incorrect | 620 | 314 | a.1, a.2 |
+| iol-2017-individual-p2-sub-a | 2017-individual-2 | subproblem | 0 | 16 | 0.000 | incorrect | 508 | 386 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16 |
+| iol-2017-team-p1 | 2017-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 2777 | 1452 | full.1 |
+| iol-2018-individual-p4-sub-a | 2018-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 490 | 489 | a.1 |
+| iol-2019-individual-p2-sub-a | 2019-individual-2 | subproblem | 0 | 17 | 0.000 | incorrect | 1114 | 438 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17 |
+| iol-2019-team-p1 | 2019-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 2677 | full.1 |
+| iol-2021-individual-p3-sub-a | 2021-individual-3 | subproblem | 0 | 4 | 0.000 | incorrect | 945 | 625 | a.1, a.2, a.3, a.4 |
+| iol-2022-individual-p1-sub-a | 2022-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 958 | 328 | a.1 |
+| iol-2022-individual-p5-sub-a | 2022-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 241 | 765 | a.1 |
+| iol-2023-individual-p3 | 2023-individual-3 | full_problem | 0 | 13 | 0.000 | incorrect | 938 | 617 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13 |
+| iol-2024-individual-p1 | 2024-individual-1 | full_problem | 0 | 17 | 0.000 | incorrect | 937 | 727 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15, full.16, full.17 |
+| iol-2024-individual-p4-sub-a | 2024-individual-4 | subproblem | 0 | 19 | 0.000 | incorrect | 8192 | 436 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17, a.18, a.19 |
+| iol-2025-individual-p2-sub-a | 2025-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 560 | a.1 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 6777 | 2292 | a.1 |
+
+### microsoft__Phi-4-mini-instruct
+
+Model: `microsoft/Phi-4-mini-instruct`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1 | 2006-individual-1 | full_problem | 0 | 9 | 0.000 | incorrect | 8192 | 540 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4, assignment_2.1, assignment_3.1, assignment_3.2, assignment_3.3, assignment_3.4 |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 153 | 461 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p1-sub-assignment_2 | 2006-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 357 | assignment_2.1 |
+
+### microsoft__Phi-4-mini-instruct__diverse5
+
+Model: `microsoft/Phi-4-mini-instruct`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 52 | 461 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 1 | 483 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 384 | 1219 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 8192 | 546 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2270 | a.1 |
+
+### microsoft__Phi-4-mini-instruct__strict30
+
+Model: `microsoft/Phi-4-mini-instruct`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 1 | 4 | 0.250 | partial | 8192 | 461 | assignment_1.1, assignment_1.2, assignment_1.3 |
+| iol-2006-individual-p5 | 2006-individual-5 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 546 | full.1 |
+| iol-2007-individual-p5 | 2007-individual-5 | full_problem | 0 | 4 | 0.000 | incorrect | 8192 | 505 | full.1, full.2, full.3, full.4 |
+| iol-2008-individual-p2-sub-b | 2008-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 1460 | b.1 |
+| iol-2008-team-p1-sub-a | 2008-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 6180 | 577 | a.1 |
+| iol-2009-individual-p4-sub-a | 2009-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 1 | 377 | a.1 |
+| iol-2010-individual-p2-sub-a | 2010-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 885 | 186 | a.1 |
+| iol-2010-team-p1 | 2010-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1406 | full.1 |
+| iol-2011-individual-p3-sub-a | 2011-individual-3 | subproblem | 0 | 1 | 0.000 | incorrect | 1 | 444 | a.1 |
+| iol-2012-individual-p1-sub-a | 2012-individual-1 | subproblem | 1 | 1 | 1.000 | correct | 8192 | 826 |  |
+| iol-2012-individual-p5-sub-a | 2012-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 324 | 453 | a.1 |
+| iol-2013-individual-p3 | 2013-individual-3 | full_problem | 0 | 5 | 0.000 | incorrect | 8192 | 842 | full.1, full.2, full.3, full.4, full.5 |
+| iol-2014-individual-p2 | 2014-individual-2 | full_problem | 0 | 1 | 0.000 | incorrect | 2008 | 473 | full.1 |
+| iol-2014-individual-p5-sub-a | 2014-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 520 | 276 | a.1 |
+| iol-2015-individual-p3-sub-a | 2015-individual-3 | subproblem | 0 | 5 | 0.000 | incorrect | 502 | 374 | a.1, a.2, a.3, a.4, a.5 |
+| iol-2016-individual-p1 | 2016-individual-1 | full_problem | 0 | 15 | 0.000 | incorrect | 8192 | 869 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15 |
+| iol-2016-individual-p5-sub-a | 2016-individual-5 | subproblem | 0 | 2 | 0.000 | incorrect | 373 | 306 | a.1, a.2 |
+| iol-2017-individual-p2-sub-a | 2017-individual-2 | subproblem | 0 | 16 | 0.000 | incorrect | 118 | 374 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16 |
+| iol-2017-team-p1 | 2017-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1335 | full.1 |
+| iol-2018-individual-p4-sub-a | 2018-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 467 | a.1 |
+| iol-2019-individual-p2-sub-a | 2019-individual-2 | subproblem | 0 | 17 | 0.000 | incorrect | 163 | 406 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17 |
+| iol-2019-team-p1 | 2019-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 2663 | full.1 |
+| iol-2021-individual-p3-sub-a | 2021-individual-3 | subproblem | 0 | 4 | 0.000 | incorrect | 1 | 577 | a.1, a.2, a.3, a.4 |
+| iol-2022-individual-p1-sub-a | 2022-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 314 | a.1 |
+| iol-2022-individual-p5-sub-a | 2022-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 740 | a.1 |
+| iol-2023-individual-p3 | 2023-individual-3 | full_problem | 0 | 13 | 0.000 | incorrect | 8192 | 608 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13 |
+| iol-2024-individual-p1 | 2024-individual-1 | full_problem | 0 | 17 | 0.000 | incorrect | 8192 | 665 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15, full.16, full.17 |
+| iol-2024-individual-p4-sub-a | 2024-individual-4 | subproblem | 0 | 19 | 0.000 | incorrect | 8192 | 424 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17, a.18, a.19 |
+| iol-2025-individual-p2-sub-a | 2025-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 5 | 501 | a.1 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 2270 | a.1 |
+
+### mistralai__Mistral-7B-Instruct-v0.3__diverse5
+
+Model: `mistralai/Mistral-7B-Instruct-v0.3`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 146 | 538 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2011-individual-p1-sub-a | 2011-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 115 | 534 | a.1 |
+| iol-2015-individual-p5-sub-b | 2015-individual-5 | subproblem | 0 | 9 | 0.000 | incorrect | 434 | 1497 | b.1, b.2, b.3, b.4, b.5, b.6, b.7, b.8, b.9 |
+| iol-2021-individual-p2-sub-a | 2021-individual-2 | subproblem | 0 | 9 | 0.000 | incorrect | 566 | 634 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 759 | 2671 | a.1 |
+
+### mistralai__Mistral-7B-Instruct-v0.3__strict30
+
+Model: `mistralai/Mistral-7B-Instruct-v0.3`
+
+| record_id | source | item_type | score | max_score | fraction | verdict | generated_tokens | prompt_tokens | missed_units |
+| --- | --- | --- | ---: | ---: | ---: | --- | ---: | ---: | --- |
+| iol-2006-individual-p1-sub-assignment_1 | 2006-individual-1 | subproblem | 0 | 4 | 0.000 | incorrect | 99 | 538 | assignment_1.1, assignment_1.2, assignment_1.3, assignment_1.4 |
+| iol-2006-individual-p5 | 2006-individual-5 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 692 | full.1 |
+| iol-2007-individual-p5 | 2007-individual-5 | full_problem | 1 | 4 | 0.250 | partial | 1711 | 641 | full.1, full.2, full.3 |
+| iol-2008-individual-p2-sub-b | 2008-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 2040 | 1700 | b.1 |
+| iol-2008-team-p1-sub-a | 2008-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 672 | a.1 |
+| iol-2009-individual-p4-sub-a | 2009-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 190 | 477 | a.1 |
+| iol-2010-individual-p2-sub-a | 2010-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 1020 | 242 | a.1 |
+| iol-2010-team-p1 | 2010-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1633 | full.1 |
+| iol-2011-individual-p3-sub-a | 2011-individual-3 | subproblem | 0 | 1 | 0.000 | incorrect | 305 | 495 | a.1 |
+| iol-2012-individual-p1-sub-a | 2012-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 295 | 985 | a.1 |
+| iol-2012-individual-p5-sub-a | 2012-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 595 | 545 | a.1 |
+| iol-2013-individual-p3 | 2013-individual-3 | full_problem | 0 | 5 | 0.000 | incorrect | 312 | 1030 | full.1, full.2, full.3, full.4, full.5 |
+| iol-2014-individual-p2 | 2014-individual-2 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 545 | full.1 |
+| iol-2014-individual-p5-sub-a | 2014-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 2083 | 322 | a.1 |
+| iol-2015-individual-p3-sub-a | 2015-individual-3 | subproblem | 0 | 5 | 0.000 | incorrect | 8192 | 453 | a.1, a.2, a.3, a.4, a.5 |
+| iol-2016-individual-p1 | 2016-individual-1 | full_problem | 0 | 15 | 0.000 | incorrect | 8192 | 1107 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15 |
+| iol-2016-individual-p5-sub-a | 2016-individual-5 | subproblem | 0 | 2 | 0.000 | incorrect | 168 | 350 | a.1, a.2 |
+| iol-2017-individual-p2-sub-a | 2017-individual-2 | subproblem | 0 | 16 | 0.000 | incorrect | 1001 | 471 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16 |
+| iol-2017-team-p1 | 2017-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 1698 | full.1 |
+| iol-2018-individual-p4-sub-a | 2018-individual-4 | subproblem | 0 | 1 | 0.000 | incorrect | 342 | 514 | a.1 |
+| iol-2019-individual-p2-sub-a | 2019-individual-2 | subproblem | 0 | 17 | 0.000 | incorrect | 1269 | 454 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17 |
+| iol-2019-team-p1 | 2019-team-1 | full_problem | 0 | 1 | 0.000 | incorrect | 8192 | 3182 | full.1 |
+| iol-2021-individual-p3-sub-a | 2021-individual-3 | subproblem | 4 | 4 | 1.000 | correct | 223 | 715 |  |
+| iol-2022-individual-p1-sub-a | 2022-individual-1 | subproblem | 0 | 1 | 0.000 | incorrect | 627 | 324 | a.1 |
+| iol-2022-individual-p5-sub-a | 2022-individual-5 | subproblem | 0 | 1 | 0.000 | incorrect | 8192 | 898 | a.1 |
+| iol-2023-individual-p3 | 2023-individual-3 | full_problem | 0 | 13 | 0.000 | incorrect | 226 | 705 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13 |
+| iol-2024-individual-p1 | 2024-individual-1 | full_problem | 0 | 17 | 0.000 | incorrect | 710 | 788 | full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15, full.16, full.17 |
+| iol-2024-individual-p4-sub-a | 2024-individual-4 | subproblem | 0 | 19 | 0.000 | incorrect | 1275 | 494 | a.1, a.2, a.3, a.4, a.5, a.6, a.7, a.8, a.9, a.10, a.11, a.12, a.13, a.14, a.15, a.16, a.17, a.18, a.19 |
+| iol-2025-individual-p2-sub-a | 2025-individual-2 | subproblem | 0 | 1 | 0.000 | incorrect | 1486 | 547 | a.1 |
+| iol-2025-team-p1-sub-a | 2025-team-1 | subproblem | 0 | 1 | 0.000 | incorrect | 420 | 2671 | a.1 |

benchmark/IOL/ioling_hf/reports/extraction_errors.json

@@ -0,0 +1,18 @@
+[
+  {
+    "contest": "2004-team",
+    "error": "missing selected problem or solution PDF"
+  },
+  {
+    "contest": "2005-team",
+    "error": "missing selected problem or solution PDF"
+  },
+  {
+    "contest": "2006-team",
+    "error": "missing selected problem or solution PDF"
+  },
+  {
+    "contest": "2013-team",
+    "error": "missing selected problem or solution PDF"
+  }
+]

benchmark/IOL/ioling_hf/reports/focused_reasoning/qwen3_4b_v16_answer_distill_manual_audit.json

@@ -0,0 +1,38 @@
+{
+  "probe": "reports/focused_reasoning/mnt_disk_ioling_checkpoints_qwen3-4b-ioling-microsteps-v16-breadth-e1-lr5e7-merged__qwen3_4b_v16_answer_distill_restart.json",
+  "review_scope": "All 38 exact-final samples in the answer-given probe were inspected for process faithfulness. An exact final answer is not sufficient for promotion to SFT.",
+  "rubric": [
+    "Every lexical, morphological, arithmetic, and word-order claim must be supported by the displayed problem, a checked non-answer rule card, or a checked intermediate parse.",
+    "The trace must not cite the target answer, a target-only example, or the answer itself as evidence.",
+    "The trace must not contain a contradictory mapping, invented example number, unsupported operation, or wrong tense/number/case assignment.",
+    "The final box must be exactly correct and use the required unit id."
+  ],
+  "counts": {
+    "samples_reviewed": 38,
+    "exact_final_samples": 38,
+    "raw_traces_accepted": 9,
+    "raw_traces_rejected": 29,
+    "accepted_by_method": {
+      "answer_plus_rule": 4,
+      "answer_rule_decomposition": 5
+    }
+  },
+  "accepted_raw_samples": [
+    {"method": "answer_plus_rule", "record_id": "iol-2015-individual-p1-sub-b-atomic-b.1", "sample_index": 0, "reason": "Correct 42 = 2*20+2 derivation, correct bound coefficient, correct -om- linker, no unsupported citation."},
+    {"method": "answer_plus_rule", "record_id": "iol-2015-individual-p1-sub-b-atomic-b.1", "sample_index": 3, "reason": "Uses equation (2), correctly maps 2 and 20, and checks the arithmetic."},
+    {"method": "answer_rule_decomposition", "record_id": "iol-2015-individual-p1-sub-b-atomic-b.1", "sample_index": 2, "reason": "Compact and faithful application of the checked rule card."},
+    {"method": "answer_rule_decomposition", "record_id": "iol-2015-individual-p1-sub-b-atomic-b.1", "sample_index": 3, "reason": "Compact arithmetic and linker derivation without a false citation."},
+    {"method": "answer_plus_rule", "record_id": "iol-2015-individual-p1-sub-c-atomic-c.1", "sample_index": 1, "reason": "Correct base-six decomposition and descending place order."},
+    {"method": "answer_plus_rule", "record_id": "iol-2015-individual-p1-sub-c-atomic-c.1", "sample_index": 2, "reason": "Correct checked mappings and arithmetic; no contradictory claim."},
+    {"method": "answer_rule_decomposition", "record_id": "iol-2015-individual-p1-sub-c-atomic-c.1", "sample_index": 0, "reason": "Correct mappings, place order, and arithmetic."},
+    {"method": "answer_rule_decomposition", "record_id": "iol-2015-individual-p1-sub-c-atomic-c.1", "sample_index": 1, "reason": "Correct mappings and descending place order."},
+    {"method": "answer_rule_decomposition", "record_id": "iol-2015-individual-p1-sub-c-atomic-c.1", "sample_index": 2, "reason": "Correct mappings and arithmetic check."}
+  ],
+  "rejection_patterns": {
+    "target_or_answer_used_as_evidence": 10,
+    "invented_or_wrong_example_citation": 12,
+    "wrong_morphology_or_grammatical_assignment": 11,
+    "nonsensical_or_off_topic_reasoning": 8
+  },
+  "promotion_policy": "Only accepted_raw_samples may be used as generated trace candidates. For all other exact finals, write a fresh manually checked trace from the ledger; do not repair them automatically."
+}

benchmark/IOL/ioling_hf/reports/focused_reasoning/qwen3_4b_v16_answer_distill_manual_audit.md

@@ -0,0 +1,17 @@
+# Answer-Given Distillation Audit
+
+The v16 Qwen3-4B answer-given probe produced 38 exact final boxes out of 72 samples. Exact final accuracy is therefore not a process-supervision signal by itself.
+
+I manually inspected all 38 exact samples. Only 9 traces were faithful enough to promote as raw trace candidates. The rest had at least one of these defects: citing the target or answer as evidence, inventing example numbers, assigning the wrong tense/case/number, or producing unrelated text while guessing the correct box.
+
+The accepted samples are recorded in the adjacent JSON file. They are not automatically added to training. The production SFT set continues to use manually written, checked traces; generated traces require this same per-record audit before promotion.
+
+Probe totals:
+
+| method | exact finals | manually accepted traces |
+| --- | ---: | ---: |
+| `answer_given` | 1/24 | 0 |
+| `answer_plus_rule` | 16/24 | 4 |
+| `answer_rule_decomposition` | 21/24 | 5 |
+
+The useful conclusion is narrow: a checked rule card plus a strict four-line decomposition makes the model much more likely to emit the known answer, but it does not make the generated rationale trustworthy. Any self-distillation pipeline must include manual or independently verified process checks before assigning a positive training weight.

benchmark/IOL/ioling_hf/reports/focused_reasoning_summary.md

@@ -0,0 +1,105 @@
+# Focused Reasoning and Aggregation Investigation
+
+Generated 2026-07-11. All generations used temperature `1.0`, top-p `1.0`, a
+4,096-token output budget, 16,384-token context, and one H100. The focus set has
+six atomic targets from three source problems. Every target and answer was
+independently compared with rendered official problem and solution pages before
+the experiments.
+
+## Focus Set
+
+| source | target | official answer |
+| --- | --- | --- |
+| 2015 Nahuatl | write 42 | `öm-pöhualli-om-öme` |
+| 2015 Arammba | write 43 | `fete nimbo ngámbi` |
+| 2016 Kunuz Nubian | `magasi argi ajomirra` | The thieves are striking us. |
+| 2016 Kunuz Nubian | `ay darbadki bijaːndi` | I will buy the chicken. |
+| 2021 Kilivila | `Amagudina gwadi lekota?` | Which child arrived? |
+| 2021 Kilivila | `Legisesi ketala waga vivila minasiwena.` | Those women saw one canoe. |
+
+## Main Results
+
+The counts below are rescored with the current strict keyed scorer, including
+two manually reviewed English aliases. `records` is pass@k coverage on the six
+focus records.
+
+| model | condition | exact samples | records | truncation | interpretation |
+| --- | --- | ---: | ---: | ---: | --- |
+| Qwen3 4B | IID reasoning | 0/48 | 0/6 | 62.5% | rule induction fails and reasoning does not terminate |
+| Qwen3 4B | evidence ledger | 0/48 | 0/6 | 56.3% | structure alone does not install mappings |
+| Qwen3 4B | target decomposition only | 0/48 | 0/6 | 95.8% | decomposition without trusted rules increases conflict/looping |
+| Qwen3 4B | aggregate eight IID attempts | 0/24 | 0/6 | 4.2% | consensus amplifies shared wrong analyses |
+| Qwen3 4B | original evidence GROOT 1x4 | 0/20 | 0/6 | 45.0% | branches invent incompatible grammars |
+| Qwen3 4B | official rule + target decomposition | **30/48** | **6/6** | **0.0%** | answer-hidden but heavily scaffolded execution |
+| Qwen3 4B | aggregate informed attempts | 12/24 | 4/6 | 25.0% | worse than direct pass@8; long candidate audits reintroduce drift |
+| Qwen3 4B | rule-grounded GROOT 1x4 | 6/20 | 4/6 | 40.0% | complementary branches still add length and errors |
+| Qwen3 4B | trusted target parse, no answer | 27/48 | 5/6 | 0.0% | local execution is still brittle without the answer |
+| Qwen3 4B | trusted parse + answer | 48/48 | 6/6 | 0.0% | faithful constrained imitation, not independent solving |
+| Qwen2.5 7B | official rule + decomposition | 5/48 | 3/6 | 0.0% | short but substantially less accurate |
+| Qwen2.5 7B | trusted target parse | 12/48 | 5/6 | 0.0% | follows format, often changes supplied meanings |
+| Gemma 3 4B | official rule + decomposition | 0/48 | 0/6 | 0.0% | content and boxed-contract failures |
+| Gemma 3 4B | trusted target parse | 0/48 | 0/6 | 0.0% | content and boxed-contract failures |
+
+## What Caused the Bad Reasoning
+
+The long outputs are not productive search loops toward the answer. The model
+repeatedly changes which surface token is a noun, verb, tense marker, or
+grammatical role. Once an early alignment is wrong, later prose elaborates that
+alignment rather than falsifying it. Four specific effects were observed:
+
+1. A 4K cap is adequate for a correct local derivation, but free reasoning uses
+   the budget to restart and reaches the box too late.
+2. Supplying only the target decomposition makes truncation worse because the
+   model tries to reconcile it with independently invented lexical mappings.
+3. Aggregation cannot recover when all candidates share the same mistaken
+   lexicon. Candidate frequency is then anti-evidence.
+4. GROOT was applied structurally as anchor -> planner -> four guided solves,
+   but “different approaches” is the wrong diversity axis for a task with one
+   latent grammar. Even the corrected rule-grounded variant is worse than
+   direct rule + decomposition prompting.
+
+The useful intervention is therefore not more free-form reasoning. It is a
+manually verified intermediate state: lexical/morpheme rules plus a target
+decomposition, followed by short composition and checking.
+
+## Manual Rationale Audit
+
+Every one of the 30 exact outputs from the heavily scaffolded Qwen3 condition
+was read manually. Only **16/30** pass a strict SFT standard requiring every
+intermediate claim to be true. Correct boxes with wrong example citations,
+case/tense errors, or self-corrected false paths were rejected. The full
+sample-index audit is stored in
+`data/manual_overrides/focused_reasoning_trace_audit_v1.json`.
+
+The 48 answer-plus-trusted-parse traces are all faithful and average 85 tokens,
+but both the parse and final answer were in the prompt. They are appropriate as
+constrained imitation/SFT targets after deduplication; they are not evidence of
+independent model reasoning and should not be reported as benchmark accuracy.
+
+## Scoring Changes
+
+The strict keyed scorer remains unchanged. Two narrow accepted values were
+added at dataset construction time after manual review:
+
+- `The thieves strike us.` for the aspect-neutral Kunuz present;
+- `One canoe was seen by those women.` for the equivalent Kilivila passive.
+
+This raises the heavily scaffolded condition from 28/48 to 30/48 without fuzzy
+matching. The associated generated rationales were still rejected where their
+intermediate claims were false.
+
+## Training Decision
+
+Use Qwen3-4B-Instruct-2507 rather than Gemma 3 4B or Qwen2.5 7B for the next
+small run. Do not start with accuracy-only RL on raw full problems. First build
+manually checked references for the training targets, generate candidates from
+the same 4B model using the lowest hint level that works, and select only traces
+that pass manual review and have high likelihood under the raw prompt. Perform
+a short LoRA SFT stage, then re-run raw-prompt pass@8. Start verifier-grounded
+online training only if the SFT model produces both correct and incorrect raw
+samples.
+
+The detailed training design is in
+[`docs/on_policy_distillation.md`](../docs/on_policy_distillation.md).
+
+The focused trace viewer is [here](focused_reasoning/index.html).

benchmark/IOL/ioling_hf/reports/index.html

@@ -0,0 +1,271 @@
+<!doctype html>
+<html lang="en">
+<head>
+  <meta charset="utf-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1">
+  <title>IOLing Dataset Report</title>
+  <style>
+    :root { --ink:#202124; --muted:#5f6368; --line:#d8dee4; --soft:#f6f8fa; --accent:#0b7285; --good:#1a7f37; --warn:#9a6700; }
+    * { box-sizing:border-box; }
+    body { margin:0; font:14px/1.5 system-ui, -apple-system, Segoe UI, sans-serif; color:var(--ink); background:white; }
+    header { padding:22px 28px; border-bottom:1px solid var(--line); background:var(--soft); }
+    h1 { margin:0 0 6px; font-size:24px; }
+    h2 { margin:28px 0 10px; font-size:18px; }
+    h3 { margin:20px 0 8px; font-size:14px; color:var(--muted); text-transform:uppercase; letter-spacing:.04em; }
+    a { color:var(--accent); }
+    main { max-width:1440px; margin:0 auto; padding:20px 28px 36px; }
+    .links { display:flex; gap:12px; flex-wrap:wrap; margin-top:10px; }
+    .links a { border:1px solid var(--line); border-radius:6px; padding:6px 10px; text-decoration:none; background:white; }
+    .metrics { display:grid; grid-template-columns:repeat(auto-fit, minmax(170px, 1fr)); gap:10px; margin:14px 0; }
+    .metric { border:1px solid var(--line); border-radius:6px; padding:11px 12px; background:#fff; }
+    .metric strong { display:block; font-size:24px; line-height:1.1; }
+    .metric label { display:block; color:var(--muted); margin-top:3px; }
+    .metric span { display:block; color:var(--muted); font-size:12px; margin-top:3px; }
+    .grid { display:grid; grid-template-columns:repeat(auto-fit, minmax(320px, 1fr)); gap:16px; }
+    table { border-collapse:collapse; width:100%; margin:8px 0 18px; }
+    th, td { border-bottom:1px solid var(--line); padding:8px 9px; text-align:left; vertical-align:top; }
+    th { background:#f1f3f4; color:var(--muted); font-size:12px; font-weight:650; }
+    .table-wrap { overflow:auto; border:1px solid var(--line); border-radius:6px; }
+    .table-wrap table { margin:0; }
+    .run { overflow-wrap:anywhere; min-width:230px; }
+    .muted { color:var(--muted); }
+    .callout { border-left:4px solid var(--warn); background:#fff8c5; padding:10px 12px; margin:12px 0; }
+    .ok { border-left-color:var(--good); background:#dafbe1; }
+    .bar { height:6px; background:#eaeef2; border-radius:999px; margin-top:5px; min-width:72px; overflow:hidden; }
+    .bar span { display:block; height:100%; background:var(--accent); }
+    code { background:#f1f3f4; padding:1px 4px; border-radius:4px; }
+  </style>
+</head>
+<body>
+  <header>
+    <h1>IOLing Dataset Report</h1>
+    <div class="muted">Generated 2026-07-11 21:26:52 UTC. Build timestamp: 2026-07-02T22:27:24.208740+00:00.</div>
+    <nav class="links">
+      <a href="../viewer/index.html">Record viewer</a>
+      <a href="../docs/index.html">Current research report</a>
+      <a href="eval_summary.md">Markdown eval summary</a>
+      <a href="../docs/scoring.md">Scoring definition</a>
+      <a href="rl_training_summary.md">RL training summary</a>
+      <a href="rl_strict_probe_summary.md">RL strict probes</a>
+      <a href="rl_probe_qualitative_error_analysis.md">Qualitative errors</a>
+      <a href="rl_false_negative_manual_review.md">False-negative audit</a>
+      <a href="rl_validation_zero_diagnosis.md">RL validation diagnosis</a>
+      <a href="ioling_hint_diversity_summary.md">Hint/diversity probes</a>
+      <a href="focused_reasoning/index.html">Focused reasoning report</a>
+      <a href="../docs/on_policy_distillation.md">SFT and online training plan</a>
+      <a href="https://huggingface.co/datasets/agurung/ioling">Hugging Face dataset</a>
+      <a href="https://github.com/Alex-Gurung/ioling">GitHub repo</a>
+    </nav>
+  </header>
+  <main>
+    <section>
+      <h2>Dataset</h2>
+      <div class="metrics">
+        <div class="metric"><strong>555</strong><label>Full solution-backed records</label><span>130 source problems</span></div>
+        <div class="metric"><strong>478</strong><label>Strict text records</label><span>111 source problems</span></div>
+        <div class="metric"><strong>15</strong><label>RL reward candidates</label><span>conservative schema-gated subset</span></div>
+        <div class="metric"><strong>9/6</strong><label>RL v1 train/val</label><span>strict_keyed_units_v5</span></div>
+        <div class="metric"><strong>9/6</strong><label>RL v7 concise train/val</label><span>strict_keyed_units_v6</span></div>
+        <div class="metric"><strong>4</strong><label>Problem-only gaps</label><span>excluded from strict scoring</span></div>
+        <div class="metric"><strong>2003-2025</strong><label>Full years</label><span>2020 not listed on source site</span></div>
+        <div class="metric"><strong>2006-2025</strong><label>Strict years</label><span>2006-2019, 2021-2025</span></div>
+        <div class="metric"><strong>305/152</strong><label>Strict table/image flags</label><span>tables / images</span></div>
+      </div>
+      <div class="grid">
+        <div>
+          <h3>Strict Record Types</h3>
+          <div class="table-wrap"><table><thead><tr><th>type</th><th>records</th></tr></thead><tbody><tr><td>full_problem</td><td>111</td></tr><tr><td>partial_variant</td><td>144</td></tr><tr><td>subproblem</td><td>223</td></tr></tbody></table></div>
+        </div>
+        <div>
+          <h3>Strict Curation Status</h3>
+          <div class="table-wrap"><table><thead><tr><th>status</th><th>records</th></tr></thead><tbody><tr><td>auto_extracted_text_high_confidence</td><td>384</td></tr><tr><td>manual_reviewed_source_text</td><td>94</td></tr></tbody></table></div>
+        </div>
+        <div>
+          <h3>Strict Training Readiness</h3>
+          <div class="table-wrap"><table><thead><tr><th>status</th><th>records</th></tr></thead><tbody><tr><td>not_ready_manual_fix_needed</td><td>478</td></tr></tbody></table></div>
+        </div>
+        <div>
+          <h3>Full Record Types</h3>
+          <div class="table-wrap"><table><thead><tr><th>type</th><th>records</th></tr></thead><tbody><tr><td>full_problem</td><td>130</td></tr><tr><td>partial_variant</td><td>166</td></tr><tr><td>subproblem</td><td>259</td></tr></tbody></table></div>
+        </div>
+      </div>
+    </section>
+
+    <section>
+      <h2>Manual Verification</h2>
+      <div class="callout">Manual source-level verification covered all 130/130 solution-backed source problems. Every source currently has at least one unresolved finding, so the dataset is an audited curation queue, not an RL-ready answer-label corpus.</div>
+      <div class="grid">
+        <div>
+          <h3>Verdicts</h3>
+          <div class="table-wrap"><table><thead><tr><th>verdict</th><th>sources</th></tr></thead><tbody><tr><td>fix_needed</td><td>130</td></tr></tbody></table></div>
+        </div>
+        <div>
+          <h3>Issue Severity</h3>
+          <div class="table-wrap"><table><thead><tr><th>severity</th><th>issues</th></tr></thead><tbody><tr><td>critical</td><td>61</td></tr><tr><td>high</td><td>101</td></tr><tr><td>low</td><td>27</td></tr><tr><td>medium</td><td>83</td></tr></tbody></table></div>
+        </div>
+        <div>
+          <h3>Worst Source Severity</h3>
+          <div class="table-wrap"><table><thead><tr><th>severity</th><th>sources</th></tr></thead><tbody><tr><td>critical</td><td>44</td></tr><tr><td>high</td><td>54</td></tr><tr><td>medium</td><td>32</td></tr></tbody></table></div>
+        </div>
+        <div>
+          <h3>Issue Fields</h3>
+          <div class="table-wrap"><table><thead><tr><th>field</th><th>issues</th></tr></thead><tbody><tr><td>answer_units</td><td>128</td></tr><tr><td>canonical</td><td>45</td></tr><tr><td>metadata</td><td>1</td></tr><tr><td>other</td><td>2</td></tr><tr><td>prompt</td><td>65</td></tr><tr><td>provenance</td><td>11</td></tr><tr><td>subparts</td><td>20</td></tr></tbody></table></div>
+        </div>
+      </div>
+      <p>Detailed findings: <a href="manual_verification/verification_summary.md">manual verification summary</a>. Reward audit: <a href="answer_unit_audit_text_strict.md">strict answer-unit audit</a>.</p>
+    </section>
+
+    <section>
+      <h2>Strict30 Model Results</h2>
+      <div class="callout ok">These runs use 30 diverse strict records, 8k max generated tokens, and vLLM defaults except <code>max_model_len=16384</code>.</div>
+      <div class="callout">Scores are deterministic answer-unit lower bounds. They are useful for comparing runs on this sample, but all strict records still need human or judge-model review for final grading.</div>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>run</th><th>model</th><th>n</th><th>score</th><th>fraction</th><th>sources</th><th>generated tokens</th><th>verdicts</th></tr></thead>
+          <tbody><tr><td class="run">meta-llama__Llama-3.2-3B-Instruct__strict30</td><td>meta-llama/Llama-3.2-3B-Instruct</td><td>30</td><td>17/139</td><td>12.2%<div class="bar"><span style="width:12.23%"></span></div></td><td>30</td><td>99132 total<br><span class="muted">3304.4 avg, 9 capped</span></td><td>correct=3, incorrect=27</td></tr>
+<tr><td class="run">Qwen__Qwen3-4B-Thinking-2507__strict30</td><td>Qwen/Qwen3-4B-Thinking-2507</td><td>30</td><td>7/139</td><td>5.0%<div class="bar"><span style="width:5.04%"></span></div></td><td>30</td><td>245760 total<br><span class="muted">8192.0 avg, 30 capped</span></td><td>correct=1, incorrect=27, partial=2</td></tr>
+<tr><td class="run">mistralai__Mistral-7B-Instruct-v0.3__strict30</td><td>mistralai/Mistral-7B-Instruct-v0.3</td><td>30</td><td>5/139</td><td>3.6%<div class="bar"><span style="width:3.60%"></span></div></td><td>30</td><td>90125 total<br><span class="muted">3004.2 avg, 9 capped</span></td><td>correct=1, incorrect=28, partial=1</td></tr>
+<tr><td class="run">Qwen__Qwen3-4B-Instruct-2507__strict30</td><td>Qwen/Qwen3-4B-Instruct-2507</td><td>30</td><td>2/139</td><td>1.4%<div class="bar"><span style="width:1.44%"></span></div></td><td>30</td><td>235730 total<br><span class="muted">7857.7 avg, 28 capped</span></td><td>incorrect=28, partial=2</td></tr>
+<tr><td class="run">microsoft__Phi-4-mini-instruct__strict30</td><td>microsoft/Phi-4-mini-instruct</td><td>30</td><td>2/139</td><td>1.4%<div class="bar"><span style="width:1.44%"></span></div></td><td>30</td><td>150345 total<br><span class="muted">5011.5 avg, 17 capped</span></td><td>correct=1, incorrect=28, partial=1</td></tr>
+<tr><td class="run">google__gemma-3-12b-it__strict30</td><td>google/gemma-3-12b-it</td><td>30</td><td>0/139</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>30</td><td>34755 total<br><span class="muted">1158.5 avg, 0 capped</span></td><td>incorrect=30</td></tr></tbody>
+        </table>
+      </div>
+      <h3>Positive Deterministic Matches In Strict30</h3>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>run</th><th>record</th><th>source</th><th>title</th><th>score</th><th>matched units</th></tr></thead>
+          <tbody><tr><td>microsoft__Phi-4-mini-instruct__strict30</td><td>iol-2006-individual-p1-sub-assignment_1</td><td>2006-individual-1</td><td>Lakhota</td><td>1/4</td><td>assignment_1.4</td></tr>
+<tr><td>mistralai__Mistral-7B-Instruct-v0.3__strict30</td><td>iol-2007-individual-p5</td><td>2007-individual-5</td><td>Turkish and Tatar</td><td>1/4</td><td>full.4</td></tr>
+<tr><td>meta-llama__Llama-3.2-3B-Instruct__strict30</td><td>iol-2009-individual-p4-sub-a</td><td>2009-individual-4</td><td>Old Indic</td><td>1/1</td><td>a.1</td></tr>
+<tr><td>meta-llama__Llama-3.2-3B-Instruct__strict30</td><td>iol-2010-team-p1</td><td>2010-team-1</td><td>Mongolian</td><td>1/1</td><td>full.1</td></tr>
+<tr><td>microsoft__Phi-4-mini-instruct__strict30</td><td>iol-2012-individual-p1-sub-a</td><td>2012-individual-1</td><td>Dyirbal</td><td>1/1</td><td>a.1</td></tr>
+<tr><td>meta-llama__Llama-3.2-3B-Instruct__strict30</td><td>iol-2016-individual-p1</td><td>2016-individual-1</td><td>Aralle-Tabulahan</td><td>15/15</td><td>full.1, full.2, full.3, full.4, full.5, full.6, full.7, full.8, full.9, full.10, full.11, full.12, full.13, full.14, full.15</td></tr>
+<tr><td>Qwen__Qwen3-4B-Thinking-2507__strict30</td><td>iol-2017-individual-p2-sub-a</td><td>2017-individual-2</td><td>Abui</td><td>2/16</td><td>a.11, a.15</td></tr>
+<tr><td>Qwen__Qwen3-4B-Instruct-2507__strict30</td><td>iol-2021-individual-p3-sub-a</td><td>2021-individual-3</td><td>Kilivila</td><td>1/4</td><td>a.1</td></tr>
+<tr><td>Qwen__Qwen3-4B-Thinking-2507__strict30</td><td>iol-2021-individual-p3-sub-a</td><td>2021-individual-3</td><td>Kilivila</td><td>4/4</td><td>a.1, a.2, a.3, a.4</td></tr>
+<tr><td>mistralai__Mistral-7B-Instruct-v0.3__strict30</td><td>iol-2021-individual-p3-sub-a</td><td>2021-individual-3</td><td>Kilivila</td><td>4/4</td><td>a.1, a.2, a.3, a.4</td></tr>
+<tr><td>Qwen__Qwen3-4B-Instruct-2507__strict30</td><td>iol-2024-individual-p1</td><td>2024-individual-1</td><td>Koryak</td><td>1/17</td><td>full.5</td></tr>
+<tr><td>Qwen__Qwen3-4B-Thinking-2507__strict30</td><td>iol-2024-individual-p1</td><td>2024-individual-1</td><td>Koryak</td><td>1/17</td><td>full.5</td></tr></tbody>
+        </table>
+      </div>
+    </section>
+
+    <section>
+      <h2>Other Local Eval Runs</h2>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>run</th><th>model</th><th>n</th><th>score</th><th>fraction</th><th>sources</th><th>generated tokens</th><th>verdicts</th></tr></thead>
+          <tbody><tr><td class="run">HuggingFaceTB__SmolLM3-3B__diverse5</td><td>HuggingFaceTB/SmolLM3-3B</td><td>5</td><td>0/24</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>5</td><td>19842 total<br><span class="muted">3968.4 avg, 2 capped</span></td><td>incorrect=5</td></tr>
+<tr><td class="run">Qwen__Qwen2.5-3B-Instruct__diverse5</td><td>Qwen/Qwen2.5-3B-Instruct</td><td>5</td><td>0/24</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>5</td><td>18402 total<br><span class="muted">3680.4 avg, 2 capped</span></td><td>incorrect=5</td></tr>
+<tr><td class="run">Qwen__Qwen3-4B-Instruct-2507</td><td>Qwen/Qwen3-4B-Instruct-2507</td><td>3</td><td>0/14</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>1</td><td>18526 total<br><span class="muted">6175.3 avg, 2 capped</span></td><td>incorrect=3</td></tr>
+<tr><td class="run">Qwen__Qwen3-4B-Instruct-2507__diverse5</td><td>Qwen/Qwen3-4B-Instruct-2507</td><td>5</td><td>1/24</td><td>4.2%<div class="bar"><span style="width:4.17%"></span></div></td><td>5</td><td>38530 total<br><span class="muted">7706.0 avg, 4 capped</span></td><td>incorrect=4, partial=1</td></tr>
+<tr><td class="run">Qwen__Qwen3-4B-Thinking-2507__diverse5</td><td>Qwen/Qwen3-4B-Thinking-2507</td><td>5</td><td>1/24</td><td>4.2%<div class="bar"><span style="width:4.17%"></span></div></td><td>5</td><td>40960 total<br><span class="muted">8192.0 avg, 5 capped</span></td><td>incorrect=4, partial=1</td></tr>
+<tr><td class="run">google__gemma-3-12b-it__diverse5</td><td>google/gemma-3-12b-it</td><td>5</td><td>0/24</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>5</td><td>4271 total<br><span class="muted">854.2 avg, 0 capped</span></td><td>incorrect=5</td></tr>
+<tr><td class="run">google__gemma-3-4b-it__diverse5</td><td>google/gemma-3-4b-it</td><td>5</td><td>0/24</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>5</td><td>17208 total<br><span class="muted">3441.6 avg, 2 capped</span></td><td>incorrect=5</td></tr>
+<tr><td class="run">meta-llama__Llama-3.2-3B-Instruct__diverse5</td><td>meta-llama/Llama-3.2-3B-Instruct</td><td>5</td><td>1/24</td><td>4.2%<div class="bar"><span style="width:4.17%"></span></div></td><td>5</td><td>14943 total<br><span class="muted">2988.6 avg, 1 capped</span></td><td>incorrect=4, partial=1</td></tr>
+<tr><td class="run">microsoft__Phi-4-mini-instruct</td><td>microsoft/Phi-4-mini-instruct</td><td>3</td><td>0/14</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>1</td><td>16537 total<br><span class="muted">5512.3 avg, 2 capped</span></td><td>incorrect=3</td></tr>
+<tr><td class="run">microsoft__Phi-4-mini-instruct__diverse5</td><td>microsoft/Phi-4-mini-instruct</td><td>5</td><td>0/24</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>5</td><td>16821 total<br><span class="muted">3364.2 avg, 2 capped</span></td><td>incorrect=5</td></tr>
+<tr><td class="run">mistralai__Mistral-7B-Instruct-v0.3__diverse5</td><td>mistralai/Mistral-7B-Instruct-v0.3</td><td>5</td><td>0/24</td><td>0.0%<div class="bar"><span style="width:0.00%"></span></div></td><td>5</td><td>2020 total<br><span class="muted">404.0 avg, 0 capped</span></td><td>incorrect=5</td></tr></tbody>
+        </table>
+      </div>
+      <div class="muted">Startup attempts without score files: 2.</div>
+    </section>
+
+    <section>
+      <h2>Coverage Gaps</h2>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>source problem</th><th>title</th><th>missing reason</th><th>notes</th></tr></thead>
+          <tbody><tr><td>2004-team-1</td><td>Armenian</td><td>2004-team-solution</td><td>Excluded from strict eval dataset because no official English solution-backed verifier source was selected.</td></tr>
+<tr><td>2005-team-1</td><td>Figuig</td><td>2005-team-solution</td><td>Excluded from strict eval dataset because no official English solution-backed verifier source was selected.</td></tr>
+<tr><td>2006-team-1</td><td>American Sign Language</td><td>2006-team-problem, 2006-team-solution</td><td>Excluded from strict eval dataset because no official English solution-backed verifier source was selected.</td></tr>
+<tr><td>2013-team-1</td><td>Georgian</td><td>2013-team-solution</td><td>Excluded from strict eval dataset because no official English solution-backed verifier source was selected.</td></tr></tbody>
+        </table>
+      </div>
+    </section>
+
+    <section>
+      <h2>Training Readiness</h2>
+      <div class="callout">The broad strict split is still not suitable as a blanket RL or SFT label source. The repository also exports a smaller manually PDF-checked RL subset whose prompts, answer units, aliases, and final-answer schema are designed for conservative deterministic rewards.</div>
+      <p>The manually curated deterministic RL-candidate export contains <code>15</code> rows. Prepared OpenRLHF data under <code>data/rl/ioling_qwen3_4b_manual_v1</code> contains <code>9</code> train rows and <code>6</code> validation rows using <code>strict_keyed_units_v5</code>. See <a href="rl_manual_pdf_audit.md">the PDF audit</a>, <a href="rl_validation_zero_diagnosis.md">the validation diagnosis</a>, and <a href="../docs/scoring.md">the scorer definition</a>.</p>
+    </section>
+
+    <section>
+      <h2>RL Training Runs</h2>
+      <div class="callout">These runs use manually checked splits and the strict keyed final boxed answer-unit reward. The latest two-step run found 10/256 exact train samples at both steps, but source-disjoint validation pass@2 stayed zero. Final-answer reward does not certify the reasoning trace.</div>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>run</th><th>status</th><th>steps</th><th>train/val</th><th>samples/step</th><th>best train reward</th><th>last train reward</th><th>eval pass@1</th><th>train exact</th><th>train unit acc</th><th>format valid</th><th>train trunc</th><th>eval trunc</th><th>wandb</th></tr></thead>
+          <tbody><tr><td class="run">distill-v1-strict-ep2-g8</td><td>failed_before_update</td><td>0</td><td>38/24</td><td>0</td><td>0</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td><a href="https://wandb.ai/alexgurung/ioling-rl/runs/b02jotne">run</a></td></tr>
+<tr><td class="run">distill-v1-strict-ep2-g8-retry1</td><td>completed_with_cleanup_warning</td><td>2</td><td>38/24</td><td>256</td><td>0.039</td><td>0.039</td><td>0</td><td>3.9%</td><td>3.9%</td><td>97.7%</td><td>0.0%</td><td>0.0%</td><td><a href="https://wandb.ai/alexgurung/ioling-rl/runs/lcy6fkw8">run</a></td></tr>
+<tr><td class="run">qwen3-4b-ioling-manual-v1-lr5e-7-t1-g8-b32-ep2-4k</td><td>failed_before_update</td><td>0</td><td>11/5</td><td>0</td><td>0</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td><a href="https://wandb.ai/alexgurung/ioling-rl/runs/lfj1as6k">run</a></td></tr>
+<tr><td class="run">qwen3-4b-ioling-manual-v1-lr5e-7-t1-g8-b32-ep2-4k-retry</td><td>no_complete_ppo_update</td><td>0</td><td>11/5</td><td>0</td><td>0</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td><a href="https://wandb.ai/alexgurung/ioling-rl/runs/igj2qwjk">run</a></td></tr>
+<tr><td class="run">qwen3-4b-ioling-manual-v1-lr5e-7-t1-g8-b8-ep2-4k</td><td>completed_with_cleanup_warning</td><td>2</td><td>11/5</td><td>64</td><td>0.018</td><td>0.018</td><td>0</td><td>0.0%</td><td>1.8%</td><td>18.8%</td><td>81.2%</td><td>90.0%</td><td><a href="https://wandb.ai/alexgurung/ioling-rl/runs/s0b7a5z6">run</a></td></tr>
+<tr><td class="run">qwen3-4b-ioling-manual-v1-lr5e-7-t1-g8-b8-ep2-8k</td><td>completed_with_cleanup_warning</td><td>2</td><td>11/5</td><td>64</td><td>0.039</td><td>0.034</td><td>0</td><td>0.0%</td><td>3.4%</td><td>78.1%</td><td>17.2%</td><td>20.0%</td><td><a href="https://wandb.ai/alexgurung/ioling-rl/runs/efu9wtf1">run</a></td></tr></tbody>
+        </table>
+      </div>
+    </section>
+
+    <section>
+      <h2>RL Strict Validation Probes</h2>
+      <div class="callout">These probes run sampled vLLM inference on the fixed manual-v1 validation split and score with the same strict boxed-answer reward path used by OpenRLHF. They are intended to answer whether base models can solve any held-out records after repeated sampling.</div>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>model</th><th>n</th><th>samples</th><th>records positive</th><th>records exact</th><th>sample positive</th><th>sample exact</th><th>mean score</th><th>format valid</th><th>truncation</th><th>file</th></tr></thead>
+          <tbody><tr><td>Qwen/Qwen3.6-35B-A3B-FP8</td><td>8</td><td>192</td><td>1</td><td>1</td><td>2.1%</td><td>2.1%</td><td>0.021</td><td>34.4%</td><td>53.1%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3.6-35B-A3B-FP8__manual_v9_atomic_prior_nothink_finalonly_8x4k.json</td></tr>
+<tr><td>google/gemma-3-12b-it</td><td>8</td><td>48</td><td>1</td><td>0</td><td>2.1%</td><td>0.0%</td><td>0.004</td><td>64.6%</td><td>0.0%</td><td class="run">reports/rl_strict_probes/google_gemma-3-12b-it__val8x8192_fixed.json</td></tr>
+<tr><td>google/gemma-3-12b-it</td><td>32</td><td>192</td><td>1</td><td>0</td><td>1.6%</td><td>0.0%</td><td>0.003</td><td>65.1%</td><td>0.0%</td><td class="run">reports/rl_strict_probes/google_gemma-3-12b-it__val32x8192_fixed.json</td></tr>
+<tr><td>mistralai/Mistral-7B-Instruct-v0.3</td><td>8</td><td>48</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>50.0%</td><td>0.0%</td><td class="run">reports/rl_strict_probes/mistralai_Mistral-7B-Instruct-v0.3__val8x8192_fixed.json</td></tr>
+<tr><td>microsoft/Phi-4-mini-instruct</td><td>8</td><td>48</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>58.3%</td><td>4.2%</td><td class="run">reports/rl_strict_probes/microsoft_Phi-4-mini-instruct__val8x8192_fixed.json</td></tr>
+<tr><td>google/gemma-3-4b-it</td><td>8</td><td>48</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>0.0%</td><td>0.0%</td><td class="run">reports/rl_strict_probes/google_gemma-3-4b-it__val8x8192_fixed.json</td></tr>
+<tr><td>Qwen/Qwen3.6-35B-A3B-FP8</td><td>8</td><td>192</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>0.5%</td><td>98.4%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3.6-35B-A3B-FP8__manual_v9_atomic_prior_8x4k.json</td></tr>
+<tr><td>Qwen/Qwen3.6-35B-A3B-FP8</td><td>8</td><td>192</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>0.0%</td><td>100.0%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3.6-35B-A3B-FP8__manual_v9_atomic_prior_nothink_8x4k.json</td></tr>
+<tr><td>Qwen/Qwen3-4B-Thinking-2507</td><td>8</td><td>48</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>0.0%</td><td>100.0%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3-4B-Thinking-2507__val8x8192_fixed.json</td></tr>
+<tr><td>Qwen/Qwen3-4B-Instruct-2507</td><td>8</td><td>192</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>79.7%</td><td>0.0%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3-4B-Instruct-2507__manual_v8_atomic_val_8x4k.json</td></tr>
+<tr><td>Qwen/Qwen3-4B-Instruct-2507</td><td>8</td><td>192</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>7.8%</td><td>92.2%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3-4B-Instruct-2507__manual_v9_atomic_prior_antiloop_val_8x4k.json</td></tr>
+<tr><td>Qwen/Qwen3-4B-Instruct-2507</td><td>8</td><td>192</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>4.7%</td><td>95.3%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3-4B-Instruct-2507__manual_v9_atomic_prior_reasoning_val_8x4k.json</td></tr>
+<tr><td>Qwen/Qwen3-4B-Instruct-2507</td><td>8</td><td>192</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>87.0%</td><td>0.0%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3-4B-Instruct-2507__manual_v9_atomic_with_prior_val_8x4k.json</td></tr>
+<tr><td>Qwen/Qwen3-4B-Instruct-2507</td><td>8</td><td>48</td><td>0</td><td>0</td><td>0.0%</td><td>0.0%</td><td>0</td><td>66.7%</td><td>27.1%</td><td class="run">reports/rl_strict_probes/Qwen_Qwen3-4B-Instruct-2507__val8x8192_fixed.json</td></tr></tbody>
+        </table>
+      </div>
+      <p>Detailed positives and boxed outputs: <a href="rl_strict_probe_summary.md">RL strict probe summary</a>. Manual qualitative read: <a href="rl_probe_qualitative_error_analysis.md">qualitative error analysis</a>.</p>
+      <p>Relaxed-format and near-miss audit: <a href="rl_false_negative_audit.md">diagnostic audit</a> and <a href="rl_false_negative_manual_review.md">manual review</a>.</p>
+    </section>
+
+    <section>
+      <h2>Hint/Distillation Probes</h2>
+      <div class="callout">These probes test whether generic hints, redacted solution hints, Groot-style planner diversity, or answer-conditioned scaffolds can create positive samples with a 4B model under the strict keyed reward. The current fresh split is <code>data/rl/ioling_qwen3_4b_manual_v7_concise</code>.</div>
+      <p>Human-readable synthesis: <a href="ioling_hint_diversity_summary.md">hint/diversity investigation summary</a>. Answer-given rationale and atomic curriculum audit: <a href="answer_given_atomic_audit.md">manual audit and 4k probes</a>.</p>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>run</th><th>data</th><th>splits</th><th>samples</th><th>positive</th><th>exact</th><th>best split</th><th>best method</th><th>sample +</th><th>sample exact</th><th>format</th><th>trunc</th><th>report</th></tr></thead>
+          <tbody><tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v1_val_derivation_scaffold_16x_4k</td><td>ioling_qwen3_4b_manual_v1</td><td>val</td><td>32</td><td>0</td><td>0</td><td>val</td><td>derivation_scaffold_hint</td><td>0.0%</td><td>0.0%</td><td>0.0%</td><td>100.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v1_val_derivation_scaffold_16x_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_concise_hints_groot_4k</td><td>ioling_qwen3_4b_manual_v6_concise</td><td>train, val</td><td>656</td><td>1</td><td>0</td><td>train</td><td>groot_1x4</td><td>2.3%</td><td>0.0%</td><td>100.0%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_concise_hints_groot_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_component_scaffold_16x_4k</td><td>ioling_qwen3_4b_manual_v6_concise</td><td>val</td><td>32</td><td>16</td><td>15</td><td>val</td><td>derivation_scaffold_hint</td><td>50.0%</td><td>46.9%</td><td>78.1%</td><td>21.9%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_component_scaffold_16x_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_component_scaffold_v2_16x_4k</td><td>ioling_qwen3_4b_manual_v6_concise</td><td>val</td><td>32</td><td>32</td><td>31</td><td>val</td><td>derivation_scaffold_hint</td><td>100.0%</td><td>96.9%</td><td>100.0%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_component_scaffold_v2_16x_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_derivation_scaffold_16x_4k</td><td>ioling_qwen3_4b_manual_v6_concise</td><td>val</td><td>32</td><td>0</td><td>0</td><td>val</td><td>derivation_scaffold_hint</td><td>0.0%</td><td>0.0%</td><td>100.0%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_derivation_scaffold_16x_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v7_concise_hints_groot_4k</td><td>ioling_qwen3_4b_manual_v7_concise</td><td>train, val</td><td>345</td><td>1</td><td>0</td><td>train</td><td>groot_1x4</td><td>2.8%</td><td>0.0%</td><td>100.0%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v7_concise_hints_groot_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v7_val_component_scaffold_v2_16x_4k</td><td>ioling_qwen3_4b_manual_v7_concise</td><td>val</td><td>32</td><td>30</td><td>29</td><td>val</td><td>derivation_scaffold_hint</td><td>93.8%</td><td>90.6%</td><td>93.8%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v7_val_component_scaffold_v2_16x_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_trainval_hints_groot_4k</td><td>ioling_qwen3_4b_manual_v1</td><td>train, val</td><td>375</td><td>9</td><td>3</td><td>train</td><td>redacted_solution_hint_mask0</td><td>100.0%</td><td>75.0%</td><td>100.0%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_trainval_hints_groot_4k.md">md</a></td></tr>
+<tr><td class="run">Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_v6_val_manual_rule_cards_16x_4k</td><td>ioling_qwen3_4b_manual_v6_concise</td><td>val</td><td>80</td><td>0</td><td>0</td><td>val</td><td>manual_rule_hint</td><td>0.0%</td><td>0.0%</td><td>100.0%</td><td>0.0%</td><td><a href="ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_v6_val_manual_rule_cards_16x_4k.md">md</a></td></tr></tbody>
+        </table>
+      </div>
+    </section>
+
+    <section>
+      <h2>RL Base Probes</h2>
+      <div class="callout">These are historical local probes from the manual-v1 development loop. They use the same vLLM chat-template path as OpenRLHF, but were run before the final prompt-furniture cleanup and scorer hardening. Positive means at least one exact answer unit in a sample; exact means every required unit matched. A semantic paraphrase is not counted unless manually registered as an alias.</div>
+      <div class="table-wrap">
+        <table>
+          <thead><tr><th>report</th><th>max tokens</th><th>samples/prompt</th><th>records</th><th>records with positive</th><th>records exact</th><th>format valid</th><th>truncated</th><th>mean sample score</th></tr></thead>
+          <tbody><tr><td>rl_base_probe_qwen3_4b_manual_v1.json</td><td>4096</td><td>8</td><td>16</td><td>1</td><td>0</td><td>7.8%</td><td>89.1%</td><td>0.005</td></tr>
+<tr><td>rl_base_probe_qwen3_4b_manual_v1_bounded.json</td><td>4096</td><td>8</td><td>16</td><td>2</td><td>0</td><td>2.3%</td><td>93.8%</td><td>0.004</td></tr>
+<tr><td>rl_base_probe_qwen3_4b_manual_v1_chat4096.json</td><td>4096</td><td>4</td><td>16</td><td>1</td><td>0</td><td>15.6%</td><td>84.4%</td><td>0.005</td></tr>
+<tr><td>rl_base_probe_qwen3_4b_manual_v1_chat8192.json</td><td>8192</td><td>2</td><td>16</td><td>2</td><td>0</td><td>75.0%</td><td>25.0%</td><td>0.026</td></tr>
+<tr><td>rl_base_probe_qwen3_4b_manual_v6_concise4096.json</td><td>4096</td><td>4</td><td>16</td><td>0</td><td>0</td><td>89.1%</td><td>0.0%</td><td>0</td></tr></tbody>
+        </table>
+      </div>
+    </section>
+  </main>
+</body>
+</html>

benchmark/IOL/ioling_hf/reports/ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v1_val_derivation_scaffold_16x_4k.md

@@ -0,0 +1,17 @@
+# IOLing Hint Diversity Probe
+
+- generated: `2026-07-10T23:30:48.034587+00:00`
+- model: `Qwen/Qwen3-4B-Instruct-2507`
+- samples: `32`
+- positive samples: `0`
+- exact samples: `0`
+
+## Method Summary
+
+| split | method | records | samples | records + | records exact | sample + | sample exact | mean best | format | unique boxes |
+| --- | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| val | derivation_scaffold_hint | 2 | 32 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.000 | 1 |
+
+## Positive Samples
+
+No positive samples found.

benchmark/IOL/ioling_hf/reports/ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_concise_hints_groot_4k.md

@@ -0,0 +1,39 @@
+# IOLing Hint Diversity Probe
+
+- generated: `2026-07-10T23:23:58.279994+00:00`
+- model: `Qwen/Qwen3-4B-Instruct-2507`
+- samples: `656`
+- positive samples: `1`
+- exact samples: `0`
+
+## Method Summary
+
+| split | method | records | samples | records + | records exact | sample + | sample exact | mean best | format | unique boxes |
+| --- | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| train | anti_nearmiss | 11 | 88 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 69 |
+| train | groot_1x4 | 11 | 44 | 1 | 0 | 0.023 | 0.000 | 0.015 | 1.000 | 43 |
+| train | groot_anchor_iid | 11 | 11 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.909 | 11 |
+| train | iid | 11 | 88 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.920 | 67 |
+| train | redacted_solution_hint | 11 | 44 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 40 |
+| train | solver_checklist | 11 | 88 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 72 |
+| train | unit_audit | 11 | 88 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.966 | 67 |
+| val | anti_nearmiss | 5 | 40 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 35 |
+| val | groot_1x4 | 5 | 20 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 16 |
+| val | groot_anchor_iid | 5 | 5 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.800 | 5 |
+| val | iid | 5 | 40 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.775 | 29 |
+| val | redacted_solution_hint | 5 | 20 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 18 |
+| val | solver_checklist | 5 | 40 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.975 | 31 |
+| val | unit_audit | 5 | 40 | 0 | 0 | 0.000 | 0.000 | 0.000 | 0.950 | 32 |
+
+## Positive Samples
+
+### `train` `iol-2008-individual-p5-sub-a` `groot_1x4` sample `3`
+
+- score: `0.16666666666666666`
+- matched: `a.5`
+- missed: `a.1, a.2, a.3, a.4, a.6`
+
+```text
+a.1: The hunter came to you; a.2: Your boy cured a shaman; a.3: The hunter cured you; a.4: The dog shot the teacher; a.5: You came; a.6: The hunter cured a doctor
+```
+

benchmark/IOL/ioling_hf/reports/ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_manual_v6_val_component_scaffold_16x_4k.md

@@ -0,0 +1,176 @@
+# IOLing Hint Diversity Probe
+
+- generated: `2026-07-10T23:33:22.033676+00:00`
+- model: `Qwen/Qwen3-4B-Instruct-2507`
+- samples: `32`
+- positive samples: `16`
+- exact samples: `15`
+
+## Method Summary
+
+| split | method | records | samples | records + | records exact | sample + | sample exact | mean best | format | unique boxes |
+| --- | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| val | derivation_scaffold_hint | 2 | 32 | 1 | 1 | 0.500 | 0.469 | 0.500 | 0.781 | 6 |
+
+## Positive Samples
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `0`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `1`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `2`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `3`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `4`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `5`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `6`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `7`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `8`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `9`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `10`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `11`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `12`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `14`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `15`
+
+- score: `1.0`
+- matched: `b.1, b.2`
+- missed: ``
+
+```text
+b.1: öm-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+
+### `val` `iol-2015-individual-p1-sub-b` `derivation_scaffold_hint` sample `13`
+
+- score: `0.5`
+- matched: `b.2`
+- missed: `b.1`
+
+```text
+b.1: om-pöhualli-om-öme; b.2: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui
+```
+

benchmark/IOL/ioling_hf/reports/ioling_hint_diversity/Qwen_Qwen3-4B-Instruct-2507__qwen3_4b_v6_val_manual_rule_cards_16x_4k.md

@@ -0,0 +1,17 @@
+# IOLing Hint Diversity Probe
+
+- generated: `2026-07-10T23:27:21.299086+00:00`
+- model: `Qwen/Qwen3-4B-Instruct-2507`
+- samples: `80`
+- positive samples: `0`
+- exact samples: `0`
+
+## Method Summary
+
+| split | method | records | samples | records + | records exact | sample + | sample exact | mean best | format | unique boxes |
+| --- | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| val | manual_rule_hint | 5 | 80 | 0 | 0 | 0.000 | 0.000 | 0.000 | 1.000 | 49 |
+
+## Positive Samples
+
+No positive samples found.

benchmark/IOL/ioling_hf/reports/manual_verification/README.md

@@ -0,0 +1,9 @@
+# Manual Verification
+
+Each batch is a disjoint source-problem review manifest. Review source problems, not only rows: one verified source problem covers all derived full/subproblem/partial records for that source.
+
+Verdict schema for reviewers:
+
+```json
+{"source_problem_id":"YYYY-round-N","verdict":"pass|fix_needed|uncertain","record_ids_checked":["..."],"issues":[{"severity":"critical|high|medium|low","record_id":"optional","field":"prompt|canonical|answer_units|provenance|subparts","description":"...","suggested_fix":"..."}],"notes":"..."}
+```

benchmark/IOL/ioling_hf/reports/manual_verification/agent_batch_01_findings.jsonl

@@ -0,0 +1,22 @@
+{"source_problem_id":"2003-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2003-individual-p1","iol-2003-individual-p1-sub-assignment_1","iol-2003-individual-p1-sub-assignment_2","iol-2003-individual-p1-partial-through-assignment_1-target-assignment_2"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The rendered problem PDF uses special mathematical/fraction glyphs for the artificial language, but the dataset OCR replaces them with ordinary letters and punctuation, so the examples and assignments are not faithful or reliably answerable from text.","suggested_fix":"Re-extract this problem from rendered pages or encode the glyph inventory explicitly so all examples and assignment items match the source."},{"severity":"critical","record_id":null,"field":"answer_units","description":"Official answers for assignment items 9, 13, and 15 are missing from answer_units; several remaining formula answers are OCR-corrupted.","suggested_fix":"Regenerate answer_units from the rendered official solution with one required unit for each target item, including 9, 13, and 15."},{"severity":"high","record_id":null,"field":"canonical","description":"The canonical solution text corrupts the source glyphs throughout the noun/verb rules and Assignment 2 formulas.","suggested_fix":"Replace canonical solution_text with a faithful transcription or rendered-safe notation for the source symbols."}],"notes":"Problem and solution pages are image-only by text extraction, but rendered pages make the OCR/glyph failures clear. Provenance page range points to the right source pages."}
+{"source_problem_id":"2003-team-2","verdict":"fix_needed","record_ids_checked":["iol-2003-team-p2","iol-2003-team-p2-sub-a","iol-2003-team-p2-sub-b","iol-2003-team-p2-sub-c","iol-2003-team-p2-partial-through-a-target-b","iol-2003-team-p2-partial-through-ab-target-c"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The source relies on subscripted variables and asterisk notation, but the dataset flattens these into semicolons and stray punctuation, changing the linguistic data.","suggested_fix":"Rebuild the prompt from a rendered transcription that preserves subscripts, asterisks, and sentence pair structure."},{"severity":"critical","record_id":null,"field":"answer_units","description":"Assignment 1 explanation is omitted from answer_units, while the a/b/c solved forms contain corrupted subscript alternatives; the c unit also includes a footer.","suggested_fix":"Create faithful units for the Assignment 1 rule explanation and each requested translation, without page headers or footers."},{"severity":"high","record_id":null,"field":"canonical","description":"The canonical solution has the same subscript/OCR corruption as the prompt, so partial variants show unreliable solved forms.","suggested_fix":"Replace solution_text and shown partial solutions with a notation-preserving transcription."}],"notes":"The local rendered problem and solution pages are the cited pages; the extraction is not text-faithful enough for this source."}
+{"source_problem_id":"2004-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2004-individual-p5"],"issues":[{"severity":"critical","record_id":"iol-2004-individual-p5","field":"prompt","description":"Chuvash letters and diacritics from the source are replaced by font-encoding artifacts such as commas, ampersands, apostrophes, and parentheses, which changes the forms needed to solve the problem.","suggested_fix":"Re-extract or manually transcribe the Chuvash data with the correct Unicode letters and diacritics."},{"severity":"critical","record_id":"iol-2004-individual-p5","field":"answer_units","description":"The answer_units contain only two suffix-rule statements and omit the completed table/gap answers required by the official solution.","suggested_fix":"Add granular answer_units for each requested filled form and keep rule statements separate from final answers."},{"severity":"high","record_id":"iol-2004-individual-p5","field":"canonical","description":"The canonical solution loses much of the completed answer table and ends with garbled symbol strings.","suggested_fix":"Replace the canonical answer with a clean transcription of the official solution pages."}],"notes":"The cited page range is correct, but the prompt and solution extraction are not faithful enough for scoring."}
+{"source_problem_id":"2006-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2006-individual-p1","iol-2006-individual-p1-sub-assignment_1","iol-2006-individual-p1-sub-assignment_2","iol-2006-individual-p1-sub-assignment_3","iol-2006-individual-p1-partial-through-assignment_1-target-assignment_2","iol-2006-individual-p1-partial-through-assignment_1assignment_2-target-assignment_3"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Some answer units are not granular or complete enough for RL scoring: assignment_1.3 is truncated after 'we can't', and assignment_2.1 combines two possible translations in one required unit.","suggested_fix":"Split alternatives into separate acceptable values and restore the missing continuation of the assignment_1.3 explanation."},{"severity":"low","record_id":null,"field":"prompt","description":"The prompt includes a stray page/footer line from the next page, though the problem text itself is otherwise faithful.","suggested_fix":"Trim page headers and footers from the extracted prompt."}],"notes":"Prompt, solution text, subproblem labels, and partial variants otherwise match the cited source pages."}
+{"source_problem_id":"2007-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2007-individual-p2"],"issues":[{"severity":"high","record_id":"iol-2007-individual-p2","field":"prompt","description":"Movima consonant symbols from the source are encoded as ASCII punctuation in the prompt and notes, making examples such as the special c/l/n letters visually non-faithful.","suggested_fix":"Use correct Unicode characters or an explicit normalized notation for the Movima symbols in the prompt."},{"severity":"medium","record_id":"iol-2007-individual-p2","field":"canonical","description":"The canonical solution uses the same degraded symbol encoding, so the official forms are not faithful to the PDF.","suggested_fix":"Correct the solution transcription to use the same notation as the repaired prompt."},{"severity":"medium","record_id":"iol-2007-individual-p2","field":"answer_units","description":"The single full.1 answer unit contains the whole solution and a footer, which is too coarse for scoring individual targets.","suggested_fix":"Split the final answers and rule explanations into separate units and remove page footer text."}],"notes":"The source is answerable in the PDF, but the dataset text relies on lossy character substitutions."}
+{"source_problem_id":"2008-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2008-individual-p2","iol-2008-individual-p2-sub-b"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"The two-column verse and word-list layout is interleaved in plain text, and the boxed answer slots are lost, making part b hard to reconstruct from the dataset prompt.","suggested_fix":"Recreate the stanza and word list with stable line/table structure and explicit blank labels."},{"severity":"critical","record_id":null,"field":"answer_units","description":"The part b answer unit does not contain the a-m gap assignments from the official solution; it mostly records leftover words and includes a footer. Full-problem units also omit useful part a scoring units.","suggested_fix":"Add one unit per filled blank or translation target and separate part a rules/answers from part b final assignments."},{"severity":"medium","record_id":null,"field":"subparts","description":"The source has multiple parts, but the dataset only exposes a sub-b variant and no usable sub-a record.","suggested_fix":"Generate subproblem records for each source subpart with matching target labels."}],"notes":"Rendered PDF confirms the source layout is much clearer than the extracted dataset text."}
+{"source_problem_id":"2009-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2009-individual-p2","iol-2009-individual-p2-sub-a","iol-2009-individual-p2-sub-b","iol-2009-individual-p2-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The N'Ko-script column present in the source problem is omitted from the dataset prompt, and the script note is collapsed into garbled run-on text.","suggested_fix":"Rebuild the prompt from rendered pages so both the N'Ko and Roman-script columns and the explanatory note are preserved."},{"severity":"high","record_id":null,"field":"canonical","description":"The official solution text does not faithfully represent the script-side answers from the PDF, so solved forms are incomplete in canonical and partial records.","suggested_fix":"Transcribe the official solution with the N'Ko forms or an explicit source-faithful notation."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Answer units are coarse and one part b unit includes footer text.","suggested_fix":"Split the official answers into one unit per requested item and remove headers/footers."}],"notes":"This source is not answerable as text because a whole source-script column is missing."}
+{"source_problem_id":"2010-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2010-individual-p2","iol-2010-individual-p2-sub-a","iol-2010-individual-p2-sub-b","iol-2010-individual-p2-sub-c","iol-2010-individual-p2-partial-through-a-target-b","iol-2010-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Answer units are too coarse for RL scoring: a.1 bundles many correspondences, b.1 bundles multiple equations, and c.1 bundles all Drehu forms plus footer text.","suggested_fix":"Split answer_units into one required unit per correspondence, equation, or translation target, and remove footer text."}],"notes":"Prompt, canonical solution, labels, partial-variant shown solutions, and provenance otherwise match the cited PDF pages."}
+{"source_problem_id":"2011-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2011-individual-p2","iol-2011-individual-p2-sub-a"],"issues":[{"severity":"critical","record_id":null,"field":"answer_units","description":"The answer_units capture only two rule lines and omit the seven actual gap transcriptions for part a; there is no record or unit coverage for part b.","suggested_fix":"Regenerate units from the official solution with one answer per gap for part a and add the missing part b target coverage."},{"severity":"medium","record_id":null,"field":"canonical","description":"The side-by-side official solution layout is interleaved in extraction, mixing part a answers and part b rules.","suggested_fix":"Clean the canonical solution into separate part a and part b sections."},{"severity":"medium","record_id":null,"field":"subparts","description":"The source has more than one part, but the dataset only includes the full problem and sub-a record.","suggested_fix":"Generate subproblem and partial records that match all source subpart labels."}],"notes":"The prompt text itself is mostly faithful aside from normal header/footer extraction."}
+{"source_problem_id":"2012-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2012-individual-p2","iol-2012-individual-p2-sub-a","iol-2012-individual-p2-sub-b","iol-2012-individual-p2-partial-through-a-target-b"],"issues":[{"severity":"medium","record_id":null,"field":"prompt","description":"A boxed comparison in the source is collapsed into the surrounding numeral list, weakening the table/layout fidelity needed to read the data.","suggested_fix":"Represent the tables and boxed comparison with explicit table markup or clear separated lines."},{"severity":"medium","record_id":null,"field":"answer_units","description":"The part a and part b answer units each bundle multiple final answers, which is too coarse for item-level scoring.","suggested_fix":"Split units so each requested numeral or expression has its own required answer unit."}],"notes":"The main text and page provenance are otherwise consistent with the PDF."}
+{"source_problem_id":"2013-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2013-individual-p2","iol-2013-individual-p2-sub-a","iol-2013-individual-p2-sub-b","iol-2013-individual-p2-sub-c","iol-2013-individual-p2-partial-through-a-target-b","iol-2013-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"The answer units are too coarse for RL scoring: part a, b, and c each bundle several final answers into one unit, and c includes footer text.","suggested_fix":"Split the official answers into one unit per requested item and strip page footer text."}],"notes":"Prompt, canonical answer, subpart labels, partial variants, and provenance are faithful; the fix is mainly scoring granularity."}
+{"source_problem_id":"2014-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2014-individual-p3","iol-2014-individual-p3-sub-a","iol-2014-individual-p3-sub-b","iol-2014-individual-p3-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The source problem depends on a family-tree diagram, but the dataset prompt reduces the diagram to scattered punctuation and letters, losing the relationships needed to solve the task.","suggested_fix":"Include the rendered diagram as an image asset or transcribe the family tree into explicit relationship edges."},{"severity":"critical","record_id":null,"field":"canonical","description":"The official solution family tree is also collapsed; the part a answer unit is effectively just '(a) 1' rather than the reconstructed tree.","suggested_fix":"Rebuild the canonical and part a answer_units from the rendered solution diagram."},{"severity":"high","record_id":"iol-2014-individual-p3-partial-through-a-target-b","field":"subparts","description":"The partial variant shows solved part a as an unusable collapsed tree, so the visible solved context is not faithful.","suggested_fix":"Use a diagram image or explicit relationship transcription for shown solved part a."}],"notes":"This is a diagram-dependent source; text extraction alone is insufficient."}
+{"source_problem_id":"2015-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2015-individual-p3","iol-2015-individual-p3-sub-a","iol-2015-individual-p3-sub-b","iol-2015-individual-p3-sub-c","iol-2015-individual-p3-partial-through-a-target-b","iol-2015-individual-p3-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"The Soundex rule units are incomplete: step 1 is missing, step 2 is truncated mid-sentence, and step 5 is truncated. Part b and c units bundle many codes into single units.","suggested_fix":"Restore all Soundex algorithm steps and split surname/code outputs into item-level answer_units."},{"severity":"low","record_id":null,"field":"prompt","description":"The source includes a sample census-card image, but the dataset prompt keeps only surrounding text/source URL context.","suggested_fix":"Include the image or a concise transcription of the relevant card fields if they are needed as context."}],"notes":"The main prompt and canonical answer are mostly faithful; answer-unit completeness and granularity are the scoring blockers."}
+{"source_problem_id":"2016-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2016-individual-p3","iol-2016-individual-p3-sub-a","iol-2016-individual-p3-sub-b","iol-2016-individual-p3-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"answer_units","description":"Answer units are on the wrong side of the translation task: part a units contain source Kunuz Nubian sentences instead of English translations and miss item 11; part b units contain English prompts instead of Kunuz Nubian translations and miss item 16.","suggested_fix":"Regenerate answer_units from the official solution so each target item stores the requested translated answer, including items 11 and 16."}],"notes":"Prompt, canonical solution, provenance, and partial labels are otherwise faithful."}
+{"source_problem_id":"2017-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2017-individual-p3","iol-2017-individual-p3-sub-a","iol-2017-individual-p3-sub-b","iol-2017-individual-p3-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"answer_units","description":"Answer units are on the wrong side of the translation task: part a units contain Kimbundu source sentences instead of English translations and miss item 12; part b units contain English prompts instead of Kimbundu translations and miss item 16.","suggested_fix":"Regenerate answer_units from the official solution with one requested translated answer per numbered target."}],"notes":"Prompt and canonical text are faithful; the issue is answer-unit extraction."}
+{"source_problem_id":"2018-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2018-individual-p3","iol-2018-individual-p3-sub-a","iol-2018-individual-p3-sub-b","iol-2018-individual-p3-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Part a answer_units bundle all fourteen gaps into one unit, and the part b unit omits explanatory content from the official solution.","suggested_fix":"Split part a into one unit per gap and include the complete part b answer/rationale expected by the official solution."},{"severity":"medium","record_id":null,"field":"prompt","description":"The extracted table and language note are linearized and partially mangled, reducing fidelity even though the main data are mostly present.","suggested_fix":"Clean table structure and note text against the PDF."}],"notes":"The problem remains broadly readable, but answer-unit granularity is too coarse for scoring."}
+{"source_problem_id":"2019-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2019-individual-p3","iol-2019-individual-p3-sub-a","iol-2019-individual-p3-sub-b","iol-2019-individual-p3-sub-c","iol-2019-individual-p3-sub-d","iol-2019-individual-p3-partial-through-a-target-b","iol-2019-individual-p3-partial-through-ab-target-c","iol-2019-individual-p3-partial-through-abc-target-d"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The source uses Book Pahlavi glyphs, but the dataset exposes internal font-encoding strings instead of the visual forms and ligatures needed to solve the problem.","suggested_fix":"Use rendered image crops or a faithful custom transcription for the Book Pahlavi forms."},{"severity":"critical","record_id":null,"field":"canonical","description":"The canonical solution and shown partial solutions use the same font-encoding artifacts, so solved script answers are not source-faithful.","suggested_fix":"Rebuild solution_text and partial solved sections from rendered pages."},{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units bundle table rows, include stray artifacts and footers, and do not provide clean item-level Pahlavi answers.","suggested_fix":"Create one clean unit per requested subpart item using the repaired script representation."}],"notes":"The cited pages are correct, but this glyph-heavy source cannot be represented by the current extracted text."}
+{"source_problem_id":"2021-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2021-individual-p3","iol-2021-individual-p3-sub-a","iol-2021-individual-p3-sub-b","iol-2021-individual-p3-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"answer_units","description":"Answer units are on the wrong side of the translation task: part a units contain Kilivila prompts instead of English translations and miss item 14; part b units contain English prompts instead of Kilivila translations and miss item 19.","suggested_fix":"Regenerate answer_units from the official solution with one target-side answer per numbered item."}],"notes":"Prompt, canonical solution text, partial labels, and provenance otherwise match the source."}
+{"source_problem_id":"2022-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2022-individual-p3"],"issues":[{"severity":"high","record_id":"iol-2022-individual-p3","field":"prompt","description":"The two-column task list is interleaved in extraction, with later items appearing inside or after earlier item text, which makes the prompt hard to follow.","suggested_fix":"Recreate the task list with explicit columns or separate numbered lines."},{"severity":"critical","record_id":"iol-2022-individual-p3","field":"answer_units","description":"Answer units omit source targets 17 and 23, truncate the answer for 26, and degrade click-symbol forms to ASCII punctuation in some entries.","suggested_fix":"Regenerate all answer_units from the official solution, preserving click characters and complete text for every numbered item."},{"severity":"medium","record_id":"iol-2022-individual-p3","field":"canonical","description":"The word-order chart/rule section is mangled by brace and line-layout extraction, although many final answers are present.","suggested_fix":"Clean the canonical rule chart into a readable structured form."}],"notes":"Only one full-problem record exists for this source."}
+{"source_problem_id":"2023-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2023-individual-p3"],"issues":[{"severity":"critical","record_id":"iol-2023-individual-p3","field":"answer_units","description":"Several answer_units are rule fragments rather than requested final answers, and actual official answers for items 18 and 23 are missing.","suggested_fix":"Replace rule-fragment units with item-level final-answer units and add the missing items 18 and 23."}],"notes":"Prompt and canonical text are otherwise faithful to the cited pages."}
+{"source_problem_id":"2024-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2024-individual-p3","iol-2024-individual-p3-sub-a","iol-2024-individual-p3-sub-b","iol-2024-individual-p3-sub-c","iol-2024-individual-p3-partial-through-a-target-b","iol-2024-individual-p3-partial-through-ab-target-c"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The source problem includes a family-tree diagram above the statements, but the dataset prompt omits the actual tree structure and keeps only a name plus legend words.","suggested_fix":"Include the rendered family tree or transcribe it into explicit relationship edges and node labels."},{"severity":"high","record_id":null,"field":"canonical","description":"The part a official answer is reduced to a loose list of names, losing the solved family-tree structure from the PDF.","suggested_fix":"Represent the solved tree with an image or explicit relationship structure in canonical and answer_units."},{"severity":"critical","record_id":null,"field":"answer_units","description":"Part b answer_units miss gap 1 and part a has only a coarse text list rather than a scoreable tree answer; part c includes footer text.","suggested_fix":"Add the missing part b gap 1 answer, split units by target, remove footers, and encode the tree answer structurally."},{"severity":"high","record_id":"iol-2024-individual-p3-partial-through-a-target-b","field":"subparts","description":"The partial variant shows solved part a as the same unstructured name list, so prior solved context is not faithful.","suggested_fix":"Use the repaired tree representation in shown solved subparts."}],"notes":"Rendered source pages confirm this is diagram-dependent and not fully answerable from the current prompt text."}
+{"source_problem_id":"2025-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2025-individual-p3","iol-2025-individual-p3-sub-a","iol-2025-individual-p3-sub-b","iol-2025-individual-p3-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"answer_units","description":"Answer units for the combined b/c target omit official answers 21 and 25, and c answers are mislabeled under b.","suggested_fix":"Add units for 21 and 25, preserve source labels b and c, and keep one required unit per numbered target."},{"severity":"high","record_id":null,"field":"subparts","description":"The source has subparts a, b, and c, but the dataset has only sub-a and sub-b records; the sub-b and partial target-b records include both b and c tasks.","suggested_fix":"Generate separate sub-b and sub-c records, plus partial variants whose target labels match the source."},{"severity":"low","record_id":null,"field":"prompt","description":"The full and partial prompts include an irrelevant header from page 4 after the Problem 3 text.","suggested_fix":"Trim extraction at the end of Problem 3 rather than carrying over the next page header."}],"notes":"The Kuria prompt and canonical solution are otherwise faithful to the local PDF; the main failures are subpart labeling and missing units."}

benchmark/IOL/ioling_hf/reports/manual_verification/agent_batch_02_findings.jsonl

@@ -0,0 +1,22 @@
+{"source_problem_id":"2003-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2003-individual-p2","iol-2003-individual-p2-sub-assignment_1","iol-2003-individual-p2-sub-assignment_2","iol-2003-individual-p2-partial-through-assignment_1-target-assignment_2"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"The image-PDF prompt OCR replaces source transliteration symbols and fraction layout with incorrect characters such as yen/2/c-like placeholders, changing the Arabic forms and notes from the rendered source page.","suggested_fix":"Re-transcribe the problem from the rendered PDF, preserving the source transliteration letters and displayed fractions."},{"severity":"critical","record_id":null,"field":"canonical","description":"The solution/canonical text and shown solution for assignment 1 contain garbled equations such as \"$+3=$\" instead of the official fraction equations; assignment 2 also has corrupted fractions and transliteration.","suggested_fix":"Rebuild canonical answers and shown solutions from the rendered solution page, e.g. assignment 1 should contain the five explicit fraction equalities and assignment 2 the square-root or square sign alternatives."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Assignment 1 is one bundled answer unit for five equalities, which is too coarse for partial-credit RL scoring even after transcription is fixed.","suggested_fix":"Split assignment 1 into separate answer units for each numbered equality and keep assignment 2 as its own unit(s)."},{"severity":"low","record_id":null,"field":"provenance","description":"The solution page range includes page 3, which is the next problem; Problem 2 solution is on page 2.","suggested_fix":"Set the solution page range for this source to page 2 only."}],"notes":"Rendered problem and solution pages were needed because the cited PDFs have no embedded text on these pages."}
+{"source_problem_id":"2003-team-3","verdict":"fix_needed","record_ids_checked":["iol-2003-team-p3","iol-2003-team-p3-sub-assignment_1","iol-2003-team-p3-sub-assignment_2","iol-2003-team-p3-sub-assignment_3","iol-2003-team-p3-partial-through-assignment_1-target-assignment_2","iol-2003-team-p3-partial-through-assignment_1assignment_2-target-assignment_3"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Several answer units bundle multiple independently scorable items: assignment 2 has seven verbs in one unit and assignment 3 asks for two examples in one unit.","suggested_fix":"Split assignment 2 into one unit per accepted verb and assignment 3 into separate example units or an explicit two-example set rubric."},{"severity":"low","record_id":null,"field":"prompt","description":"The prompt has an OCR typo: \"Tt is known\" should be \"It is known\".","suggested_fix":"Correct the prompt text against the rendered source page."},{"severity":"low","record_id":null,"field":"canonical","description":"The solution text has minor OCR errors in prose, including a missing opening quote before \"I denounce\" and \"of\" where the source says \"or\" in the parenthetical example.","suggested_fix":"Correct the solution prose from the rendered solution page."}],"notes":"Problem and solution content are otherwise faithful and answerable."}
+{"source_problem_id":"2005-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2005-individual-p1","iol-2005-individual-p1-sub-assignment_1","iol-2005-individual-p1-sub-assignment_2","iol-2005-individual-p1-partial-through-assignment_1-target-assignment_2"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Assignment 1's four translations and assignment 2's five translations are each bundled into a single answer unit, which is too coarse for RL scoring.","suggested_fix":"Create separate answer units for problem items 16-24."},{"severity":"medium","record_id":null,"field":"canonical","description":"The canonical solution includes OCR artifacts, including a slash in the final answer (\"Ch/abal\" instead of the source form) and page-footer text from the next page.","suggested_fix":"Clean the canonical/subpart answers against the OCR-rendered solution page and remove page footer noise."},{"severity":"low","record_id":null,"field":"provenance","description":"The cited solution range includes page 2, which starts Problem 2; Problem 1 solution is on page 1.","suggested_fix":"Set the solution page range to page 1 only for this source."}],"notes":"The PDF has custom-encoded embedded text, so the repository OCR text was used for verification."}
+{"source_problem_id":"2006-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2006-individual-p2","iol-2006-individual-p2-sub-assignment_1","iol-2006-individual-p2-sub-assignment_2","iol-2006-individual-p2-sub-assignment_3","iol-2006-individual-p2-partial-through-assignment_1-target-assignment_2","iol-2006-individual-p2-partial-through-assignment_1assignment_2-target-assignment_3"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Assignment 1 is one large answer unit covering all missing Catalan forms, which is too coarse for partial-credit scoring.","suggested_fix":"Split assignment 1 into separate units for each missing singular/plural/article form or row."},{"severity":"low","record_id":null,"field":"canonical","description":"The canonical and subpart answers include page-footer noise and a wrong footer page number.","suggested_fix":"Remove contest footer text from canonical/subpart answers."},{"severity":"low","record_id":null,"field":"provenance","description":"Problem and solution page ranges include page 3, which is the next problem; the Catalan problem and solution are on page 2.","suggested_fix":"Narrow both problem and solution ranges to page 2."}],"notes":"The Catalan prompt and main answers are otherwise readable and source-faithful."}
+{"source_problem_id":"2007-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2007-individual-p3"],"issues":[{"severity":"critical","record_id":"iol-2007-individual-p3","field":"canonical","description":"The canonical solution is corrupted by interleaving prose with the layout table, making the explanation hard to read and not faithful to the solution PDF order.","suggested_fix":"Re-extract or manually transcribe the solution so prose, final correspondence table, and tense table are separated in source order."},{"severity":"critical","record_id":"iol-2007-individual-p3","field":"answer_units","description":"The only answer unit has key \"keni\" and value \"tthe suffix -eb by contrasting visadileb and vsadilob. We can't\", which is a stray fragment rather than the requested Georgian-English correspondences.","suggested_fix":"Replace with structured answer units for the 15 final correspondences and the ambiguity/plausibility note."},{"severity":"low","record_id":"iol-2007-individual-p3","field":"prompt","description":"The prompt has minor OCR artifacts such as \"Problem !3\" and a garbled section marker before the task.","suggested_fix":"Correct headings/section markers from the source PDF."}],"notes":"Embedded PDF text was available, and the source solution clearly shows final correspondences absent from the answer unit."}
+{"source_problem_id":"2008-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2008-individual-p3","iol-2008-individual-p3-sub-a","iol-2008-individual-p3-sub-b","iol-2008-individual-p3-sub-c","iol-2008-individual-p3-partial-through-a-target-b","iol-2008-individual-p3-partial-through-ab-target-c"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Answer units are too coarse: subpart (a) bundles all correspondences, subpart (b) bundles four lexical meanings, and subpart (c) bundles two translations.","suggested_fix":"Split (a) by correspondence row, (b) by requested word, and (c) by requested Drehu translation."}],"notes":"Prompt, partial shown solutions, and canonical text match the cited pages, but the scoring granularity is not adequate."}
+{"source_problem_id":"2009-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2009-individual-p3"],"issues":[{"severity":"medium","record_id":"iol-2009-individual-p3","field":"prompt","description":"The phonetic note in the prompt is line-smashed into an unreadable fragment around the definitions of c, E, h, and N.","suggested_fix":"Reformat the note as normal prose from the source PDF."},{"severity":"medium","record_id":"iol-2009-individual-p3","field":"answer_units","description":"All six Burmese child name/date answers are in a single free-text unit with reasoning, which is too coarse for RL scoring.","suggested_fix":"Create one answer unit per requested child/date assignment."}],"notes":"The main names/date table is faithful and answerable."}
+{"source_problem_id":"2010-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2010-individual-p3","iol-2010-individual-p3-sub-a","iol-2010-individual-p3-sub-b","iol-2010-individual-p3-sub-c","iol-2010-individual-p3-partial-through-a-target-b","iol-2010-individual-p3-partial-through-ab-target-c"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The Blissymbolics symbols are omitted from the model-facing prompt; the prompt lists only English words and blank tasks, so subparts (a) and (b) are not answerable.","suggested_fix":"Include rendered images or an exact textual/vector representation of every Blissymbolic symbol in the examples and target questions."},{"severity":"critical","record_id":null,"field":"canonical","description":"The canonical solution and shown solutions describe symbol compositions in words but do not preserve the actual Blissymbolic glyphs needed for source fidelity.","suggested_fix":"Store the official symbol glyphs or linked rendered crops alongside the textual composition descriptions."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Answer units are large composition tables rather than individual symbol correspondences/outputs.","suggested_fix":"Split by each requested symbol correspondence and each requested produced Blissymbolic word."},{"severity":"low","record_id":null,"field":"provenance","description":"The cited page ranges include adjacent Problem 4 pages although Problem 3 content is on page 2.","suggested_fix":"Narrow problem and solution ranges to the page containing Problem 3, while retaining any image assets from that page."}],"notes":"Rendered PDF inspection confirms the omitted content is essential."}
+{"source_problem_id":"2011-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2011-individual-p3","iol-2011-individual-p3-sub-a","iol-2011-individual-p3-sub-b","iol-2011-individual-p3-sub-c","iol-2011-individual-p3-partial-through-a-target-b","iol-2011-individual-p3-partial-through-ab-target-c"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Subpart (a) has three translations in one unit and subpart (c) has four Vai translations in one unit, which is too coarse for partial-credit RL scoring.","suggested_fix":"Split answer units by each requested phrase/translation in (a) and (c)."}],"notes":"Prompt, canonical answer, and partial variants are otherwise faithful to the cited pages."}
+{"source_problem_id":"2012-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2012-individual-p3","iol-2012-individual-p3-sub-a","iol-2012-individual-p3-sub-b","iol-2012-individual-p3-sub-c","iol-2012-individual-p3-sub-d","iol-2012-individual-p3-partial-through-a-target-b","iol-2012-individual-p3-partial-through-ab-target-c","iol-2012-individual-p3-partial-through-abc-target-d"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Answer units are too coarse and partly layout-scrambled: subpart (a) is one correspondence table unit, while (b) and (c) each bundle two answers into one unit.","suggested_fix":"Split (a) by Basque-English correspondence and split (b)/(c) by requested sentence."},{"severity":"medium","record_id":null,"field":"canonical","description":"The canonical subpart (a) correspondence table is flattened into a hard-to-read layout and subpart (d) includes solution footer text.","suggested_fix":"Reformat the solution table and remove page footers from canonical/subpart answers."},{"severity":"low","record_id":null,"field":"provenance","description":"The page ranges include adjacent Problems 2 and 4 pages around Problem 3.","suggested_fix":"Use page spans or text spans that isolate Problem 3 where possible."}],"notes":"Main prompt content is answerable, but solution/scoring structure needs cleanup."}
+{"source_problem_id":"2013-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2013-individual-p3"],"issues":[{"severity":"critical","record_id":"iol-2013-individual-p3","field":"answer_units","description":"The five answer units are rule fragments only; none of the ten requested Piraha pronunciations are represented as answer units.","suggested_fix":"Create answer units for each requested item: xaaibi, xaapisi, xitiixisi, bigi, kagahoaogii, kaibai, kapiigaiitoii, poogaihiai toio, xabagi kapioxio, and xabagi xogiai."},{"severity":"medium","record_id":"iol-2013-individual-p3","field":"prompt","description":"The prompt table has layout artifacts, including a misplaced \"toucan\" label near the kaoaibogi row and page-footer noise.","suggested_fix":"Reformat the training and target tables from the source PDF and remove footers."},{"severity":"low","record_id":"iol-2013-individual-p3","field":"provenance","description":"The problem page range includes page 4, which is the next problem.","suggested_fix":"Narrow the problem page range to page 3 for this source."}],"notes":"The canonical text contains the final pronunciations, but the scoring units do not."}
+{"source_problem_id":"2014-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2014-individual-p4","iol-2014-individual-p4-sub-a","iol-2014-individual-p4-sub-b","iol-2014-individual-p4-sub-c","iol-2014-individual-p4-partial-through-a-target-b","iol-2014-individual-p4-partial-through-ab-target-c"],"issues":[{"severity":"critical","record_id":null,"field":"answer_units","description":"Answer units omit major required answers: subpart (a) lacks item 8 and truncates items 9/10; subpart (b) lacks item 11 and the second line of item 12; subpart (c) includes footer text.","suggested_fix":"Rebuild answer units from the canonical answer so every numbered English/Engenni sentence and base-form explanation is separately represented."},{"severity":"medium","record_id":null,"field":"prompt","description":"The note explaining tone marks loses the mark symbols, and several pronoun glosses are mashed together as strings like \"hethe\".","suggested_fix":"Re-transcribe the note and affected English glosses with correct spacing and tone mark names/symbols."},{"severity":"low","record_id":null,"field":"canonical","description":"Canonical prose contains OCR artifacts such as missing letters in \"affected\"/\"affirmative\" and includes solution footnote/footer noise in subpart answers.","suggested_fix":"Clean OCR artifacts and remove footers from canonical and subpart_solutions."}],"notes":"The official solution on the cited pages contains the missing answers clearly."}
+{"source_problem_id":"2015-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2015-individual-p4","iol-2015-individual-p4-sub-a","iol-2015-individual-p4-sub-b","iol-2015-individual-p4-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first answer in both subparts: item 18 in (a) and item 24 in (b). The unit keys also use solution-local numbering 2-6 and 8-11 instead of prompt item numbers 18-28.","suggested_fix":"Add the missing units for 18 and 24 and renumber/split units to match prompt items 18-28."},{"severity":"low","record_id":null,"field":"provenance","description":"Problem and solution ranges include adjacent Problem 5 pages, and subpart prompts include page-footer text.","suggested_fix":"Narrow page ranges or text spans to Problem 4 and strip footers."}],"notes":"The prompt and canonical answer are otherwise source-faithful."}
+{"source_problem_id":"2016-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2016-individual-p4","iol-2016-individual-p4-sub-a","iol-2016-individual-p4-sub-b","iol-2016-individual-p4-sub-c","iol-2016-individual-p4-partial-through-a-target-b","iol-2016-individual-p4-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first required answer in every subpart: (a) item 1, (b) item 1 plus the synonym explanation, and (c) item 1.","suggested_fix":"Add units for gu-na vaala = canoe, vaala = canoe with the synonym explanation, and pig = bak; align keys with source item numbering."},{"severity":"low","record_id":null,"field":"canonical","description":"Canonical/subpart answers include page-footer text from the solution PDF.","suggested_fix":"Remove footer text from canonical answers."}],"notes":"Prompt and partial-variant shown solutions are otherwise faithful."}
+{"source_problem_id":"2017-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2017-individual-p4"],"issues":[{"severity":"critical","record_id":"iol-2017-individual-p4","field":"prompt","description":"The Khom script column and source illustrations are omitted from the prompt; the task asks to fill question-mark cells including script cells, so the record is not answerable from text alone.","suggested_fix":"Include rendered table/image assets or exact Khom glyph representations for all source and target cells."},{"severity":"critical","record_id":"iol-2017-individual-p4","field":"canonical","description":"The canonical answer represents Khom-script answers with arrows and blank positions rather than actual glyphs, losing the official written answers.","suggested_fix":"Store the rendered Khom glyph answers or a faithful transcription/image crop for each target cell."},{"severity":"medium","record_id":"iol-2017-individual-p4","field":"answer_units","description":"All target cells are in one large free-text unit, too coarse for scoring even if glyphs are restored.","suggested_fix":"Create one answer unit per filled cell or per row/cell target."}],"notes":"Rendered page inspection confirms essential visual/script content is missing."}
+{"source_problem_id":"2018-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2018-individual-p4","iol-2018-individual-p4-sub-a","iol-2018-individual-p4-sub-b","iol-2018-individual-p4-sub-c","iol-2018-individual-p4-partial-through-a-target-b","iol-2018-individual-p4-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit required first answers in translation subparts: item 12 in (b) and item 17 in (c).","suggested_fix":"Add answer units for 12. biabys atub bobys = 3 betel nuts and 17. 4 betel nuts = biabys atubys atub bobys, preserving source diacritics."},{"severity":"medium","record_id":null,"field":"prompt","description":"The subpart (c) note is line-scrambled into text like \"15,000Mountainpeople inArapesh...\", reducing prompt fidelity.","suggested_fix":"Reformat the note from the source PDF with normal spacing."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Subpart (a)'s multi-part explanation is a single unit; this is coarse for partial credit.","suggested_fix":"Split subpart (a) into units for word X, examples 6/9/11, and how the word is implied."}],"notes":"The main data table and canonical answers match the cited source."}
+{"source_problem_id":"2019-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2019-individual-p4"],"issues":[{"severity":"medium","record_id":"iol-2019-individual-p4","field":"answer_units","description":"All nine North-dialect and nine Coast-dialect gap fills are bundled into one free-text answer unit, which is too coarse for RL scoring.","suggested_fix":"Create separate answer units for each requested form or each row/dialect cell."},{"severity":"low","record_id":"iol-2019-individual-p4","field":"provenance","description":"Problem and solution page ranges include adjacent problems; the prompt also contains a page footer between the example table and target table.","suggested_fix":"Use ranges/text spans that isolate Problem 4 and strip footers."}],"notes":"The prompt and canonical content are otherwise answerable and faithful."}
+{"source_problem_id":"2021-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2021-individual-p4","iol-2021-individual-p4-sub-a","iol-2021-individual-p4-sub-b","iol-2021-individual-p4-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Subpart (b) answer units omit item 1; units start at key 2 even though the canonical has the form for rutu/chewing-stick.","suggested_fix":"Add an answer unit for item 1 and align keys with source numbering."},{"severity":"low","record_id":null,"field":"canonical","description":"The rule table in the canonical contains glyph artifacts shown as question marks and includes page-footer text.","suggested_fix":"Clean or replace the rule-table extraction and remove footer text."}],"notes":"Prompt and target subparts are otherwise faithful."}
+{"source_problem_id":"2022-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2022-individual-p4"],"issues":[{"severity":"critical","record_id":"iol-2022-individual-p4","field":"prompt","description":"The Arabana family-tree diagrams are linearized into symbols and numbers, losing marriage, sibling, and parent-child structure. The record is not answerable without the rendered PDF.","suggested_fix":"Include rendered family-tree images or a structured tree representation with all relationships."},{"severity":"critical","record_id":"iol-2022-individual-p4","field":"canonical","description":"The canonical solution is a lossy one-unit diagram dump rather than structured A-X position mappings and blank (i)-(ix) answers.","suggested_fix":"Represent the solution as explicit mappings for A-X to 1-24 and separate values for blanks i-ix, with any diagrams preserved as images if needed."},{"severity":"medium","record_id":"iol-2022-individual-p4","field":"answer_units","description":"The entire solution is one answer unit, too coarse for scoring the 24 mappings and 9 blanks.","suggested_fix":"Create one unit per family-member mapping and one unit per blank."}],"notes":"Rendered source page confirms the lost tree structure is essential."}
+{"source_problem_id":"2023-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2023-individual-p4"],"issues":[{"severity":"high","record_id":"iol-2023-individual-p4","field":"answer_units","description":"Answer units omit requested items 22 and 26; units start at 23 for subpart (a) and 27 for subpart (b).","suggested_fix":"Add units for 22. e-wapamikoyek = as he sees you(pl) and 26. if we ask them = kakwecimayahkwawi, preserving source diacritics."},{"severity":"low","record_id":"iol-2023-individual-p4","field":"prompt","description":"The note for the long-vowel mark is awkwardly formatted as \"mark\" joined with the macron symbol.","suggested_fix":"Reformat the note so it clearly says the macron indicates a long vowel."}],"notes":"Main Plains Cree prompt and canonical answer are otherwise faithful."}
+{"source_problem_id":"2024-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2024-individual-p4","iol-2024-individual-p4-sub-a","iol-2024-individual-p4-sub-b","iol-2024-individual-p4-sub-c","iol-2024-individual-p4-partial-through-a-target-b","iol-2024-individual-p4-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first required answer in each subpart: (a) item 1, (b) item 21, and (c) item 24.","suggested_fix":"Add units for 1. cum aa = C flip-flops, 21. daw sob piis = finger, and 24. brook/stream = nax piis, preserving source characters."},{"severity":"medium","record_id":null,"field":"prompt","description":"The source problem refers to images on the next page, but the prompt only contains captions for capybara, tapir, and anteater without the images.","suggested_fix":"Include the rendered animal images or mark them as omitted only if the dataset intentionally relies on the textual note instead."}],"notes":"Prompt text is largely faithful, but scoring coverage is incomplete."}
+{"source_problem_id":"2025-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2025-individual-p4","iol-2025-individual-p4-sub-a","iol-2025-individual-p4-sub-b","iol-2025-individual-p4-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"subparts","description":"Subpart (b) records and the partial variant incorrectly merge the two-column (b)/(c) target into one target question, so a record labeled target (b) also asks for (c).","suggested_fix":"Separate subpart (b) and subpart (c) into distinct records/targets, or label the current record as a combined b+c task if that is intended."},{"severity":"high","record_id":null,"field":"answer_units","description":"The subpart (b) answer units mix (b) and (c) answers under label b, omit item 40 from (b), and omit item 47 from (c).","suggested_fix":"Create separate b units for 40-46 and c units for 47-53, including the missing repena = tree/fire and white man = bali answers."},{"severity":"medium","record_id":"iol-2025-individual-p4-sub-a","field":"answer_units","description":"Subpart (a)'s 25 matching answers are bundled into one unit, too coarse for RL scoring.","suggested_fix":"Split subpart (a) into one answer unit for each item 15-39."},{"severity":"low","record_id":null,"field":"provenance","description":"The problem page range includes page 5, which starts Problem 5, and prompts include page-footer text.","suggested_fix":"Narrow the page/text span to Problem 4 and remove footer noise."}],"notes":"The full problem prompt is readable, but the derived subproblem/partial records are not faithful to the source subpart boundaries."}

benchmark/IOL/ioling_hf/reports/manual_verification/agent_batch_03_findings.jsonl

@@ -0,0 +1,22 @@
+{"issues": [{"description": "OCR text drops the visible underline blanks in the Basque/English lists and carries in an unrelated Egyptian Arabic footnote from the same PDF page; the prompt is mostly readable but not a clean rendering of the cited problem.", "field": "prompt", "record_id": null, "severity": "medium", "suggested_fix": "Manually transcribe Problem 3 from page 2, preserving blank positions and only the Basque footnote."}, {"description": "Each assignment is a single answer unit even where Assignment 1 has ten matches/fill-ins and Assignment 2 has four translations, which is too coarse for RL scoring.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split answer_units into one unit per requested date/match/translation and keep the literal-name explanation as separate units."}], "notes": "Checked rendered OCR pages for the image-only 2003 problem and solution. Canonical solution is broadly from the cited solution pages but has OCR artifacts around blanks.", "record_ids_checked": ["iol-2003-individual-p3", "iol-2003-individual-p3-sub-assignment_1", "iol-2003-individual-p3-sub-assignment_2", "iol-2003-individual-p3-sub-assignment_3", "iol-2003-individual-p3-partial-through-assignment_1-target-assignment_2", "iol-2003-individual-p3-partial-through-assignment_1assignment_2-target-assignment_3"], "source_problem_id": "2003-individual-3", "verdict": "fix_needed"}
+{"issues": [{"description": "Kayapo special vowels are corrupted in the prompt: PDF forms such as mé/mẽ and aje/ije are represented with ! and quotation-mark artifacts (for example \"Ba m! m! anhê\", \"Ij\" m! akuru kêt\", \"M! aj\" inhêrê kêt\").", "field": "prompt", "record_id": null, "severity": "high", "suggested_fix": "Re-extract or manually transcribe the Kayapo forms using the PDF glyphs, e.g. mé, aje, ije, and the note vowels."}, {"description": "The solution answer for Assignment 1 contains OCR damage: \"Tréré két Tam not swimming\" instead of the PDF answer \"Irêrê két — I am not swimming\".", "field": "canonical", "record_id": null, "severity": "high", "suggested_fix": "Correct the Assignment 1 canonical answer and derived answer_units from the rendered solution page."}, {"description": "Assignment-level answer units group several translations together, making partial scoring too coarse.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Create separate units for each Assignment 1 sentence and each Assignment 2 translation."}], "notes": "Verified against rendered page 1 of both problem and solution PDFs.", "record_ids_checked": ["iol-2004-individual-p1", "iol-2004-individual-p1-sub-assignment_1", "iol-2004-individual-p1-sub-assignment_2", "iol-2004-individual-p1-partial-through-assignment_1-target-assignment_2"], "source_problem_id": "2004-individual-1", "verdict": "fix_needed"}
+{"issues": [{"description": "The Lango prompt loses distinctive source glyphs and tones: rendered PDF has forms such as dyè ɔ̀t, nìg wàŋ, ɔ̀t cèm, but the dataset has plain/incorrect forms such as dye ot, nig way, at cem.", "field": "prompt", "record_id": null, "severity": "high", "suggested_fix": "Manually transcribe the Lango forms and note from the PDF, preserving ɔ/ɛ, ŋ, and tone marks."}, {"description": "The canonical solution contains OCR substitutions in target forms, including \"dye dt\", \"St cém\", and \"way ot\" where the PDF shows dyè ɔ̀t, ɔ̀t cèm, and wàŋ ɔ̀t.", "field": "canonical", "record_id": null, "severity": "high", "suggested_fix": "Correct the solution_text, canonical answers, and answer_units from the rendered solution page."}, {"description": "Assignment 1 is one large unit with eight correspondences, too coarse for RL scoring.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split into one unit per correspondence and one unit per short answer in Assignments 2 and 3."}], "notes": "Rendered problem and solution page 2; OCR text is not faithful for Lango diacritics.", "record_ids_checked": ["iol-2005-individual-p2", "iol-2005-individual-p2-sub-assignment_1", "iol-2005-individual-p2-sub-assignment_2", "iol-2005-individual-p2-sub-assignment_3", "iol-2005-individual-p2-partial-through-assignment_1-target-assignment_2", "iol-2005-individual-p2-partial-through-assignment_1assignment_2-target-assignment_3"], "source_problem_id": "2005-individual-2", "verdict": "fix_needed"}
+{"issues": [{"description": "Assignment 1 and Assignment 2 are each single answer units even though they contain many correspondences/romanizations; this is too coarse for RL scoring.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split Assignment 1 into individual word correspondence units and Assignment 2 into one unit per requested form."}], "notes": "Rendered/PyMuPDF text matches the cited Khmer pages closely; issue is scoring granularity rather than source fidelity.", "record_ids_checked": ["iol-2006-individual-p3", "iol-2006-individual-p3-sub-assignment_1", "iol-2006-individual-p3-sub-assignment_2", "iol-2006-individual-p3-partial-through-assignment_1-target-assignment_2"], "source_problem_id": "2006-individual-3", "verdict": "fix_needed"}
+{"issues": [{"description": "Important mathematical symbols are corrupted: the prompt has \"mer abo sas $ meregh\" where the PDF has a multiplication sign, and problem number/sign glyphs are also mapped incorrectly.", "field": "prompt", "record_id": "iol-2007-individual-p4", "severity": "high", "suggested_fix": "Manually correct multiplication signs, problem markers, and symbols from the rendered PDF."}, {"description": "The solution_text/canonical maps multiplication and exponent notation to punctuation such as %, &, and apostrophes (for example 6& and 7'x) instead of 6² and 7×x.", "field": "canonical", "record_id": "iol-2007-individual-p4", "severity": "high", "suggested_fix": "Replace the corrupted notation with the actual PDF math symbols throughout the canonical solution."}, {"description": "The full problem has a single full-answer unit, too coarse for the four requested tasks.", "field": "answer_units", "record_id": "iol-2007-individual-p4", "severity": "medium", "suggested_fix": "Create units for §1 square identifications, §2 equality, §3 numerals, and §4 Ndom forms."}], "notes": "Rendered pages 3-4; the source page range is correct but embedded-font math extraction is not faithful.", "record_ids_checked": ["iol-2007-individual-p4"], "source_problem_id": "2007-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "Copainalá Zoque transcription glyphs are corrupted: PDF characters such as ʌ, ŋ, ʔ, and š appear as 2, N, P, and ˇ/letter sequences in the prompt.", "field": "prompt", "record_id": "iol-2008-individual-p4", "severity": "high", "suggested_fix": "Manually transcribe the problem with the actual IPA-like characters shown in the PDF."}, {"description": "The same transcription corruption appears in canonical answers, changing forms such as -kʌsmʌ, ʔas, ŋgumguy, and šeh.", "field": "canonical", "record_id": "iol-2008-individual-p4", "severity": "high", "suggested_fix": "Correct solution_text/canonical answer forms from the rendered solution page."}, {"description": "The answer_units capture numbered suffix-rule bullets rather than the requested final answers for parts (a) and (b), so scoring would not check the actual task outputs.", "field": "answer_units", "record_id": "iol-2008-individual-p4", "severity": "critical", "suggested_fix": "Replace units with the six English translations in part (a) and six Copainalá Zoque translations in part (b)."}], "notes": "Rendered problem page 3 and solution page 2; source range is correct but text encoding is lossy.", "record_ids_checked": ["iol-2008-individual-p4"], "source_problem_id": "2008-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "Old Indic diacritics and the explanatory note are badly garbled: the model-facing prompt uses lossy forms like v´r.k-a- and the note about ṇ, ṣ, ṭ, and ṛ is scrambled, even though stress/diacritics are central to the problem.", "field": "prompt", "record_id": null, "severity": "high", "suggested_fix": "Manually transcribe the Sanskrit stems and note using proper Unicode diacritics from the PDF."}, {"description": "Canonical answer forms use decomposed/placeholder stress and length notation rather than the PDF glyphs.", "field": "canonical", "record_id": null, "severity": "medium", "suggested_fix": "Normalize canonical solution forms to the PDF’s Unicode diacritics and stress marks."}, {"description": "Part (b) is a single unit containing all stressed stems, too coarse for scoring individual answers.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split part (b) into one unit per stem and keep part (a) as a separate explanatory unit."}], "notes": "Rendered page 3 confirms the prompt note and stem diacritics are not faithfully represented.", "record_ids_checked": ["iol-2009-individual-p4", "iol-2009-individual-p4-sub-a", "iol-2009-individual-p4-sub-b", "iol-2009-individual-p4-partial-through-a-target-b"], "source_problem_id": "2009-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "The full-problem answer_units cover only part (c); answers for parts (a) and (b) are present in canonical text but unscored.", "field": "answer_units", "record_id": "iol-2010-individual-p4", "severity": "high", "suggested_fix": "Add units for the synthesized polypeptide answer in (a) and the valid mRNA-sequence choices/count in (b)."}, {"description": "The solution table and part (b) combinatorics are linearized with glyph artifacts such as private-use braces and !/\" markers, making the canonical solution hard to verify mechanically.", "field": "canonical", "record_id": null, "severity": "medium", "suggested_fix": "Manually clean the codon table and part (b) answer formatting from the solution PDF."}], "notes": "Problem prompt is answerable; issue is incomplete/coarse scoring and some solution formatting artifacts.", "record_ids_checked": ["iol-2010-individual-p4", "iol-2010-individual-p4-sub-c"], "source_problem_id": "2010-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "Each subpart is a single large answer unit even though part (a) has many correspondences and parts (b)/(c) each have multiple translations; too coarse for RL scoring.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split part (a) by word correspondence and parts (b)/(c) by requested word or phrase."}], "notes": "Text matches the cited Nahuatl pages; no prompt or canonical fidelity defect found beyond unit granularity.", "record_ids_checked": ["iol-2011-individual-p4", "iol-2011-individual-p4-sub-a", "iol-2011-individual-p4-sub-b", "iol-2011-individual-p4-sub-c", "iol-2011-individual-p4-partial-through-a-target-b", "iol-2011-individual-p4-partial-through-ab-target-c"], "source_problem_id": "2011-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "The first answer on each labeled solution line is omitted from answer_units: part (a) misses 13, part (b) misses 16, and part (c) misses the first target answer.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Regenerate units so answers 13, 16, and the first part (c) translation are included, and split every numbered target separately."}, {"description": "Part (b) canonical includes explanatory rule text after the target translations, which is acceptable for solution_text but noisy as the canonical answer for the subproblem/partial variant.", "field": "canonical", "record_id": null, "severity": "medium", "suggested_fix": "Keep rationale in solution_text but set canonical for subproblem records to the requested translations only."}], "notes": "Prompt and cited page ranges are correct; issue is answer-unit completeness and canonical scoping.", "record_ids_checked": ["iol-2012-individual-p4", "iol-2012-individual-p4-sub-a", "iol-2012-individual-p4-sub-b", "iol-2012-individual-p4-sub-c", "iol-2012-individual-p4-partial-through-a-target-b", "iol-2012-individual-p4-partial-through-ab-target-c"], "source_problem_id": "2012-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "Subpart solution splitting is wrong: sub-a asks for items 12 and 13, but the canonical omits item 12 and includes answers for items 14 and 15, which belong to part (b).", "field": "subparts", "record_id": "iol-2013-individual-p4-sub-a", "severity": "critical", "suggested_fix": "Manually split the solution so part (a) contains answers 12 and 13 only."}, {"description": "Sub-b asks for items 14-17, but its canonical contains only items 16 and 17 because answers 14 and 15 were attached to part (a).", "field": "canonical", "record_id": "iol-2013-individual-p4-sub-b", "severity": "critical", "suggested_fix": "Move answers 14 and 15 into the part (b) canonical and units."}, {"description": "The partial-through-a variant leaks part (b) answers 14 and 15 in the shown official solution for part (a).", "field": "subparts", "record_id": "iol-2013-individual-p4-partial-through-a-target-b", "severity": "high", "suggested_fix": "After correcting subpart splits, regenerate partial variants so shown_solutions contain only prior target answers."}], "notes": "Rendered/source text is otherwise faithful; the issue is caused by the official solution’s visual placement of (a)/(b) markers.", "record_ids_checked": ["iol-2013-individual-p4", "iol-2013-individual-p4-sub-a", "iol-2013-individual-p4-sub-b", "iol-2013-individual-p4-partial-through-a-target-b"], "source_problem_id": "2013-individual-4", "verdict": "fix_needed"}
+{"issues": [{"description": "The fi ligature is dropped in several English words, producing \"eld edge\" and \"good eld\" instead of \"field edge\" and \"good field\" in prompt/canonical text.", "field": "prompt", "record_id": null, "severity": "medium", "suggested_fix": "Normalize PDF ligatures so field/simplified are restored."}, {"description": "Subpart units are large blocks: part (a) has all correspondences as one unit, and parts (b)/(c) each group several answers.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split answer_units by each correspondence/translation requested."}], "notes": "PDF page ranges are correct; source is mostly faithful but ligature loss affects target words.", "record_ids_checked": ["iol-2014-individual-p5", "iol-2014-individual-p5-sub-a", "iol-2014-individual-p5-sub-b", "iol-2014-individual-p5-sub-c", "iol-2014-individual-p5-partial-through-a-target-b", "iol-2014-individual-p5-partial-through-ab-target-c"], "source_problem_id": "2014-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "Full-problem answer_units only cover part (b) and omit the descriptive answer for part (a). Within part (b), item 36 is also omitted because it is on the same line as the (b) marker.", "field": "answer_units", "record_id": "iol-2015-individual-p5", "severity": "high", "suggested_fix": "Add a part (a) meter-structure unit and include item 36 plus one unit for each 36-45 genuine/rearranged judgment."}], "notes": "Problem and solution text are source-faithful enough; issue is scoring completeness.", "record_ids_checked": ["iol-2015-individual-p5", "iol-2015-individual-p5-sub-b"], "source_problem_id": "2015-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "The first numbered answer in each subpart is missing from answer_units: item 1 in part (a) and item 4 in part (b) are present in canonical text but not unitized.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Regenerate units to include items 1-3 and 4-9 separately."}], "notes": "Jaqaru prompt/canonical match the cited pages; scoring units are incomplete.", "record_ids_checked": ["iol-2016-individual-p5", "iol-2016-individual-p5-sub-a", "iol-2016-individual-p5-sub-b", "iol-2016-individual-p5-partial-through-a-target-b"], "source_problem_id": "2016-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "Part (b)/(c) answer_units omit item 1 (laradi) and item 8 (heart) because those answers share lines with subpart markers.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Include all items 1-14 as separate units, with part (a) as its own literal-translation unit."}], "notes": "Madak source text is faithful; the failure is answer-unit coverage.", "record_ids_checked": ["iol-2017-individual-p5", "iol-2017-individual-p5-sub-a", "iol-2017-individual-p5-sub-b", "iol-2017-individual-p5-partial-through-a-target-b"], "source_problem_id": "2017-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "The family-tree diagram is not represented coherently in the text prompt; the PDF geometry/legend needed for part (a) becomes a jumbled sequence of names and symbols.", "field": "prompt", "record_id": null, "severity": "critical", "suggested_fix": "Provide the family tree as an image attachment or manually encode the graph structure in text."}, {"description": "Part (b) answer_units only include statement 2 and omit statement 1; part (a) has only a shape sequence rather than a scoreable named family tree.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Create scoreable units for the names in the family tree and both filled statements."}], "notes": "Rendered PDF confirms the problem depends on a diagram; current text-only extraction is not answerable for part (a).", "record_ids_checked": ["iol-2018-individual-p5", "iol-2018-individual-p5-sub-a", "iol-2018-individual-p5-sub-b", "iol-2018-individual-p5-partial-through-a-target-b"], "source_problem_id": "2018-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "Answer_units omit the first item in parts (a) and (c): item 16 and item 20 are in canonical text but not scored. Part (b) is also a single list unit.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Create separate units for items 16-19, the week-order list, and items 20-24."}], "notes": "Nooni prompt and solution text match the cited pages closely; issue is unit completeness/granularity.", "record_ids_checked": ["iol-2019-individual-p5", "iol-2019-individual-p5-sub-a", "iol-2019-individual-p5-sub-b", "iol-2019-individual-p5-sub-c", "iol-2019-individual-p5-partial-through-a-target-b", "iol-2019-individual-p5-partial-through-ab-target-c"], "source_problem_id": "2019-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "The two-column table with bracketed alternative translations is scrambled in the prompt; brackets and alternatives are interleaved with unrelated entries, making the text record unreliable and not fully answerable.", "field": "prompt", "record_id": "iol-2021-individual-p5", "severity": "critical", "suggested_fix": "Manually encode the Rikbaktsa table row-by-row, preserving bracketed alternative translations."}, {"description": "The solution’s paradigm table and several answers are similarly scrambled by linearization, including bracket artifacts in part (b).", "field": "canonical", "record_id": "iol-2021-individual-p5", "severity": "high", "suggested_fix": "Clean the solution_text/canonical from the rendered solution page."}, {"description": "The full record has a single full-answer unit, too coarse for multiple requested translations.", "field": "answer_units", "record_id": "iol-2021-individual-p5", "severity": "medium", "suggested_fix": "Split into one unit per part (a) and part (b) requested item."}], "notes": "Rendered page 5 shows the original table layout; the text extraction does not preserve it.", "record_ids_checked": ["iol-2021-individual-p5"], "source_problem_id": "2021-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "Part (b) has 18 blanks but is represented as one answer unit, too coarse for RL scoring.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Split part (b) into one unit per numbered blank and keep part (a) as a separate unit."}, {"description": "The source is table-heavy and the linearized prompt moves some column labels/entries out of visual order; values remain present but require careful manual cleanup for fidelity.", "field": "prompt", "record_id": null, "severity": "low", "suggested_fix": "Prefer a manually transcribed Markdown table or image-backed prompt for this problem."}], "notes": "Rendered table pages were checked; answer values are present, but scoring granularity is insufficient.", "record_ids_checked": ["iol-2022-individual-p5", "iol-2022-individual-p5-sub-a", "iol-2022-individual-p5-sub-b", "iol-2022-individual-p5-partial-through-a-target-b"], "source_problem_id": "2022-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "Part (a) is one unit for two digit answers and part (b) is one unit for five Supyire numerals, too coarse for RL scoring.", "field": "answer_units", "record_id": null, "severity": "medium", "suggested_fix": "Create one answer unit per requested number/numeral."}], "notes": "Supyire prompt and solution text are faithful; only scoring granularity needs correction.", "record_ids_checked": ["iol-2023-individual-p5", "iol-2023-individual-p5-sub-a", "iol-2023-individual-p5-sub-b", "iol-2023-individual-p5-partial-through-a-target-b"], "source_problem_id": "2023-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "The first answer in every subpart is omitted from answer_units (15 in parts a/b and 19 in part c). Part (c) units also capture only the English prompts for 20-22, not the required Yanyuwa translations.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Regenerate units for items 15-22, including the actual women/men Yanyuwa forms for each part (c) answer."}, {"description": "Records are marked requires_images even though the cited Yanyuwa problem itself is textual; the image count appears to come from unrelated material on the page range.", "field": "metadata", "record_id": null, "severity": "low", "suggested_fix": "Set requires_images based on the extracted source problem span rather than whole cited pages."}], "notes": "Rendered problem/solution pages show text is faithful; answer_units are incomplete and partly wrong.", "record_ids_checked": ["iol-2024-individual-p5", "iol-2024-individual-p5-sub-a", "iol-2024-individual-p5-sub-b", "iol-2024-individual-p5-sub-c", "iol-2024-individual-p5-partial-through-a-target-b", "iol-2024-individual-p5-partial-through-ab-target-c"], "source_problem_id": "2024-individual-5", "verdict": "fix_needed"}
+{"issues": [{"description": "The Kaqchikel task depends on colored stick-figure pictures, but the dataset prompt contains only sentence text/gaps and no image descriptions or attachments. Parts (a), (b), and (c) are not answerable from the text record alone.", "field": "prompt", "record_id": null, "severity": "critical", "suggested_fix": "Attach/render the relevant PDF images or manually encode each picture scene in text."}, {"description": "Part (d) is merged into the sub-c target_question and canonical answer, so a record labeled target_subparts=[\"c\"] actually asks for and answers both (c) and (d).", "field": "subparts", "record_id": "iol-2025-individual-p5-sub-c", "severity": "high", "suggested_fix": "Split part (d) into its own subproblem/partial target or exclude it from sub-c."}, {"description": "Answer_units are incomplete/coarse: part (a) omits item 1, part (b) groups two picture-description answers as one unit, and part (c)/(d) is one large merged unit.", "field": "answer_units", "record_id": null, "severity": "high", "suggested_fix": "Create per-item units for A-I, 9-10, the four possible sentences in (c), and the three cortex predictions in (d)."}], "notes": "Rendered pages 5-6 confirm missing visual stimuli and the separate part (d).", "record_ids_checked": ["iol-2025-individual-p5", "iol-2025-individual-p5-sub-a", "iol-2025-individual-p5-sub-b", "iol-2025-individual-p5-sub-c", "iol-2025-individual-p5-partial-through-a-target-b", "iol-2025-individual-p5-partial-through-ab-target-c"], "source_problem_id": "2025-individual-5", "verdict": "fix_needed"}

benchmark/IOL/ioling_hf/reports/manual_verification/agent_batch_04_findings.jsonl

@@ -0,0 +1,22 @@
+{"source_problem_id":"2003-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2003-individual-p4","iol-2003-individual-p4-sub-assignment_1","iol-2003-individual-p4-sub-assignment_2","iol-2003-individual-p4-sub-assignment_3","iol-2003-individual-p4-sub-assignment_4","iol-2003-individual-p4-partial-through-assignment_1-target-assignment_2","iol-2003-individual-p4-partial-through-assignment_1assignment_2-target-assignment_3","iol-2003-individual-p4-partial-through-assignment_1assignment_2assignment_3-target-assignment_4"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The image-only problem page OCR does not match the dataset on target Adyghe forms; for example sentence 8 reads as a throw-form ending like \"deza\" in the rendered/OCR source, while the dataset prompt has \"dega\", and other special consonants drift in sentences 4 and 6.","suggested_fix":"Manually transcribe the Adyghe forms from the rendered PDF and correct the prompt text before using these records."},{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit required numbered answers: assignment 2 lacks item 8, assignment 3 lacks items 10 and 11, and assignment 4 collapses three possible answers into one unit.","suggested_fix":"Create separate answer_units for every numbered answer and every assignment 4 alternative, including all units in full, subproblem, and partial records."},{"severity":"medium","record_id":null,"field":"canonical","description":"The canonical solution has OCR artifacts such as \"ll.\" for item 11 and questionable Adyghe special-consonant transcription.","suggested_fix":"Recheck the rendered solution page and repair the canonical solution text."}],"notes":"Problem and solution PDFs are image-only; verification used PyMuPDF OCR. Exact special consonants remain hard to verify from OCR alone, but the prompt/unit defects are clear."}
+{"source_problem_id":"2004-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2004-individual-p2"],"issues":[{"severity":"low","record_id":null,"field":"prompt","description":"The prompt includes a stray Kayapo footnote from Problem 1 and has \"SwiftNews\" without the source-space in \"Swift News\".","suggested_fix":"Remove the unrelated Kayapo footnote from Problem 2 and restore the agency name spacing."},{"severity":"high","record_id":"iol-2004-individual-p2","field":"answer_units","description":"The five answer_units are truncated fragments: the corrected titles for 6, 7, and 9 omit essential continuation text, and the explanatory group units stop mid-sentence.","suggested_fix":"Rebuild units as complete explanation components and complete corrected titles, e.g. including \"Fraught with Tax Reform\" and the full 7 and 9 titles."}],"notes":"OCR of the solution pages matches the dataset canonical answer; the main problem is prompt contamination and unusable scoring units."}
+{"source_problem_id":"2005-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2005-individual-p3","iol-2005-individual-p3-sub-assignment_1","iol-2005-individual-p3-sub-assignment_2","iol-2005-individual-p3-partial-through-assignment_1-target-assignment_2"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"The Mansi prompt has OCR/provenance contamination: the 918 example is rendered as \"... nollowxujplow\" in the source OCR but appears as \"... hollowxujplow\" in the dataset, and an unrelated Lango footnote plus page footer are included.","suggested_fix":"Manually re-extract Problem 3 from the rendered PDF, correct the 918 numeral, and remove unrelated footnotes/footers."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Assignments 1 and 2 each contain three requested answers but are stored as one coarse unit per assignment; the assignment 2 unit also includes the solution page footer.","suggested_fix":"Split into six numbered answer_units and remove page footer text from canonical/unit values."}],"notes":"The 2005 PDF text layer is font-encoded; OCR was used for comparison."}
+{"source_problem_id":"2006-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2006-individual-p4","iol-2006-individual-p4-sub-assignment_1","iol-2006-individual-p4-sub-assignment_2","iol-2006-individual-p4-sub-assignment_3","iol-2006-individual-p4-sub-assignment_4","iol-2006-individual-p4-partial-through-assignment_1-target-assignment_2","iol-2006-individual-p4-partial-through-assignment_1assignment_2-target-assignment_3","iol-2006-individual-p4-partial-through-assignment_1assignment_2assignment_3-target-assignment_4"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Each assignment is a single coarse unit even when it contains multiple translations or yes/no judgments; assignment 4 units also include source footer text.","suggested_fix":"Split assignment 1-4 answers into individual phrase/judgment units and remove attribution/page footer text from answer values."}],"notes":"Prompt and canonical solution text align with the cited pages; scoring granularity is the blocker."}
+{"source_problem_id":"2007-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2007-individual-p5"],"issues":[{"severity":"medium","record_id":"iol-2007-individual-p5","field":"answer_units","description":"The four units pair answers 19-22 with 23-26 in the same values and do not provide separate keys/units for items 23-26.","suggested_fix":"Create eight answer_units, one for each missing Turkish/Tatar word pair from 19 through 26."}],"notes":"Prompt and canonical table are faithful to the PDF text layer, including its source glyph conventions."}
+{"source_problem_id":"2008-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2008-individual-p5","iol-2008-individual-p5-sub-a","iol-2008-individual-p5-sub-b","iol-2008-individual-p5-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"The answer_units omit the first answer in each subpart: item 13 in part (a) and item 19 in part (b) are present in the canonical solution but absent from units.","suggested_fix":"Add units for 13 \"The wolf saw your shaman\" and 19 \"Angatkuup aanniqtaatit\" across the full, subproblem, and partial records."}],"notes":"Problem and solution text otherwise match the cited PDF page."}
+{"source_problem_id":"2009-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2009-individual-p5","iol-2009-individual-p5-sub-a","iol-2009-individual-p5-sub-b","iol-2009-individual-p5-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"The units omit item 18 from part (a) and item 24 from part (b), even though both are in the canonical solution.","suggested_fix":"Add separate units for 18 and 24 in the full, subproblem, and partial records."},{"severity":"low","record_id":null,"field":"prompt","description":"The Nahuatl note is garbled into a fused line beginning \"TheClassicalNahuatl...\".","suggested_fix":"Repair the note text from the PDF so it reads as the Classical Nahuatl background and orthography note."}],"notes":"Main task data and canonical answers are otherwise faithful."}
+{"source_problem_id":"2010-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2010-individual-p5","iol-2010-individual-p5-sub-a","iol-2010-individual-p5-sub-b","iol-2010-individual-p5-sub-c","iol-2010-individual-p5-sub-d","iol-2010-individual-p5-partial-through-a-target-b","iol-2010-individual-p5-partial-through-ab-target-c","iol-2010-individual-p5-partial-through-abc-target-d"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Part (a) stores all gap fills as one unit and part (d) stores both requested plural translations as one unit, which is too coarse for RL scoring.","suggested_fix":"Split part (a) into individual gap/table-row units and part (d) into separate units for \"elms\" and \"angles\"."}],"notes":"Prompt and canonical solution match the cited page aside from minor whitespace artifacts."}
+{"source_problem_id":"2011-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2011-individual-p5","iol-2011-individual-p5-sub-a","iol-2011-individual-p5-sub-b","iol-2011-individual-p5-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The barcode source requires diagrams, barcode images, and a partially filled grid, but the dataset prompt contains only arrows/labels and omits the visual information needed to solve the problem.","suggested_fix":"Attach rendered page images or encode the barcode/grid data structurally; otherwise exclude this source from text-only records."},{"severity":"critical","record_id":null,"field":"subparts","description":"Official part (c) is not represented as its own target; the sub-b and partial-b records ask only to draw barcode (b), but their canonical answer is the textual part (c) Norway/code answer.","suggested_fix":"Create separate records for (b) and (c), or mark drawing-only (b) unsupported; align each target with its own canonical answer."},{"severity":"high","record_id":null,"field":"answer_units","description":"Part (a) units start at item 2 and omit item 1, and the current b unit combines empty (b) text with the part (c) answer.","suggested_fix":"Add item 1 to part (a) units and rebuild units after separating (b) and (c)."}],"notes":"Image dependence makes current text prompts unanswerable."}
+{"source_problem_id":"2012-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2012-individual-p5","iol-2012-individual-p5-sub-a","iol-2012-individual-p5-sub-b","iol-2012-individual-p5-sub-c","iol-2012-individual-p5-sub-d","iol-2012-individual-p5-partial-through-a-target-b","iol-2012-individual-p5-partial-through-ab-target-c","iol-2012-individual-p5-partial-through-abc-target-d"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Subpart (b) has a single unit \"poga: hole\" and omits the other four requested translations; parts (a), (c), and (d) are also grouped as coarse multi-answer units.","suggested_fix":"Create separate units for every requested body-part correspondence, every part (b) translation, every part (c) translation, and each theoretical form in part (d)."}],"notes":"PDF prompt and canonical solution text are otherwise faithful."}
+{"source_problem_id":"2013-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2013-individual-p5"],"issues":[{"severity":"medium","record_id":"iol-2013-individual-p5","field":"answer_units","description":"The full record has one large unit containing the whole explanation/table instead of separate scoreable correspondences for the six numbered words.","suggested_fix":"Split the answer into units for refrigerator, cow, bed, butterfly, spoon, and cat, with their corresponding numbers/locations."}],"notes":"Source text and canonical solution match the cited pages."}
+{"source_problem_id":"2014-team-1","verdict":"fix_needed","record_ids_checked":["iol-2014-team-p1"],"issues":[{"severity":"high","record_id":"iol-2014-team-p1","field":"answer_units","description":"Although there are 30 units, multi-line Armenian articles are truncated to their first line in answer_units; for example article 2 and articles 29-30 omit most of their official solution text.","suggested_fix":"Store the complete Armenian text for each numbered article, or split long articles into complete sub-units without dropping continuation lines."}],"notes":"Prompt and canonical solution are faithful, but unit values are incomplete for many articles."}
+{"source_problem_id":"2015-team-1","verdict":"fix_needed","record_ids_checked":["iol-2015-team-p1","iol-2015-team-p1-sub-a","iol-2015-team-p1-sub-b","iol-2015-team-p1-sub-c","iol-2015-team-p1-partial-through-a-target-b","iol-2015-team-p1-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"The official problem includes part (d) asking for a bilingual dictionary and grammar, but derived records and target_subparts only cover (a)-(c); the full record units also omit the dictionary/grammar answer material.","suggested_fix":"Add a part (d) record/partial sequence and answer_units for the dictionary/grammar solution, or explicitly mark that open-ended part unsupported while keeping provenance clear."}],"notes":"The extracted prompt and solution text for parts (a)-(c) are faithful."}
+{"source_problem_id":"2016-team-1","verdict":"fix_needed","record_ids_checked":["iol-2016-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2016-team-p1","field":"other","description":"The source is an audio matching problem requiring 114 audio files, but the dataset record has no audio assets and metadata has has_audio=false, making the prompt unanswerable.","suggested_fix":"Include the audio files/references in the dataset task or exclude this source from non-audio records."},{"severity":"high","record_id":"iol-2016-team-p1","field":"answer_units","description":"All 114 audio-word matches are stored as a single full answer unit.","suggested_fix":"Create one unit per audio-word match and a separate unit for the observations requested in part (b)."}],"notes":"PDF text for the word list and answer table is faithful; missing audio is the critical blocker."}
+{"source_problem_id":"2017-team-1","verdict":"fix_needed","record_ids_checked":["iol-2017-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2017-team-p1","field":"prompt","description":"The emoji list needed for solving is corrupted in the prompt: many emoji render as punctuation, letters, blanks, or private-use placeholders instead of the actual glyphs.","suggested_fix":"Use rendered problem pages or verified Unicode emoji data for the numbered emoji list."},{"severity":"high","record_id":"iol-2017-team-p1","field":"canonical","description":"The canonical solution also contains corrupted/missing emoji glyphs in places, such as black circle appearing as a stray letter/blank and some symbol answers lacking the glyph.","suggested_fix":"Repair canonical emoji mappings from the rendered solution PDF."},{"severity":"medium","record_id":"iol-2017-team-p1","field":"answer_units","description":"The entire 87-item solution plus extra vocabulary is one answer_unit.","suggested_fix":"Split into per-emoji mapping units and separate units for the additional requested descriptions/dictionary/grammar outputs."}],"notes":"The source is image/glyph-heavy; current text extraction is not reliably answerable."}
+{"source_problem_id":"2018-team-1","verdict":"fix_needed","record_ids_checked":["iol-2018-team-p1","iol-2018-team-p1-sub-a","iol-2018-team-p1-sub-b","iol-2018-team-p1-sub-c","iol-2018-team-p1-sub-d","iol-2018-team-p1-partial-through-a-target-b","iol-2018-team-p1-partial-through-ab-target-c","iol-2018-team-p1-partial-through-abc-target-d"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Part (a) is one very large unit for the whole matching table, and part (c) has incorrect units: item 5 is truncated and item 6 combines answers 6 and 7.","suggested_fix":"Split part (a) into individual match units; repair part (c) so items 5, 6, and 7 are complete and separately keyed."}],"notes":"Prompt and canonical solution largely follow the PDF, though table layout is dense."}
+{"source_problem_id":"2019-team-1","verdict":"fix_needed","record_ids_checked":["iol-2019-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2019-team-p1","field":"other","description":"The source states that some entries require videos, but the dataset provides only text/PDF extraction and no video assets, so the full team problem is not answerable as given.","suggested_fix":"Include the designated videos or exclude/mark the video-dependent source unsupported."},{"severity":"high","record_id":"iol-2019-team-p1","field":"answer_units","description":"All entries 49-77 and their scores/descriptions are stored as one huge full answer unit.","suggested_fix":"Create one scoreable unit per numbered answer, with score and description kept together."}],"notes":"PDF prompt/canonical text is otherwise aligned with cited pages."}
+{"source_problem_id":"2021-team-1","verdict":"fix_needed","record_ids_checked":["iol-2021-team-p1","iol-2021-team-p1-sub-a","iol-2021-team-p1-sub-b","iol-2021-team-p1-sub-c","iol-2021-team-p1-sub-d","iol-2021-team-p1-sub-e","iol-2021-team-p1-sub-f","iol-2021-team-p1-sub-g","iol-2021-team-p1-partial-through-a-target-b","iol-2021-team-p1-partial-through-ab-target-c","iol-2021-team-p1-partial-through-abc-target-d","iol-2021-team-p1-partial-through-abcd-target-e","iol-2021-team-p1-partial-through-abcde-target-f","iol-2021-team-p1-partial-through-abcdef-target-g"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Several parts are too coarse as single-table/list units, and part (f) units omit item 1, item 14, and the N-Q translation answers while truncating items 12 and 13.","suggested_fix":"Split parts (a)-(g) into granular units and rebuild part (f) to include every numbered/lettered answer completely."}],"notes":"The long prompt and canonical solution are broadly faithful to the cited PDFs; scoring units are the primary failure."}
+{"source_problem_id":"2022-team-1","verdict":"fix_needed","record_ids_checked":["iol-2022-team-p1","iol-2022-team-p1-sub-a","iol-2022-team-p1-sub-b","iol-2022-team-p1-sub-c","iol-2022-team-p1-sub-d","iol-2022-team-p1-sub-e","iol-2022-team-p1-sub-f","iol-2022-team-p1-sub-g","iol-2022-team-p1-sub-h","iol-2022-team-p1-partial-through-a-target-b","iol-2022-team-p1-partial-through-ab-target-c","iol-2022-team-p1-partial-through-abc-target-d","iol-2022-team-p1-partial-through-abcd-target-e","iol-2022-team-p1-partial-through-abcde-target-f","iol-2022-team-p1-partial-through-abcdef-target-g","iol-2022-team-p1-partial-through-abcdefg-target-h"],"issues":[{"severity":"critical","record_id":null,"field":"prompt","description":"The Manchu script material in the prompt is garbled by font extraction, especially the initial manuscript excerpt; this is core data for the problem.","suggested_fix":"Provide rendered images or manually verified Unicode/structured Manchu script text for all Manchu-script passages."},{"severity":"critical","record_id":null,"field":"canonical","description":"The canonical answer for part (c) is also garbled Manchu script and includes extraction artifacts rather than reliably readable target script.","suggested_fix":"Repair part (c) from the rendered solution PDF or include the answer as an image/validated Unicode text."},{"severity":"high","record_id":null,"field":"answer_units","description":"Units omit answers b54, f90, and h112; part (f) units capture only source Manchu prompts rather than full translated dialogue answers, and some units include headers/chart spillover.","suggested_fix":"Rebuild answer_units from the canonical solution with one complete unit per requested item and no page headers or explanatory charts inside item values."}],"notes":"Latin transcription sections are mostly readable, but the Manchu-script extraction is not faithful enough for answerability."}
+{"source_problem_id":"2023-team-1","verdict":"fix_needed","record_ids_checked":["iol-2023-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2023-team-p1","field":"answer_units","description":"The canonical answer contains the full page 57-61 matching key plus the 60-T-54 restoration, but answer_units contain only the 60-T-54 restoration; almost all required mappings are unscored.","suggested_fix":"Create answer_units for every dictionary-entry match on pages 57-61 and a separate unit for the restored 60-T-54 class."},{"severity":"medium","record_id":"iol-2023-team-p1","field":"prompt","description":"The long dictionary tables are flattened with many adjacent entries concatenated, making parts of the prompt hard to parse.","suggested_fix":"Preserve table layout or encode dictionary entries as structured rows."}],"notes":"Solution page range matches the canonical answer; unit coverage is the severe defect."}
+{"source_problem_id":"2024-team-1","verdict":"fix_needed","record_ids_checked":["iol-2024-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2024-team-p1","field":"answer_units","description":"Only three garbled units were extracted for a multi-assignment team problem, and they do not correspond to the official assignment answers.","suggested_fix":"Rebuild answer_units by assignment from the solution material, with complete values for each final answer/tree/distance/procedure."},{"severity":"high","record_id":"iol-2024-team-p1","field":"canonical","description":"The cited solution is a slide deck with graphical trees; the dataset canonical is a flattened slide text dump rather than a clean, source-faithful answer text.","suggested_fix":"Use rendered tree images or a structured representation of each tree and extract a clean answer key from the slides."}],"notes":"Problem text is mostly present, but tree/slide layout and units are not reliable for scoring."}
+{"source_problem_id":"2025-team-1","verdict":"fix_needed","record_ids_checked":["iol-2025-team-p1","iol-2025-team-p1-sub-a","iol-2025-team-p1-sub-b","iol-2025-team-p1-sub-c","iol-2025-team-p1-sub-d","iol-2025-team-p1-sub-e","iol-2025-team-p1-sub-g","iol-2025-team-p1-sub-h","iol-2025-team-p1-partial-through-a-target-b","iol-2025-team-p1-partial-through-ab-target-c","iol-2025-team-p1-partial-through-abc-target-d","iol-2025-team-p1-partial-through-abcd-target-e","iol-2025-team-p1-partial-through-abcde-target-g","iol-2025-team-p1-partial-through-abcdeg-target-h"],"issues":[{"severity":"critical","record_id":null,"field":"subparts","description":"Official part (f) is merged into the part (e) canonical/unit, and there is no sub-f record or partial variant; the sequence jumps from e to g.","suggested_fix":"Split parts (e) and (f), add sub-f and appropriate partial records, and align visible solved parts accordingly."},{"severity":"high","record_id":null,"field":"answer_units","description":"Large matching parts (a)-(d) and (h) are single coarse units, part (g) omits item 1, and the part (e) unit includes part (f) plus page footer text.","suggested_fix":"Create granular units for every numbered/lettered answer, add g item 1, remove footers, and separate e/f units."}],"notes":"Prompt and overall solution content are mostly present, but PDF side-by-side layout caused serious subpart and unit corruption."}

benchmark/IOL/ioling_hf/reports/manual_verification/agent_batch_05_findings.jsonl

@@ -0,0 +1,21 @@
+{"source_problem_id":"2003-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2003-individual-p5"],"issues":[{"severity":"high","record_id":"iol-2003-individual-p5","field":"prompt","description":"The French table's shaded/no-counterpart cells are OCRed as long nonsense strings instead of being represented as shaded or empty cells, so the model-facing table is not faithful to the PDF.","suggested_fix":"Re-extract or manually transcribe the table, preserving blank prefixes and shaded no-counterpart cells in a consistent textual notation."},{"severity":"high","record_id":"iol-2003-individual-p5","field":"canonical","description":"The solution answer lines contain OCR garbage in the shaded-cell positions, including corrupted strings after entries such as reformer/refuter/remunerate/distribute.","suggested_fix":"Replace the corrupted solution table with a clean transcription of the intended completed rows and explanation."},{"severity":"medium","record_id":"iol-2003-individual-p5","field":"answer_units","description":"The single full.1 answer unit covers all gap fills plus the explanation, which is too coarse for RL scoring of the individual table entries.","suggested_fix":"Split the expected answer into separate units for each filled gap and a separate explanation/rule unit."}],"notes":"Problem and solution page ranges point to the right PDFs/pages, but the OCR table corruption requires source repair."}
+{"source_problem_id":"2004-individual-3","verdict":"fix_needed","record_ids_checked":["iol-2004-individual-p3","iol-2004-individual-p3-sub-assignment_1","iol-2004-individual-p3-sub-assignment_2","iol-2004-individual-p3-partial-through-assignment_1-target-assignment_2"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"Latin long-vowel marks in the problem word lists are corrupted into placeholder punctuation such as d$vidit, mord%x, f&mus, frac's, fov're, rubr$ca, combr'tum, and d'beo.","suggested_fix":"Manually restore the Latin forms with their proper macrons/long-vowel notation throughout all derived prompts."},{"severity":"high","record_id":null,"field":"canonical","description":"The Assignment 2 solution includes OCR/footer debris after the actual answer and several character confusions in the explanatory solution text.","suggested_fix":"Clean the official solution text so Assignment 2 ends after the debeo explanation and remove unrelated footer/OCR fragments."},{"severity":"medium","record_id":"iol-2004-individual-p3-sub-assignment_2","field":"answer_units","description":"Assignment 2 is represented as one unit even though it asks which of six words are excluded and requires multiple answer items.","suggested_fix":"Split Assignment 2 into separate answer units for brevis, gurdus, debeo, and the shared rationale or acceptance criteria."}],"notes":"Partial variant shows only Assignment 1 before targeting Assignment 2, but both prompts and solution text need cleanup."}
+{"source_problem_id":"2005-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2005-individual-p4","iol-2005-individual-p4-sub-assignment_1","iol-2005-individual-p4-sub-assignment_2","iol-2005-individual-p4-partial-through-assignment_1-target-assignment_2"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"Yoruba special consonants/vowels and tone marks are corrupted in the prompt, e.g. bracket loss around a city-witch phrase, s/y substitutions, and note symbols rendered as 3, c, or 0 rather than the source characters.","suggested_fix":"Manually transcribe the Yoruba forms and note from the rendered PDF, preserving brackets, special letters, and tone marks."},{"severity":"high","record_id":null,"field":"canonical","description":"The solution answers carry the same OCR damage, e.g. [oko sya], laze elu], [ila a3e], missing spacing in 'house of love', and a page footer included in the answer.","suggested_fix":"Replace the solution answer blocks with clean Yoruba forms from the PDF and remove page footer text from canonical answers."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Each assignment answer unit bundles four separate translations, which is too coarse for item-level scoring.","suggested_fix":"Create one answer unit per numbered item 14-21."}],"notes":"PDF provenance is correct, but table/text transcription is not reliable enough as-is."}
+{"source_problem_id":"2006-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2006-individual-p5"],"issues":[{"severity":"medium","record_id":"iol-2006-individual-p5","field":"answer_units","description":"The prompt and solution text match the cited pages, but the single full.1 unit covers translations, error diagnoses, and four corrected Ngoni sentences for each case.","suggested_fix":"Split the expected answer into units for each erroneous sentence's English translation, error explanation, and accepted corrected Ngoni alternatives."}],"notes":"Source text and page ranges verified against pages 5 of the problem and solution PDFs."}
+{"source_problem_id":"2007-team-1","verdict":"fix_needed","record_ids_checked":["iol-2007-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2007-team-p1","field":"prompt","description":"The genealogical tree diagram on problem page 1 is omitted from the dataset prompt. The prompt only says a tree is presented, but does not include the empty-box tree structure to fill.","suggested_fix":"Include a rendered image reference or a faithful textual encoding of the family-tree graph and empty boxes in the prompt."},{"severity":"medium","record_id":"iol-2007-team-p1","field":"answer_units","description":"The single full.1 unit is a flattened graphical tree answer for both requested tasks, too coarse for scoring tree reconstruction and the eleven numbered members separately.","suggested_fix":"Add structured answer units for each tree position/name and for each of the eleven numbered family members."}],"notes":"The missing diagram makes the source record unanswerable from text alone."}
+{"source_problem_id":"2008-team-1","verdict":"fix_needed","record_ids_checked":["iol-2008-team-p1","iol-2008-team-p1-sub-a","iol-2008-team-p1-sub-b","iol-2008-team-p1-sub-c","iol-2008-team-p1-sub-d","iol-2008-team-p1-sub-e","iol-2008-team-p1-sub-f","iol-2008-team-p1-sub-g","iol-2008-team-p1-sub-h","iol-2008-team-p1-sub-i","iol-2008-team-p1-sub-j","iol-2008-team-p1-partial-through-a-target-b","iol-2008-team-p1-partial-through-ab-target-c","iol-2008-team-p1-partial-through-abc-target-d","iol-2008-team-p1-partial-through-abcd-target-e","iol-2008-team-p1-partial-through-abcde-target-f","iol-2008-team-p1-partial-through-abcdef-target-g","iol-2008-team-p1-partial-through-abcdefg-target-h","iol-2008-team-p1-partial-through-abcdefgh-target-i","iol-2008-team-p1-partial-through-abcdefghi-target-j"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Several answer units bundle multiple numbered outputs, especially h (six missing tones), i (four Cantonese readings), and j (eleven Mandarin readings), so scoring is too coarse.","suggested_fix":"Split h, i, and j into one unit per numbered transcription/tone answer while keeping explanatory subparts as separate rubric units."}],"notes":"Prompt page ranges, target labels, and partial-variant shown-solution sequencing match the cited PDFs."}
+{"source_problem_id":"2009-team-1","verdict":"fix_needed","record_ids_checked":["iol-2009-team-p1"],"issues":[{"severity":"medium","record_id":"iol-2009-team-p1","field":"prompt","description":"The PDF highlights frequent Vietnamese words in bold inside the reading units, and the prompt explicitly refers to highlighted words; the dataset prompt flattens away that highlighting.","suggested_fix":"Preserve highlighted target words with markup, e.g. bold Markdown or another explicit token-level marker."},{"severity":"medium","record_id":"iol-2009-team-p1","field":"answer_units","description":"The full answer is one giant unit covering translations for ten reading units, which is far too coarse for scoring.","suggested_fix":"Split answer units by reading unit and preferably by sentence or highlighted vocabulary item."}],"notes":"Vietnamese text and solution pages are otherwise present; Wroclaw/font quirks are minor compared with lost highlighting."}
+{"source_problem_id":"2010-team-1","verdict":"fix_needed","record_ids_checked":["iol-2010-team-p1"],"issues":[{"severity":"medium","record_id":"iol-2010-team-p1","field":"answer_units","description":"The single full.1 unit covers all phrase translations and the large Mongolian lexicon table, which is too coarse for RL scoring.","suggested_fix":"Split part (a) phrase translations into separate units and represent part (b) lexicon entries as separate optional/graded units."}],"notes":"Problem and solution text are faithful to the cited pages."}
+{"source_problem_id":"2011-team-1","verdict":"fix_needed","record_ids_checked":["iol-2011-team-p1","iol-2011-team-p1-sub-a","iol-2011-team-p1-sub-b","iol-2011-team-p1-sub-c","iol-2011-team-p1-partial-through-a-target-b","iol-2011-team-p1-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":"iol-2011-team-p1-sub-b","field":"answer_units","description":"Part (b) answer units are incomplete: they omit the deleted words 'jarau, lagau' and several added/deleted macron corrections, even though the canonical text includes them.","suggested_fix":"Create answer units for the two deleted words, three changed letters, four added-macron removals, and five deleted-macron restorations."},{"severity":"high","record_id":"iol-2011-team-p1","field":"answer_units","description":"The full-problem answer units inherit the incomplete part (b) coverage, so the full record cannot score all requested corrections.","suggested_fix":"Use the same complete granular part (b) units in the full and partial-through-a records."}],"notes":"The timed-hint pages are in the cited problem PDF and are reflected in the long prompts; the main defect is expected-answer coverage."}
+{"source_problem_id":"2012-team-1","verdict":"fix_needed","record_ids_checked":["iol-2012-team-p1"],"issues":[{"severity":"critical","record_id":"iol-2012-team-p1","field":"prompt","description":"The prompt omits the entire list of 57 Lao country names visible on the source PDF page, leaving only the sentence 'Here are the names of 57 countries in Lao'.","suggested_fix":"Add a rendered image or faithful Unicode/manual transcription of all 57 Lao names to the prompt."},{"severity":"high","record_id":"iol-2012-team-p1","field":"answer_units","description":"The answer units are only four large, mis-grouped chunks of the 57-country solution rather than item-level country/pronunciation answers.","suggested_fix":"Split into one unit per country identification and pronunciation, or at least one unit per numbered country."}],"notes":"This record is unanswerable without the omitted Lao list."}
+{"source_problem_id":"2014-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2014-individual-p1","iol-2014-individual-p1-sub-a","iol-2014-individual-p1-sub-b","iol-2014-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"The source uses the letter sequence 'fi' in many Benabena forms, but the PDF text extraction lost the ligature/control character, producing blanks such as 'noli 'ibe', 'no ne', 'noni be', and 'li latagihe'.","suggested_fix":"Manually restore all lost 'fi' sequences in examples and targets, e.g. nolifi'ibe, nofine, nifila'ibe, nonifibe, lifilatagihe."},{"severity":"high","record_id":null,"field":"canonical","description":"The same 'fi' loss appears in the solution rules and answers, including the pierce root and answers such as kifilune and filanagihe.","suggested_fix":"Clean the canonical solution and answer units against the rendered solution PDF."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Subpart (a) bundles five translations in one unit and subpart (b) bundles four forms in one unit.","suggested_fix":"Split each requested item into its own answer unit."}],"notes":"The partial variant sequencing is otherwise correct."}
+{"source_problem_id":"2015-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2015-individual-p1","iol-2015-individual-p1-sub-b","iol-2015-individual-p1-sub-c","iol-2015-individual-p1-partial-through-b-target-c"],"issues":[{"severity":"high","record_id":"iol-2015-individual-p1","field":"answer_units","description":"The full-problem prompt asks parts (a), (b), and (c), but the answer_units cover only b and c; part (a)'s numerical equalities are missing from scoring.","suggested_fix":"Add granular answer units for all sixteen items in part (a), and add a sub-a record if subproblem coverage is intended."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Parts (b) and (c) each bundle two requested translations in a single unit.","suggested_fix":"Split b into separate units for 42 and 494, and c into separate units for 43 and 569."}],"notes":"Prompt and canonical solution text match the cited pages; the defect is in record coverage/scoring granularity."}
+{"source_problem_id":"2016-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2016-individual-p1"],"issues":[{"severity":"critical","record_id":"iol-2016-individual-p1","field":"prompt","description":"The map of Salu Leang hamlets, roads, river, bridges, stream direction, and relative heights is missing from the dataset prompt; only the legend text survives.","suggested_fix":"Include a rendered map image or a faithful textual graph with hamlet positions, road/river connections, bridges, and elevation numbers."}],"notes":"The 15 answer units are granular, but the prompt is not answerable without the map."}
+{"source_problem_id":"2017-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2017-individual-p1","iol-2017-individual-p1-sub-a","iol-2017-individual-p1-sub-b","iol-2017-individual-p1-sub-c","iol-2017-individual-p1-partial-through-a-target-b","iol-2017-individual-p1-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"Superscript exponents in the Birom equalities are flattened or line-broken, e.g. powers are shown as adjacent digits or misplaced words, and the a^k explanatory formula is garbled.","suggested_fix":"Represent powers explicitly with caret notation or Unicode superscripts in every equality and in the formula note."},{"severity":"high","record_id":null,"field":"answer_units","description":"Part (a) answer units omit equality 1 (52 + 3 + 4 = 32); parts (b) and (c) also bundle multiple requested outputs into single units.","suggested_fix":"Add the missing item 1 unit and split parts (b) and (c) into one unit per number/equality/output."}],"notes":"Canonical solution text itself contains the missing item 1, so the main scoring defect is answer_units."}
+{"source_problem_id":"2018-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2018-individual-p1"],"issues":[{"severity":"medium","record_id":"iol-2018-individual-p1","field":"canonical","description":"The official solution's stress-rule formula is flattened into a largely unreadable text block, losing the visual structure of the PDF.","suggested_fix":"Manually rewrite the rule formula in clear textual notation or preserve it as a rendered image plus text."},{"severity":"medium","record_id":"iol-2018-individual-p1","field":"answer_units","description":"The single answer unit bundles all twelve stress-marking targets, which is too coarse for RL scoring.","suggested_fix":"Create one answer unit for each target word's correctly stressed form."}],"notes":"The problem prompt is usable; the solution/rubric structure needs cleanup."}
+{"source_problem_id":"2019-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2019-individual-p1","iol-2019-individual-p1-sub-a","iol-2019-individual-p1-sub-b","iol-2019-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first requested item in each subpart: item 12 for part (a) and item 18 for part (b), although both appear in the canonical answer.","suggested_fix":"Add answer units for item 12 ('Om benu aneen' translation) and item 18 ('We will go away and they will come' in Yonggom), and propagate them to full/sub/partial records."}],"notes":"Prompt text, canonical answers, subpart labels, and partial variant are otherwise faithful."}
+{"source_problem_id":"2021-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2021-individual-p1","iol-2021-individual-p1-sub-a","iol-2021-individual-p1-sub-b","iol-2021-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Part (a) is one unit for three digit answers and part (b) is one unit for five Ekari outputs, so scoring is too coarse.","suggested_fix":"Split part (a) into units for each numeral and part (b) into units for 1, 19, 26, 104, and 292."}],"notes":"Prompt, solution text, and partial-variant shown solution are faithful to the cited pages."}
+{"source_problem_id":"2022-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2022-individual-p1","iol-2022-individual-p1-sub-a","iol-2022-individual-p1-sub-b","iol-2022-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"The source PDF has separate parts (b) and (c), but the dataset collapses both into target subpart b and has no separate c record; sub-b prompts ask both 'Translate into English' and 'Translate into Ubykh'.","suggested_fix":"Create a distinct c subproblem/partial target for items 13-17 and keep b limited to items 10-12."},{"severity":"high","record_id":null,"field":"answer_units","description":"The combined b answer units omit the first item of each source subpart: item 10 'aš'əntʷən — he gives him to us' and item 13 'they give youpl to me — ŝʷəsənatʷan'.","suggested_fix":"Add missing units for items 10 and 13 and relabel all b/c units to match the source subparts."}],"notes":"Prompt text for the full record is faithful, but derived subpart structure is not."}
+{"source_problem_id":"2023-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2023-individual-p1","iol-2023-individual-p1-sub-a","iol-2023-individual-p1-sub-b","iol-2023-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"critical","record_id":"iol-2023-individual-p1-sub-b","field":"prompt","description":"The sub-b prompt asks to fill rows 39-48 but omits the entire second table containing rows 19-48 from the source PDF, so the target is not answerable.","suggested_fix":"Include the second verb-form table (rows 19-48) in the b prompt before the target question."},{"severity":"critical","record_id":"iol-2023-individual-p1-partial-through-a-target-b","field":"prompt","description":"The partial-through-a target-b prompt has the same omission of rows 19-48, even though those rows are required for solving b.","suggested_fix":"Regenerate the partial b prompt with rows 19-48 included after the shown solution for part a."}],"notes":"The full-problem record contains the missing table and has granular answer units; only the b-derived prompts are broken."}
+{"source_problem_id":"2024-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2024-individual-p1"],"issues":[{"severity":"high","record_id":"iol-2024-individual-p1","field":"answer_units","description":"The answer_units include explanatory rule-structure bullets as required units and omit actual answer items 19 and 24 from the requested translations.","suggested_fix":"Remove rule bullets from required answer_units or mark them as rationale, and add answer units for item 19 (kullenan/kulleŋən translation) and item 24 (you(sg) will see them(du))."},{"severity":"medium","record_id":"iol-2024-individual-p1","field":"answer_units","description":"Only one full-problem record exists even though the source has parts (a) and (b), limiting subpart-level evaluation.","suggested_fix":"Add subproblem records or at least label answer units under source subparts a and b."}],"notes":"Prompt and canonical solution text match the cited Koryak pages."}
+{"source_problem_id":"2025-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2025-individual-p1","iol-2025-individual-p1-sub-a","iol-2025-individual-p1-sub-b","iol-2025-individual-p1-sub-c","iol-2025-individual-p1-partial-through-a-target-b","iol-2025-individual-p1-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"canonical","description":"Several solution equalities involving fractions are flattened incorrectly in canonical text, e.g. part (b)(5) appears as '2 x 43 + 21 = 2' instead of using 3/4 and 1/2, and (b)(6) appears with '1100 x 12' instead of '1100 x 1/2'.","suggested_fix":"Manually transcribe the fraction expressions from the rendered solution PDF using explicit slash notation such as 3/4 and 1/2."},{"severity":"high","record_id":null,"field":"answer_units","description":"Part (a) answer_units omit [X] = 60 = ɖukcu; part (b) is a single unit for ten equalities and part (c) a single unit for two outputs.","suggested_fix":"Add the missing X unit and split b into ten units and c into separate units for 75 and 570."}],"notes":"Problem prompt content is mostly faithful, though the two-column assignment/note text is interleaved in subpart prompts."}

benchmark/IOL/ioling_hf/reports/manual_verification/agent_batch_06_findings.jsonl

@@ -0,0 +1,21 @@
+{"source_problem_id":"2003-team-1","verdict":"fix_needed","record_ids_checked":["iol-2003-team-p1"],"issues":[{"severity":"high","record_id":"iol-2003-team-p1","field":"prompt","description":"The scanned source page is OCR-corrupted in the model prompt: multiple Proto-Tocharian forms and the note symbols are misread or replaced by placeholder characters, affecting core linguistic data.","suggested_fix":"Re-extract from rendered pages or manually transcribe the problem, preserving the special vowel/length/ng symbols and the reconstructed word forms."},{"severity":"medium","record_id":"iol-2003-team-p1","field":"answer_units","description":"The only answer unit is a short fragment of Assignment 1 and does not cover the language allocation, proto-form reconstructions, or stress explanation required by the full problem.","suggested_fix":"Split answer_units by the required outputs for Assignments 1, 2(a-c), and 3, with normalized acceptable values for each."}],"notes":"Problem and solution PDFs are image-only scans; rendered OCR was sufficient to identify definite prompt and scoring issues."}
+{"source_problem_id":"2004-individual-4","verdict":"fix_needed","record_ids_checked":["iol-2004-individual-p4","iol-2004-individual-p4-sub-assignment_1","iol-2004-individual-p4-sub-assignment_2","iol-2004-individual-p4-sub-assignment_3","iol-2004-individual-p4-partial-through-assignment_1-target-assignment_2","iol-2004-individual-p4-partial-through-assignment_1assignment_2-target-assignment_3"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"The Lakhota problem text has corrupted special characters in key data and target words, e.g. placeholder characters appear in forms that the rendered source shows as distinct Lakhota symbols.","suggested_fix":"Manually transcribe the Lakhota forms and notes from the rendered PDF, preserving the source transliteration consistently."},{"severity":"high","record_id":null,"field":"canonical","description":"The solution text and canonical answers contain OCR garbage in the explanatory table and answer strings, including corrupted fricative symbols and several damaged word forms.","suggested_fix":"Replace the solution_text and canonical subpart answers with a clean transcription of the official solution for Problem 4."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Assignments 1 and 2 are each represented as one free-text unit even though they require multiple separate matches/translations, which is too coarse for RL scoring.","suggested_fix":"Create separate answer_units for each matched item in Assignment 1 and each translation requested in Assignment 2."}],"notes":"The cited pages are the intended problem and solution pages, but custom-font/OCR corruption affects both prompt and answer fidelity."}
+{"source_problem_id":"2005-individual-5","verdict":"fix_needed","record_ids_checked":["iol-2005-individual-p5","iol-2005-individual-p5-sub-assignment_1","iol-2005-individual-p5-sub-assignment_2","iol-2005-individual-p5-sub-assignment_3","iol-2005-individual-p5-partial-through-assignment_1-target-assignment_2","iol-2005-individual-p5-partial-through-assignment_1assignment_2-target-assignment_3"],"issues":[{"severity":"high","record_id":null,"field":"prompt","description":"The Lithuanian accent marks and paradigm table are corrupted in the prompt, replacing source accent/intonation notation with ambiguous or wrong characters. This damages the core data needed to solve the problem.","suggested_fix":"Rebuild the prompt from a manual transcription that preserves the Lithuanian accent marks, nasal notation, and table entries."},{"severity":"high","record_id":null,"field":"canonical","description":"The official solution/canonical text loses the crucial accent-pattern notation, reducing it to ambiguous strings such as aa, dd, ad, and da instead of the source-marked patterns.","suggested_fix":"Transcribe the solution table and Saussure-law explanation with faithful accent and stress notation."},{"severity":"medium","record_id":"iol-2005-individual-p5-partial-through-assignment_1-target-assignment_2","field":"subparts","description":"The partial variant for Assignment 2 shows extra reasoning under the solved Assignment 1 block that belongs to deriving Assignment 2, giving target-relevant information before the target question.","suggested_fix":"Limit the visible solved Assignment 1 block to the actual answer to Assignment 1, or relabel/split the official solution spans more carefully."}],"notes":"Rendered OCR confirms the intended source pages, but the dataset text does not faithfully preserve the notation that makes the problem answerable."}
+{"source_problem_id":"2007-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2007-individual-p1"],"issues":[{"severity":"critical","record_id":"iol-2007-individual-p1","field":"prompt","description":"The Braille dot diagrams from the source problem are absent from the model-facing prompt, leaving only blank space after the example sentences; the task is not answerable.","suggested_fix":"Include rendered Braille diagrams as images or encode each Braille cell in an unambiguous textual representation."},{"severity":"critical","record_id":"iol-2007-individual-p1","field":"canonical","description":"The canonical answer and answer_unit are for 2007 Problem 2 (Movima), while this source problem is 2007 Problem 1 (Braille).","suggested_fix":"Replace the solution provenance and canonical answer with the official Braille solution, or remove the record until a correct source solution is available."},{"severity":"critical","record_id":"iol-2007-individual-p1","field":"provenance","description":"The cited solution page range points to the Movima Problem 2 solution, not the Braille Problem 1 solution.","suggested_fix":"Correct the solution page range/source or mark the Braille solution as unavailable if it is not present in the local PDF."}],"notes":"The problem PDF page is the Braille problem, but the solution PDF starts at Problem 2 and the dataset copied that wrong solution."}
+{"source_problem_id":"2008-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2008-individual-p1","iol-2008-individual-p1-sub-a","iol-2008-individual-p1-sub-b","iol-2008-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"Subparts (a) and (b) each have a single free-text answer unit despite requiring five transcriptions and four orthographic forms respectively.","suggested_fix":"Split answer_units into one unit per numbered item 12-20."}],"notes":"Prompt and canonical answer match the cited Micmac source closely; scoring granularity is the blocking issue."}
+{"source_problem_id":"2009-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2009-individual-p1"],"issues":[{"severity":"medium","record_id":"iol-2009-individual-p1","field":"answer_units","description":"The full-problem record uses one free-text answer unit for all eight requested translations plus explanation, which is too coarse for RL scoring.","suggested_fix":"Create separate units for each item in parts (a) and (b), with explanation/rules either separate or optional."},{"severity":"low","record_id":"iol-2009-individual-p1","field":"prompt","description":"The language note/footer text is garbled, although the core Sulka examples and questions are readable.","suggested_fix":"Clean the source note text when rebuilding the prompt."}],"notes":"Cited pages and canonical content are otherwise faithful to Problem 1."}
+{"source_problem_id":"2010-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2010-individual-p1"],"issues":[{"severity":"medium","record_id":"iol-2010-individual-p1","field":"answer_units","description":"A single full-problem answer unit contains all required Budukh table fills and rules, making scoring too coarse.","suggested_fix":"Split answer_units by each required vacant cell/form and keep rules separate from final filled forms."},{"severity":"low","record_id":null,"field":"provenance","description":"The manifest problem page range includes page 2, which contains later problems rather than Problem 1.","suggested_fix":"Set the problem page range to page 1 only for this source problem."}],"notes":"The prompt and canonical answer are faithful enough for Problem 1, aside from coarse scoring and overbroad problem provenance."}
+{"source_problem_id":"2011-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2011-individual-p1","iol-2011-individual-p1-sub-a","iol-2011-individual-p1-sub-b","iol-2011-individual-p1-partial-through-a-target-b"],"issues":[{"severity":"medium","record_id":null,"field":"answer_units","description":"The Menominee answer units are grouped by subpart, but subpart (a) has three translations and subpart (b) has four translations, so the units are too coarse for RL scoring.","suggested_fix":"Create one answer_unit for each requested form/translation in parts (a) and (b)."}],"notes":"Problem prompt, partial variants, and canonical answers match the official source; scoring granularity needs correction."}
+{"source_problem_id":"2012-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2012-individual-p1","iol-2012-individual-p1-sub-a","iol-2012-individual-p1-sub-b","iol-2012-individual-p1-sub-c","iol-2012-individual-p1-partial-through-a-target-b","iol-2012-individual-p1-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first required item in multiple subparts: part (b) starts at 18 instead of 17 and part (c) starts at 21 instead of 20. The full record has the same omissions.","suggested_fix":"Add answer_units for 17 and 20 and verify all Dyirbal numbered answers are represented individually."}],"notes":"Canonical solution text is faithful to the cited Dyirbal solution; the main defect is incomplete scoring units."}
+{"source_problem_id":"2013-individual-1","verdict":"fix_needed","record_ids_checked":["iol-2013-individual-p1","iol-2013-individual-p1-sub-a","iol-2013-individual-p1-sub-b","iol-2013-individual-p1-sub-c","iol-2013-individual-p1-partial-through-a-target-b","iol-2013-individual-p1-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Part (c) is represented by a single unit containing only the last two translations, omitting most requested Yidiny outputs. Part (b) is also a single unit for four outputs.","suggested_fix":"Create separate answer_units for every requested item in parts (a), (b), and (c), especially all eight translations in part (c)."}],"notes":"Prompt and canonical text are otherwise consistent with the cited source pages."}
+{"source_problem_id":"2014-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2014-individual-p2"],"issues":[{"severity":"critical","record_id":"iol-2014-individual-p2","field":"answer_units","description":"The only answer unit is a fragment of the rule table (fruit suffixes) rather than the filled Kiowa noun table, so the expected answer is effectively unscoreable.","suggested_fix":"Replace the full.1 unit with separate answer_units for each question-mark cell in the Kiowa table."},{"severity":"low","record_id":"iol-2014-individual-p2","field":"prompt","description":"Minor text extraction damage appears in the prompt, such as the word fish rendered as sh.","suggested_fix":"Clean ligature/font extraction artifacts in the prompt text."}],"notes":"Canonical answer contains the official filled table, but answer_units do not capture it."}
+{"source_problem_id":"2015-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2015-individual-p2","iol-2015-individual-p2-sub-a","iol-2015-individual-p2-sub-b","iol-2015-individual-p2-sub-c","iol-2015-individual-p2-partial-through-a-target-b","iol-2015-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"critical","record_id":"iol-2015-individual-p2-sub-b","field":"prompt","description":"The standalone part (b) prompt omits the second block of Beslenei examples that the source provides before question (b), so the target is not answerable from the shown data.","suggested_fix":"Include the additional Beslenei forms block before part (b) in the sub-b prompt."},{"severity":"critical","record_id":"iol-2015-individual-p2-partial-through-a-target-b","field":"prompt","description":"The partial variant targeting part (b) also omits the second Beslenei examples block, so the partial record remains not answerable despite showing part (a).","suggested_fix":"Include the omitted examples in the partial-through-a target-b prompt."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Each subpart is scored as a single free-text unit even though parts (a), (b), and (c) each require multiple separate translations/forms.","suggested_fix":"Split answer_units by each requested Kabardian/Beslenei item."}],"notes":"Full and other subpart prompts have the intended source text; the target-b prompt slicing is the critical failure."}
+{"source_problem_id":"2016-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2016-individual-p2","iol-2016-individual-p2-sub-a","iol-2016-individual-p2-sub-b","iol-2016-individual-p2-sub-c","iol-2016-individual-p2-partial-through-a-target-b","iol-2016-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Part (c) answer_units omit item 1 (the acute-sign function) in the full, sub-c, and partial-through-ab records. Part (b) is also one unit for two answers.","suggested_fix":"Add a unit for c.1 and split part (b) into two units, then verify all 13 matches in part (a) remain covered."}],"notes":"The Luwian symbol font is represented by extracted glyph codes, but the official letter labels and canonical correspondences are verifiable; scoring units need repair."}
+{"source_problem_id":"2017-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2017-individual-p2","iol-2017-individual-p2-sub-a","iol-2017-individual-p2-sub-b","iol-2017-individual-p2-sub-c","iol-2017-individual-p2-partial-through-a-target-b","iol-2017-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first required answer in part (a) (1. a-bang) and part (c) (1. pig), and part (b) is a single unit for two translations.","suggested_fix":"Add the missing first units and split all multi-answer units so every requested Abui item is scoreable."}],"notes":"Source prompt and canonical answers are faithful; incomplete answer_units are the main issue."}
+{"source_problem_id":"2018-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2018-individual-p2","iol-2018-individual-p2-sub-b"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"The source problem has parts (a) and (b), but the derived records include only a full record and a sub-b record; there is no subproblem record for part (a).","suggested_fix":"Add a subproblem record for part (a), or explain intentionally omitted subparts in metadata."},{"severity":"high","record_id":null,"field":"answer_units","description":"The full record answer_units cover only part (b) items 8-10, omitting all part (a) answers and part (b) item 7; the sub-b record also omits item 7.","suggested_fix":"Create answer_units for part (a) items 1-6 and part (b) items 7-10."}],"notes":"Prompt and canonical text are faithful to Hakhun Problem 2, but the derived subparts and scoring units are incomplete."}
+{"source_problem_id":"2019-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2019-individual-p2","iol-2019-individual-p2-sub-a","iol-2019-individual-p2-sub-b","iol-2019-individual-p2-sub-c","iol-2019-individual-p2-partial-through-a-target-b","iol-2019-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"Source parts (c) and (d) are merged into records labeled only target c; the target question says solve (c) only but includes the (d) translation requests, and no separate d record exists.","suggested_fix":"Split source parts (c) and (d) into distinct target_subparts/records or relabel the merged target as c+d consistently."},{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units repeatedly omit the first item of each section: a.1 for source item 1, b.1 for item 19, c.1 for item 26, and the first part (d) item 30 are missing from scoring.","suggested_fix":"Add units for source items 1, 19, 26, and 30 and verify all Yurok outputs are represented individually."}],"notes":"Canonical content matches the official Yurok solution, but subpart metadata and answer_units are incomplete/misgrouped."}
+{"source_problem_id":"2021-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2021-individual-p2","iol-2021-individual-p2-sub-a","iol-2021-individual-p2-sub-b","iol-2021-individual-p2-sub-c","iol-2021-individual-p2-sub-e","iol-2021-individual-p2-partial-through-a-target-b","iol-2021-individual-p2-partial-through-ab-target-c","iol-2021-individual-p2-partial-through-abc-target-e"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"Source part (d) is merged into records labeled target c, while the target question for sub-c asks only part (c); no separate d record exists.","suggested_fix":"Split part (d) into its own subproblem/partial target or label the merged record as covering both c and d."},{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first answer in parts (a), (b), (c), and (d), and the c units combine c and d answers into single paired values.","suggested_fix":"Add units for source items 1, 11, 20, and 26, and separate part (c) English translations from part (d) Zuni translations."}],"notes":"The cited source pages are correct; the table layout caused c/d answer merging and missing first-item units."}
+{"source_problem_id":"2022-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2022-individual-p2","iol-2022-individual-p2-sub-a","iol-2022-individual-p2-sub-b","iol-2022-individual-p2-sub-d","iol-2022-individual-p2-partial-through-a-target-b","iol-2022-individual-p2-partial-through-ab-target-d"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"Source parts (b) and (c) are merged into records labeled only target b; the prompt says solve (b) only while including the (c) Alabama translations.","suggested_fix":"Split part (c) into its own target_subpart/record or relabel the merged record as b+c."},{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit first answers in multiple sections: source item 1 in part (a), item 27 in part (b), and item 30 in part (c).","suggested_fix":"Add the missing units and align b/c unit ids with the source subpart labels."}],"notes":"Canonical answers are source-faithful, but subpart extraction and scoring units are incomplete."}
+{"source_problem_id":"2023-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2023-individual-p2","iol-2023-individual-p2-sub-a","iol-2023-individual-p2-sub-b","iol-2023-individual-p2-sub-c","iol-2023-individual-p2-partial-through-a-target-b","iol-2023-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit source item 17 and item 21, and the part (c) unit values contain only the English prompts after the dash, not the required Apurina translations.","suggested_fix":"Rebuild answer_units for all items 17-26, preserving both acceptable Apurina alternatives for part (c)."},{"severity":"medium","record_id":"iol-2023-individual-p2-sub-b","field":"canonical","description":"The part (b) canonical/answer unit is contaminated by page-layout spillover from the following part (c), including an unrelated answer fragment.","suggested_fix":"Clean the part (b) canonical to contain only the grammatical correction, both valid equivalents, and its English translation."}],"notes":"Prompt text is answerable and matches the cited Apurina problem; answer extraction from the braced solution layout is defective."}
+{"source_problem_id":"2024-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2024-individual-p2","iol-2024-individual-p2-sub-a","iol-2024-individual-p2-sub-b","iol-2024-individual-p2-partial-through-a-target-b"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"Source part (c) is merged into records labeled only target b; target-b prompts say solve (b) only but include the Hadza-to-English and English-to-Hadza tasks together.","suggested_fix":"Split source part (c) into its own target_subpart/record or relabel merged records as b+c."},{"severity":"high","record_id":null,"field":"answer_units","description":"Answer units omit the first item in part (a) (19), part (b) (23), and merged part (c) (29).","suggested_fix":"Add missing units for 19, 23, and 29 and align unit ids with the correct source subparts."}],"notes":"The source includes animal images; the dataset captures their labels, but subpart grouping and scoring units need repair."}
+{"source_problem_id":"2025-individual-2","verdict":"fix_needed","record_ids_checked":["iol-2025-individual-p2","iol-2025-individual-p2-sub-a","iol-2025-individual-p2-sub-b","iol-2025-individual-p2-sub-c","iol-2025-individual-p2-partial-through-a-target-b","iol-2025-individual-p2-partial-through-ab-target-c"],"issues":[{"severity":"high","record_id":null,"field":"subparts","description":"Source part (d) is merged into records labeled only target c; the target-c prompt says solve (c) only but includes both c and d requests, and no separate d record exists.","suggested_fix":"Split part (d) into its own target_subpart/record or relabel the merged target as c+d."},{"severity":"medium","record_id":null,"field":"answer_units","description":"Part (a) is a single unit for 20 correspondences, which is too coarse; merged c/d units also omit source items 21 and 26.","suggested_fix":"Create one unit per correspondence for part (a), and add separate units for all c and d translations including 21 and 26."}],"notes":"Prompt and canonical content are otherwise faithful to the Gaahmg source problem."}

benchmark/IOL/ioling_hf/reports/manual_verification/batch_01.jsonl

@@ -0,0 +1,22 @@
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_needs_manual_review": 4}, "languages": ["Transcendental Algebra"], "problem_extraction_confidence": "ocr_heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 5, "answer_unit_ids": ["assignment_1.1", "assignment_1.2", "assignment_1.3", "assignment_2.1", "assignment_2.2"], "canonical_answer_chars": 2050, "canonical_excerpt": "Solution of Problem 1\n1. Nouns:\ne A ‘man’, A ‘woman’, i ‘boy’, A ‘girl’, etter’, [-- ‘work’.\n\n— Combinations: AA ‘man + woman = husband + wife’, iA ‘boy + girl = brother\n+ sister’, AAiA ‘man + woman + boy + girl = family’.\n\n— Family members are singled out by division and cancellation: 4¢% ‘family /(woman\n+ kids) = father’, is ‘kids/girl = brother’, AAA ‘family /kids = parents’.\n\n— Missing (deceased) family members are preceded by a minus sign: “ ‘kids\n(—parents)/(—parents) = orphans’ (apparently orphaned children of one and the\nsame family).\n\ne I ‘person’, (> 1) ‘giant’.\n\n2. Pronouns are composed of the character { or A (for feminine gender) and the subscripts 1\nto 3, which indicate the per", "item_type": "full_problem", "prompt_chars": 1315, "record_id": "iol-2003-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_1.2", "assignment_1.3"], "canonical_answer_chars": 272, "canonical_excerpt": "Assignment 1. 9. He loves with an unrequited love (i. e. loves without being loved).\n10. The taciturn (or mute) daughter will write about the father and the mother.\n11. You (sg. fem.) worked quickly (or hastily) and silently.\n12. The letter was eaten by the hungry sister.", "item_type": "subproblem", "prompt_chars": 1152, "record_id": "iol-2003-individual-p1-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["assignment_2.1", "assignment_2.2"], "canonical_answer_chars": 157, "canonical_excerpt": "Assignment 2. 13. (A; + AAs —t=—ni3\n14. (njjF-2\ni. (AGP ESI =(<D-1-\n16. (nix)¥S+t\n\n1st IOL: Borovetz ’03. Solutions to the Problems of the Individual Contest", "item_type": "subproblem", "prompt_chars": 1584, "record_id": "iol-2003-individual-p1-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["assignment_2.1", "assignment_2.2"], "canonical_answer_chars": 157, "canonical_excerpt": "Assignment 2. 13. (A; + AAs —t=—ni3\n14. (njjF-2\ni. (AGP ESI =(<D-1-\n16. (nix)¥S+t\n\n1st IOL: Borovetz ’03. Solutions to the Problems of the Individual Contest", "item_type": "partial_variant", "prompt_chars": 1920, "record_id": "iol-2003-individual-p1-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-sol.en.pdf", "source_problem_id": "2003-individual-1", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Transcendental Algebra", "year": 2003}
+{"author": "Maria Rubinstein", "curation_statuses": {"auto_extracted_needs_manual_review": 6}, "languages": [], "problem_extraction_confidence": "ocr_heading", "problem_number": 2, "problem_page_end": 2, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-team-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 992, "canonical_excerpt": "Solution of Problem 2\nAssignment 1. The subscripts mark the participants in the situation (the persons mentioned\nin the sentence). Identical letters mean identical individuals, different letters mean different indi-\nviduals. In this way it is shown which pronoun can refer to which noun. If a pronoun can refer\nto more than one noun, all possible subscripts are given, separated by slashes. If a pronoun can\nrefer to an individual not mentioned in the sentence, a letter is used that doesn’t mark any other\nword in the same sentence (e. g., he in (2) may be someone other than John or Peter, let’s say\nBill, if he exists at all). An asterisk next to a letter indicates that the pronoun can’t refer to", "item_type": "full_problem", "prompt_chars": 974, "record_id": "iol-2003-team-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 45, "canonical_excerpt": "(a) She; doesn’t like this trait in herself;.", "item_type": "subproblem", "prompt_chars": 945, "record_id": "iol-2003-team-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 57, "canonical_excerpt": "(b) The father; took his; /«;/, son; to his; /;/,/, room.", "item_type": "subproblem", "prompt_chars": 939, "record_id": "iol-2003-team-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 143, "canonical_excerpt": "(c) John; knows that Peter; has given his;/;,. book to his; /;/~/1/m Sok.\n\n1st IOL: Borovetz ’03. Solutions to the Problems of the Team Contest", "item_type": "subproblem", "prompt_chars": 1009, "record_id": "iol-2003-team-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 57, "canonical_excerpt": "(b) The father; took his; /«;/, son; to his; /;/,/, room.", "item_type": "partial_variant", "prompt_chars": 1037, "record_id": "iol-2003-team-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 143, "canonical_excerpt": "(c) John; knows that Peter; has given his;/;,. book to his; /;/~/1/m Sok.\n\n1st IOL: Borovetz ’03. Solutions to the Problems of the Team Contest", "item_type": "partial_variant", "prompt_chars": 1192, "record_id": "iol-2003-team-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-team-sol.en.pdf", "source_problem_id": "2003-team-2", "target_subparts": ["a", "b", "c", "full"], "title": "Subscripts", "year": 2003}
+{"author": "Boris Iomdin", "curation_statuses": {"auto_extracted_needs_manual_review": 1}, "languages": ["Chuvash (Turkic)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 4, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["full.1", "full.2"], "canonical_answer_chars": 2890, "canonical_excerpt": "Problem No.5\n\nIt is easy to note that the left column lists only intransitive verbs and the right column only\ntransitive ones. If a verb in the left column means ‘X’, its counterpart in the right column\nmeans ‘to cause X’ (i.e. it expresses a so-called causative meaning). Apparently, the Chuvash\nverbs in each pair have the same root, but one cannot determine which of the two verbs is\nprimary and which is derived from the primary verb using a suffix: in some pairs, the\nintransitive verb is shorter, and in others the other way round. If one admitted truncations, one\nwould have to formulate rules with exceptions. Consider e.g. the verbs ¢iihen and véren, which\nhave the same structure, but diffe", "item_type": "full_problem", "prompt_chars": 2919, "record_id": "iol-2004-individual-p5", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 5, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-sol.en.pdf", "source_problem_id": "2004-individual-5", "target_subparts": ["full"], "title": "Chuvash", "year": 2004}
+{"author": "Pyotr Arkadiev", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Lakhota (Siouxan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 9, "answer_unit_ids": ["assignment_1.1", "assignment_1.2", "assignment_1.3", "assignment_1.4", "assignment_2.1", "assignment_3.1", "assignment_3.2", "assignment_3.3", "assignment_3.4"], "canonical_answer_chars": 1915, "canonical_excerpt": "Problem #1\n      The word order is SOV.\n      Nouns are followed by ki {this is actually an article}.\n        Interrogative sentences end by he.\n      The subject and object markers on the verb are as follows:\n                               subject                        object\n1sg                     wa-                    ma-\n2sg                          ya-                              ni-\n3sg                        0                          0\n3pl                             -pi                        wičha-\n         If there is an object prefix as well as a subject prefix, they come in this order.\n      The number of the nouns is not expressed by the nouns themselves, only by the verbs.", "item_type": "full_problem", "prompt_chars": 2123, "record_id": "iol-2006-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["assignment_1.1", "assignment_1.2", "assignment_1.3", "assignment_1.4"], "canonical_answer_chars": 341, "canonical_excerpt": "Assignment 1.\n         1.  I came and the Indians called the bears.\n         2.  The man killed you and called me.\n         3.  The man killed you and he called me. (So it is rendered in English at any rate; we can’t\n          be sure whether ‘he’ can be ‘the man’ from the first clause or has to be someone else.)\n         4. He killed you.", "item_type": "subproblem", "prompt_chars": 1658, "record_id": "iol-2006-individual-p1-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 145, "canonical_excerpt": "Assignment 2.\n         1. Whom did he kill? 2. Who killed him?\n      There is not enough evidence to support the existence of other translations.", "item_type": "subproblem", "prompt_chars": 1446, "record_id": "iol-2006-individual-p1-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["assignment_3.1", "assignment_3.2", "assignment_3.3", "assignment_3.4"], "canonical_answer_chars": 384, "canonical_excerpt": "Assignment 3.\n        1.  lakhota ki hokšita ki ktepi čha matho ki hi\n        2.  yahi ną lakhota ki yakte\n        3. tuwa wačho he\n        4. wičhaša ki hipi čha tuwa wičhakte\n\n                                           Problem: Pyotr Arkadiev Solution: Alexander Piperski\n\nFourth International Linguistic Olympiad. Solutions of the Problems of the Individual Contest               2", "item_type": "subproblem", "prompt_chars": 2542, "record_id": "iol-2006-individual-p1-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 145, "canonical_excerpt": "Assignment 2.\n         1. Whom did he kill? 2. Who killed him?\n      There is not enough evidence to support the existence of other translations.", "item_type": "partial_variant", "prompt_chars": 1851, "record_id": "iol-2006-individual-p1-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 4, "answer_unit_ids": ["assignment_3.1", "assignment_3.2", "assignment_3.3", "assignment_3.4"], "canonical_answer_chars": 384, "canonical_excerpt": "Assignment 3.\n        1.  lakhota ki hokšita ki ktepi čha matho ki hi\n        2.  yahi ną lakhota ki yakte\n        3. tuwa wačho he\n        4. wičhaša ki hipi čha tuwa wičhakte\n\n                                           Problem: Pyotr Arkadiev Solution: Alexander Piperski\n\nFourth International Linguistic Olympiad. Solutions of the Problems of the Individual Contest               2", "item_type": "partial_variant", "prompt_chars": 3131, "record_id": "iol-2006-individual-p1-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-sol.en.pdf", "source_problem_id": "2006-individual-1", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Lakhota", "year": 2006}
+{"author": "Dmitry Gerasimov", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Movima (Isolate)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 2181, "canonical_excerpt": "Problem #2\n\nThe negative forms are composed of a particle kas followed by a modification of the original\nform which contains the marker -(k)a’- in one form or another. The rules for insertion of this\nmarker are as follows:\n\n    (1) The marker is inserted after the first syllable of the word if this syllable is either closed\n          (i.e., ends in a consonant) or long (i.e., contains a long vowel); otherwise the marker is\n        inserted after the second syllable of the word.\n\n    (2) If the marker is inserted after a long vowel, this vowel loses its length.\n\n    (3) If the marker is inserted after an open syllable, it retains its original form -ka’-; if it is\n        inserted after a clos", "item_type": "full_problem", "prompt_chars": 2490, "record_id": "iol-2007-individual-p2", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-sol.en.pdf", "source_problem_id": "2007-individual-2", "target_subparts": ["full"], "title": "Movima", "year": 2007}
+{"author": "Alexander Piperski", "curation_statuses": {"auto_extracted_text_high_confidence": 2}, "languages": ["Old Norse (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-prob.en.pdf", "record_count": 2, "record_types": {"full_problem": 1, "subproblem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 1429, "canonical_excerpt": "Problem #2.  (a) Rules:\n\n   1. Number of syllables. Each line contains 6 syllables.\n\n   2. Alliteration. See the statement of the problem.\n\n   3. Internal rhyme. Let us denote the vowels (and diphthongs) in each line by V1, V2, . . . , V6.\n    At least one consonant immediately following V5 must immediately follow Vn (n = 1, 2\n      or 3). Also, in even lines Vn = V5.\n\n                             VFor instance, cf. lines IV, 1–6 (alliteration is marked\nin boldface, internal rhyme by underlining):            1  a  ríks  (þreifsk reiddra øxa\n    IV                                              2   b rymr    ; kno˛´ttu spjo˛´r c glymja   )\n     1  háði gramr, þars gnúðu,                      3 ", "item_type": "full_problem", "prompt_chars": 4008, "record_id": "iol-2008-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 240, "canonical_excerpt": "(b) Leftover words: hœgra, smíði.                    8    l  sigr  (flugbeiddra m  vigra   ).\n\nSixth International Olympiad in Linguistics (2008).                                          2\nSolutions of the Problems of the Individual Contest", "item_type": "subproblem", "prompt_chars": 5564, "record_id": "iol-2008-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-sol.en.pdf", "source_problem_id": "2008-individual-2", "target_subparts": ["b", "full"], "title": "Old Norse", "year": 2008}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Bamana (Mande) Maninka (Mande)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 1077, "canonical_excerpt": "Problem #2.  The N’Ko script is written and read from right to left. The script is an alphabet:\neach letter stands for a consonant or a vowel. The letters within a word are joined.\n\n (a) A tilde above a vowel letter means low tone, its absence means high tone. But a vowel has\n     middle tone if it is marked in the same way as the one before it (if both either have or lack\n       tildes).\n\n        — b`ıl´ak`or´o                                              k`orik´ori —\n        — t´ajula                                            b´aw`o  —\n\n (b) If two adjacent syllables have the same vowel and both letters should have a tilde or neither\n     should have one according to the rules, only the s", "item_type": "full_problem", "prompt_chars": 2298, "record_id": "iol-2009-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 361, "canonical_excerpt": "(a) A tilde above a vowel letter means low tone, its absence means high tone. But a vowel has\n     middle tone if it is marked in the same way as the one before it (if both either have or lack\n       tildes).\n\n        — b`ıl´ak`or´o                                              k`orik´ori —\n        — t´ajula                                            b´aw`o  —", "item_type": "subproblem", "prompt_chars": 1424, "record_id": "iol-2009-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 530, "canonical_excerpt": "(b) If two adjacent syllables have the same vowel and both letters should have a tilde or neither\n     should have one according to the rules, only the second vowel is written.\n\n       — k`OlOl´O                                                  l´EtErE  —\n       — t´amEnE                                                b`ılak´oro —\n       — w´olowolo                                       j`amanak´E —\n\nSeventh International Olympiad in Linguistics (2009).                                        2\n   Individual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3252, "record_id": "iol-2009-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 530, "canonical_excerpt": "(b) If two adjacent syllables have the same vowel and both letters should have a tilde or neither\n     should have one according to the rules, only the second vowel is written.\n\n       — k`OlOl´O                                                  l´EtErE  —\n       — t´amEnE                                                b`ılak´oro —\n       — w´olowolo                                       j`amanak´E —\n\nSeventh International Olympiad in Linguistics (2009).                                        2\n   Individual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3666, "record_id": "iol-2009-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-sol.en-gb.pdf", "source_problem_id": "2009-individual-2", "target_subparts": ["a", "b", "full"], "title": "Maninka and Bamana", "year": 2009}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Drehu (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 2, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 937, "canonical_excerpt": "Problem #2.\n   • 1–4: caa 1, lue 2, köni 3, eke 4;\n   • 5, 10, 15: β-pi = 5β (1 ≤β ≤3);\n   • 6–9, 11–14, 16–19: α-ngömen = 5 + α, α-ko = 10 + α,               -e-ko > -ako\n                    α-qaihano = 15 + α  (1 ≤α ≤4);\n   • 20, 40, 60, 80: γ-atr = 20γ (1 ≤γ);           caa-atr > caatr, eke-atr > ekaatr\n   • 21–39, 41–59, . . . : Γ nge ∆= Γ + ∆(Γ = 20γ, 1 ≤∆≤19).\n (a) caatr nge caako: 31, caatr nge caangömen: 26, caatr nge caaqaihano: 36, ekaatr nge\n     ekengömen: 89, köniatr nge köniko: 73, köniatr nge könipi: 75, köniatr nge köniqai-\n     hano: 78, lueatr nge lue: 42, lueatr nge luako: 52, lueatr nge luepi: 50.\n\n (b) köniatr nge eke: 64 + caatr nge luepi: 30 = ekaatr nge ekako: 94\n    ", "item_type": "full_problem", "prompt_chars": 1175, "record_id": "iol-2010-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 254, "canonical_excerpt": "(a) caatr nge caako: 31, caatr nge caangömen: 26, caatr nge caaqaihano: 36, ekaatr nge\n     ekengömen: 89, köniatr nge köniko: 73, köniatr nge könipi: 75, köniatr nge köniqai-\n     hano: 78, lueatr nge lue: 42, lueatr nge luako: 52, lueatr nge luepi: 50.", "item_type": "subproblem", "prompt_chars": 651, "record_id": "iol-2010-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 115, "canonical_excerpt": "(b) köniatr nge eke: 64 + caatr nge luepi: 30 = ekaatr nge ekako: 94\n     luengömen: 7 + luako: 12 = ekeqaihano: 19", "item_type": "subproblem", "prompt_chars": 859, "record_id": "iol-2010-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 192, "canonical_excerpt": "(c) 21: caatr nge caa, 48: lueatr nge köningömen, 83: ekaatr nge köni.\n\nEighth International Olympiad in Linguistics (2010).                                      2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1481, "record_id": "iol-2010-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 115, "canonical_excerpt": "(b) köniatr nge eke: 64 + caatr nge luepi: 30 = ekaatr nge ekako: 94\n     luengömen: 7 + luako: 12 = ekeqaihano: 19", "item_type": "partial_variant", "prompt_chars": 1166, "record_id": "iol-2010-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 192, "canonical_excerpt": "(c) 21: caatr nge caa, 48: lueatr nge köningömen, 83: ekaatr nge köni.\n\nEighth International Olympiad in Linguistics (2010).                                      2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 1931, "record_id": "iol-2010-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-sol.en.pdf", "source_problem_id": "2010-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Drehu Numerals", "year": 2010}
+{"author": "Alexander Piperski", "curation_statuses": {"auto_extracted_text_high_confidence": 2}, "languages": ["Faroese (North Germanic)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-prob.en-gb.pdf", "record_count": 2, "record_types": {"full_problem": 1, "subproblem": 1}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 819, "canonical_excerpt": "Problem #2.\n (a)  bøga      [bø:va]        (b) In the first syllable a [Ea], á [Oa], e [e:], ei  [ai], ey  [Ei], i  [i:],\n     knoðar   [kno:ar]      o [o:], oy [Oi], ó [Ou], u [u:], ú [0u], ø [ø:].\n     kvøða    [kvø:a]                            Between vowels ð = g. The first applicable rule is applied:\n      løgur     [lø:vur]\n      plága     [plOava]             1. ð/g [w] | [u(:)]     ;\n      skaði     [skEaji]\n                                         2. ð/g  [j] | [i(:)]    or     [i(:)];     toygur    [tOijur]\n      trúgi     [tr0uwi]                                         3. ð/g [v] |     [u(:)];\n                                         4. ð/g [v] in a noun, [∅] in a verb.\n\nNinth Int", "item_type": "full_problem", "prompt_chars": 2475, "record_id": "iol-2011-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 806, "canonical_excerpt": "(a)  bøga      [bø:va]        (b) In the first syllable a [Ea], á [Oa], e [e:], ei  [ai], ey  [Ei], i  [i:],\n     knoðar   [kno:ar]      o [o:], oy [Oi], ó [Ou], u [u:], ú [0u], ø [ø:].\n     kvøða    [kvø:a]                            Between vowels ð = g. The first applicable rule is applied:\n      løgur     [lø:vur]\n      plága     [plOava]             1. ð/g [w] | [u(:)]     ;\n      skaði     [skEaji]\n                                         2. ð/g  [j] | [i(:)]    or     [i(:)];     toygur    [tOijur]\n      trúgi     [tr0uwi]                                         3. ð/g [v] |     [u(:)];\n                                         4. ð/g [v] in a noun, [∅] in a verb.\n\nNinth International Ol", "item_type": "subproblem", "prompt_chars": 2033, "record_id": "iol-2011-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-sol.en-gb.pdf", "source_problem_id": "2011-individual-2", "target_subparts": ["a", "full"], "title": "Faroese", "year": 2011}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Umbu-Ungu (Trans-New Guinea)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 1119, "canonical_excerpt": "Problem #2.\n       Umbu-Ungu                                       Umbu-Ungu\n     1  telu                           24                       tokapu\n     2  talu                           48 = 24 × 2               tokapu talu\n     3  yepoko                        72 = 24 × 3               tokapu yepoko\n    12  rurepo                   α ¬ β := (α −4) + β,       α-nga β\n    16  malapu                    α ∈{12, 16, 20, 24, 28, 32},\n    20  supu                         β ∈{1, 2, 3}    24  tokapu                       γ + δ,                     γ δ\n    28  alapu                          γ = 24k, k ∈{1, 2, 3},\n    32  polangipu                       9 ≤δ ≤32, δ̸ = 24\n (a)  tokapu polangipu = 24 +", "item_type": "full_problem", "prompt_chars": 1253, "record_id": "iol-2012-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 216, "canonical_excerpt": "(a)  tokapu polangipu = 24 + 32 = 56,\n      tokapu talu rureponga telu = 24 × 2 + 12 ¬ 3 = 57,\n      tokapu yepoko malapunga talu = 24 × 3 + 16 ¬ 2 = 86,\n      tokapu yepoko polangipunga telu = 24 × 3 + 32 ¬ 1 = 101.", "item_type": "subproblem", "prompt_chars": 1276, "record_id": "iol-2012-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 229, "canonical_excerpt": "(b)  13 = 16 ¬ 1 = malapunga telu,\n      66 = 24 × 2 + 20 ¬ 2 = tokapu talu supunga talu,\n      72 = 24 × 3 = tokapu yepoko,\n      76 = 24 × 2 + 28 = tokapu talu alapu,\n      95 = 24 × 3 + 24 ¬ 3 = tokapu yepoko tokapunga yepoko.", "item_type": "subproblem", "prompt_chars": 1310, "record_id": "iol-2012-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 229, "canonical_excerpt": "(b)  13 = 16 ¬ 1 = malapunga telu,\n      66 = 24 × 2 + 20 ¬ 2 = tokapu talu supunga talu,\n      72 = 24 × 3 = tokapu yepoko,\n      76 = 24 × 2 + 28 = tokapu talu alapu,\n      95 = 24 × 3 + 24 ¬ 3 = tokapu yepoko tokapunga yepoko.", "item_type": "partial_variant", "prompt_chars": 1579, "record_id": "iol-2012-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-sol.en.pdf", "source_problem_id": "2012-individual-2", "target_subparts": ["a", "b", "full"], "title": "Umbu-Ungu", "year": 2012}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Tundra Yukaghir (Yukaghir)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 1445, "canonical_excerpt": "Problem #2.  The compound noun has the following structure:                  {                 }\n                               -d-  (before a vowel)\n               modifier  +                  +  modified  .\n                              -n-  (before a consonant)\n                    (-l is lost)\n\n      ilennime      herd of deer (“house of deer”)     møŋer         thunder\n     joqonnime    wooden house (“Yakut house”)     ciremennime   nest\n      saancohoje    wooden knife                    joqoncohoje    Yakut knife\n     johudawur     nose case                      saadoŋoj      wooden box\n (a)  ilenlegul       deer feed                     uoduo          grandchild\n      cireme         bir", "item_type": "full_problem", "prompt_chars": 1695, "record_id": "iol-2013-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 392, "canonical_excerpt": "(a)  ilenlegul       deer feed                     uoduo          grandchild\n      cireme         bird                          oŋoj           bag\n      johul           nose                             aariinjohul      rifle’s muzzle\n     aariinmøŋer  gunshot (“rifle thunder”)         uodawur        cradle\n       joqodile        horse (“Yakut deer”)                joqol          Yakut person", "item_type": "subproblem", "prompt_chars": 808, "record_id": "iol-2013-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 200, "canonical_excerpt": "(b) aarii — rifle, aariidoŋoj — rifle case, ciremedawur — nest (= ciremennime), ile\n   — deer, johudewce — tip of nose, legul — food, saal — wood, saannime — wooden\n     house (= joqonnime), uo — child.", "item_type": "subproblem", "prompt_chars": 1166, "record_id": "iol-2013-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 248, "canonical_excerpt": "(c) iron bird — cuoncireme, snoring — johunmøŋer, tip of knife’s blade — cohoje-\n     dewce, sack for provisions — legudoŋoj .\n\nEleventh International Olympiad in Linguistics (2013).                                    2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2214, "record_id": "iol-2013-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 200, "canonical_excerpt": "(b) aarii — rifle, aariidoŋoj — rifle case, ciremedawur — nest (= ciremennime), ile\n   — deer, johudewce — tip of nose, legul — food, saal — wood, saannime — wooden\n     house (= joqonnime), uo — child.", "item_type": "partial_variant", "prompt_chars": 1611, "record_id": "iol-2013-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 248, "canonical_excerpt": "(c) iron bird — cuoncireme, snoring — johunmøŋer, tip of knife’s blade — cohoje-\n     dewce, sack for provisions — legudoŋoj .\n\nEleventh International Olympiad in Linguistics (2013).                                    2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2887, "record_id": "iol-2013-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-sol.en-gb.pdf", "source_problem_id": "2013-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Tundra Yukaghir", "year": 2013}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Tangut (Tibeto-Burman)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 984, "canonical_excerpt": "Problem #3.  Rules:\n #     #\n\n      ny1           wiej2        wia1    ma1            la2               'i@1\n     @@       @@         HHHH       @@       @@\n  ¹we.j1 ¹we.j1     lio2  ndo.n1    lio2   me  ndo.n1    lio2  ndo.n1   ¹we.j1 ¹we.j1\n\n #     #\n\n      ny1           wiej2        wia1    ma1            la2               'i@1\n     @@       @@         HHHH       @@       @@\n  ¹we.j1 ¹we.j1   mu1   kâ.j1   mu1   me    kâ.j1   mu1   kâ.j1    ¹we.j1 ¹we.j1\n\n   △      ⃝    (    = man,     = woman)\n            (( (((( ((( (((( hhhhhhhhhhhhhHHHHH\n          Lhie2nyn2        Ngwi1mbyn2       Mbe2phon1        Ngon2ngwe1\n        @          @          @          @\n         @          @          @  ", "item_type": "full_problem", "prompt_chars": 2968, "record_id": "iol-2014-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 20, "canonical_excerpt": "(a)\n\n              1", "item_type": "subproblem", "prompt_chars": 2281, "record_id": "iol-2014-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 163, "canonical_excerpt": "(b) Nie2tse.  Sie1tsie1 'yn1 ndo.n1 ngu2.\n\nTwelfth International Olympiad in Linguistics (2014).                                     3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3733, "record_id": "iol-2014-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 163, "canonical_excerpt": "(b) Nie2tse.  Sie1tsie1 'yn1 ndo.n1 ngu2.\n\nTwelfth International Olympiad in Linguistics (2014).                                     3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3806, "record_id": "iol-2014-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-sol.en-gb.pdf", "source_problem_id": "2014-individual-3", "target_subparts": ["a", "b", "full"], "title": "Tangut", "year": 2014}
+{"author": "Alexander Piperski", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": [], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "b.1", "c.1"], "canonical_answer_chars": 1012, "canonical_excerpt": "Problem 3.\n\n (a)    1. Leave the first letter in place.\n\n         2. Delete h and w.\n\n         3. Replace all consonant letters with digits (letters whose most common sounds are\n           similar are grouped together):\n                         bpv (f)   cgjkqs (xz)   dt     l  mn   r\n                       1       2          3   4  5    6\n\n         4. Reduce any sequence of two or more identical digits to a single digit.\n\n         5. Delete all vowels (a, e, i, o, u, y).\n\n         6. Leave only the first three digits or add zeroes on the right to make the code one\n            letter and three digits long.\n\n (b) Allaway: A400, Anderson: A536, Ashcombe: A251, Buckingham: B252,\n     Chapman: C15", "item_type": "full_problem", "prompt_chars": 1605, "record_id": "iol-2015-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5"], "canonical_answer_chars": 598, "canonical_excerpt": "(a)    1. Leave the first letter in place.\n\n         2. Delete h and w.\n\n         3. Replace all consonant letters with digits (letters whose most common sounds are\n           similar are grouped together):\n                         bpv (f)   cgjkqs (xz)   dt     l  mn   r\n                       1       2          3   4  5    6\n\n         4. Reduce any sequence of two or more identical digits to a single digit.\n\n         5. Delete all vowels (a, e, i, o, u, y).\n\n         6. Leave only the first three digits or add zeroes on the right to make the code one\n            letter and three digits long.", "item_type": "subproblem", "prompt_chars": 1272, "record_id": "iol-2015-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 271, "canonical_excerpt": "(b) Allaway: A400, Anderson: A536, Ashcombe: A251, Buckingham: B252,\n     Chapman: C155, Colquhoun: C425, Evans: E152, Fairwright: F623,\n      Kingscott: K523, Lewis: L200, Littlejohns: L342, Stanmore: S356,\n      Stubbs: S312, Tocher: T260, Tonks: T520, Whytehead: W330.", "item_type": "subproblem", "prompt_chars": 1358, "record_id": "iol-2015-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 124, "canonical_excerpt": "(c) Ferguson: F622, Fitzgerald: F326, Hamnett: H530, Keefe: K100,\n     Maxwell: M240, Razey: R200, Shaw: S000, Upfield: U143.", "item_type": "subproblem", "prompt_chars": 1810, "record_id": "iol-2015-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 271, "canonical_excerpt": "(b) Allaway: A400, Anderson: A536, Ashcombe: A251, Buckingham: B252,\n     Chapman: C155, Colquhoun: C425, Evans: E152, Fairwright: F623,\n      Kingscott: K523, Lewis: L200, Littlejohns: L342, Stanmore: S356,\n      Stubbs: S312, Tocher: T260, Tonks: T520, Whytehead: W330.", "item_type": "partial_variant", "prompt_chars": 2009, "record_id": "iol-2015-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 124, "canonical_excerpt": "(c) Ferguson: F622, Fitzgerald: F326, Hamnett: H530, Keefe: K100,\n     Maxwell: M240, Razey: R200, Shaw: S000, Upfield: U143.", "item_type": "partial_variant", "prompt_chars": 2760, "record_id": "iol-2015-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-sol.en-gb.pdf", "source_problem_id": "2015-individual-3", "target_subparts": ["a", "b", "c", "full"], "title": "Soundex", "year": 2015}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Kunuz Nubian (Eastern Sudanic)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 1972, "canonical_excerpt": "Problem 3.  Rules:\n\n   • Word order: S (O′) O V\n     (S = subject, O′ = indirect object, O = direct object, V = predicate).\n\n   • noun:\n\n      – \n\n      – number:\n\n           * singular: -∅\n             {\n                               -i      after a consonant\n           * plural:\n                           -ːcci  after a vowel\n      – case:\n\n           * S: -∅            \n             -ːg  after a vowel\n           * O′, O:    -gi  after a sonorant consonant\n                                     -ki  after an obstruent consonant\n\n   • verb:\n\n      – tense (present a-, future bi-)\n\n      – \n\n      – for somebody: ‘to give’ (to the first person: deːn, to the second or third person:", "item_type": "full_problem", "prompt_chars": 1841, "record_id": "iol-2016-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4"], "canonical_answer_chars": 380, "canonical_excerpt": "(a)    1. magasi argi ajomirra.\n        The thieves are striking us.\n         2. ay kanarriːg baːbki alletirsi.\n            I repaired the door for the neighbour.\n         3. hanu tirtiːg elirsu.\n        The donkey found the owners.\n         4. tirti argi kamgi deːccirsa.\n        The owners gave us the camel.\n         5. ay darbadki bijaːndi.\n            I will buy the chicken.", "item_type": "subproblem", "prompt_chars": 1435, "record_id": "iol-2016-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 434, "canonical_excerpt": "(b)    6. The neighbours are giving the necklace to the owner.\n          kanarriːcci tirtki beyyeːg atirra.\n         7. The young man bought the dog for us.\n          jahal argi walgi jaːndeːccirsu.\n         8. We will steal him.\n         ar mangi bimaːgru.\n         9. The owners struck the thieves.\n            tirti magasiːg jomirsa.\n       10. The dogs found the chickens for the coward.\n         wali sarkaːygi darbadiːg eltirsa.", "item_type": "subproblem", "prompt_chars": 2327, "record_id": "iol-2016-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 434, "canonical_excerpt": "(b)    6. The neighbours are giving the necklace to the owner.\n          kanarriːcci tirtki beyyeːg atirra.\n         7. The young man bought the dog for us.\n          jahal argi walgi jaːndeːccirsu.\n         8. We will steal him.\n         ar mangi bimaːgru.\n         9. The owners struck the thieves.\n            tirti magasiːg jomirsa.\n       10. The dogs found the chickens for the coward.\n         wali sarkaːygi darbadiːg eltirsa.", "item_type": "partial_variant", "prompt_chars": 2760, "record_id": "iol-2016-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-sol.en.pdf", "source_problem_id": "2016-individual-3", "target_subparts": ["a", "b", "full"], "title": "Kunuz Nubian", "year": 2016}
+{"author": "Bruno L'Astorina", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Kimbundu (Atlantic-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 6, "answer_unit_ids": ["a.1", "a.2", "a.3", "b.1", "b.2", "b.3"], "canonical_answer_chars": 1532, "canonical_excerpt": "Problem 3.  Rules:\n\n   • word order:\n      ? ? ?\n      – S  V  (O) (L)    (S = subject, V = predicate, O = object, L = place)\n        6\n          ŋg-   (S = 1st person sg)\n      ? ?        ?\n      – X  V  L ≡  L  V  X\n        6        6\n      ?\n      – N  P    (N = noun, P = possessive pronoun);\n        6\n   •   6: gender and number concord —\n\n       singular   plural\n        mú-       à-\n             dì-    má-  (before a vowel i- > j-, u- > w-, a- > ∅);\n                        ʒì-\n     mù+ (place)\n\n   • ?   : the tone of the last syllable assimilates to the tone of the first syllable of the\n      following word;\n\n   • the tone of the last syllable of the last word is low in a declarative se", "item_type": "full_problem", "prompt_chars": 1923, "record_id": "iol-2017-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["a.1", "a.2", "a.3"], "canonical_answer_chars": 303, "canonical_excerpt": "(a)   12.  múlóʒí mwámónà ʒìtéténbwá\n           Did the sorcerer see the stars?\n        13.  ʒìtéténbwá ʒjálà mùdjúlù\n           There are stars in the sky. = (4)\n        14.  ŋgákínà\n               I danced.\n        15.   djálá djámónà màhónʒò mùlwándá\n           Did the man see the bananas in Luanda?", "item_type": "subproblem", "prompt_chars": 1495, "record_id": "iol-2017-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 468, "canonical_excerpt": "(b)   16.  Did I sing?\n             ŋgáʃíká\n        17.  The person met the sorcerer and the friend in the square.\n          mútú mwásáŋgá múlóʒì nì dìkámbà mùkìtándà\n        18.  My man (husband) saw the cave.\n             djálá djámí djámónà dìkúŋgù\n        19.  There are sorcerers in Luanda.\n          mùlwándà mwálà àlóʒì; àlóʒí álà mùlwándà\n\nFifteenth International Linguistics Olympiad (2017).                                      4\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2431, "record_id": "iol-2017-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 468, "canonical_excerpt": "(b)   16.  Did I sing?\n             ŋgáʃíká\n        17.  The person met the sorcerer and the friend in the square.\n          mútú mwásáŋgá múlóʒì nì dìkámbà mùkìtándà\n        18.  My man (husband) saw the cave.\n             djálá djámí djámónà dìkúŋgù\n        19.  There are sorcerers in Luanda.\n          mùlwándà mwálà àlóʒì; àlóʒí álà mùlwándà\n\nFifteenth International Linguistics Olympiad (2017).                                      4\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2787, "record_id": "iol-2017-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-sol.en.pdf", "source_problem_id": "2017-individual-3", "target_subparts": ["a", "b", "full"], "title": "Kimbundu", "year": 2017}
+{"author": "André Nikulin", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Terêna (Arawakan)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 1488, "canonical_excerpt": "Problem 3.\n\n        • 1st person sg:\n\n           –  if the word contains at least one voiceless consonant, the first one becomes nasal+voiced\n                (⟨p →mb, t →nd, h →nz, x →nj, k →ng⟩)\n           – otherwise the whole word is nasalised (-m)\n\n        • 2nd person sg:\n\n           –  if the word begins with a vowel different from i (í, î): y-\n           – otherwise the first vowel different from i (í, î) changes as follows:\n           ∗a →e, á →é, â →eâ;\n           ∗o →e, ó →é, ô →eô;\n          ∗û →iû;\n           ∗e →i, é →í, ê →î (the same happens in the immediately following syllables).\n      (a)   1st person sg  2nd person sg                 1st person sg  2nd person sg\n            mb", "item_type": "full_problem", "prompt_chars": 3688, "record_id": "iol-2018-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 755, "canonical_excerpt": "(a)   1st person sg  2nd person sg                 1st person sg  2nd person sg\n            mbîho          1pîhe            to go          noínjoa         8neíxoa         to see it\n           2mbêyo         pîyo           animal        vanénjo        9venéxo         to buy\n            yênom          3yîno           wife         mómindi       10mémiti        to be tired\n           4ênom         yêno           mother        ínzikaxovoku    11íhikexovoku   school\n             ngásaxo        5késaxo         to feel cold   12ônju         yôxu            grandfather\n                njérere           6xíriri           side        13nêwom       nîwo          nephew\n             ndôko          7teôko", "item_type": "subproblem", "prompt_chars": 2495, "record_id": "iol-2018-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 120, "canonical_excerpt": "(b)    1. 2nd person sg: in Portuguese loanwords á →eá; in native words á →é, â →eâ.\n               2. lándana, ngápana.", "item_type": "subproblem", "prompt_chars": 4947, "record_id": "iol-2018-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 120, "canonical_excerpt": "(b)    1. 2nd person sg: in Portuguese loanwords á →eá; in native words á →é, â →eâ.\n               2. lándana, ngápana.", "item_type": "partial_variant", "prompt_chars": 5755, "record_id": "iol-2018-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-sol.en.pdf", "source_problem_id": "2018-individual-3", "target_subparts": ["a", "b", "full"], "title": "Terêna", "year": 2018}
+{"author": "André Nikulin", "curation_statuses": {"auto_extracted_text_high_confidence": 8}, "languages": ["Middle Persian (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-prob.en.pdf", "record_count": 8, "record_types": {"full_problem": 1, "partial_variant": 3, "subproblem": 4}, "records": [{"answer_unit_count": 4, "answer_unit_ids": ["a.1", "b.1", "c.1", "d.1"], "canonical_answer_chars": 2677, "canonical_excerpt": "Problem 3.  The direction of writing is from right to left.\n\n   • [in the Book Pahlavi script] ↔[scientific transliteration]:\n\n                              n, r, w, '                                               n               ʾ, A, h                              s      A                   z                        II         b        Z                        c, p          b                k         c            d, g, y, Ẕ    k                      š      i            s\n        E                        l                   t   h          r          t               m               x\n   • [[ligatures]]:\n      DX ←A+A     s ←A+i       Dl ←A+r          MZ ←x+Z\n      TeX ←t+A+A         A ←i+i    ", "item_type": "full_problem", "prompt_chars": 4768, "record_id": "iol-2019-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 306, "canonical_excerpt": "(a)\n\n  1    2    3    4    5    6    7    8    9    10   11   12   13   14   15   16   17\n  E   CC G  Y  D  CC K   S   V   L     I    Z   P  DD O  M  B\n  18   19   20   21   22   23   24   25   26   27   28   29   30   31   32   33   34\n  F   X  C   BB V  AA  J   T  Q  H  T  U  W  R  A   F   N\n\n         ?", "item_type": "subproblem", "prompt_chars": 4885, "record_id": "iol-2019-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 75, "canonical_excerpt": "(b) U.    =       ʾhlmn' Ahreman ‘the evil spirit’.  $    nHrDX           6", "item_type": "subproblem", "prompt_chars": 1859, "record_id": "iol-2019-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 425, "canonical_excerpt": "(c)  EE.           bʾnbšn'  bāmbišn  queen              nnsbnAb\n      FF.            chʾl      čahār     four             LDXc\n     GG.               tl          tar       over, through, across           rt\n     HH.            hʾk'      xāk       earth, dust                n%DX\n           II.          hwn'     xōn       blood                         nnnA\n        JJ.          hwk'     hūg       pig                     nknA", "item_type": "subproblem", "prompt_chars": 2489, "record_id": "iol-2019-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 295, "canonical_excerpt": "(d)\n\nKK. DKRA             LL. dlwnd         MM. stwl           NN. cmbl             ankI                            innLI                rnTF                rbGc\n\nSeventeenth International Linguistics Olympiad (2019)                                    4\nIndividual Contest Solutions\n\n          R", "item_type": "subproblem", "prompt_chars": 3563, "record_id": "iol-2019-individual-p3-sub-d", "target_subparts": ["d"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 75, "canonical_excerpt": "(b) U.    =       ʾhlmn' Ahreman ‘the evil spirit’.  $    nHrDX           6", "item_type": "partial_variant", "prompt_chars": 2218, "record_id": "iol-2019-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 425, "canonical_excerpt": "(c)  EE.           bʾnbšn'  bāmbišn  queen              nnsbnAb\n      FF.            chʾl      čahār     four             LDXc\n     GG.               tl          tar       over, through, across           rt\n     HH.            hʾk'      xāk       earth, dust                n%DX\n           II.          hwn'     xōn       blood                         nnnA\n        JJ.          hwk'     hūg       pig                     nknA", "item_type": "partial_variant", "prompt_chars": 2951, "record_id": "iol-2019-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 295, "canonical_excerpt": "(d)\n\nKK. DKRA             LL. dlwnd         MM. stwl           NN. cmbl             ankI                            innLI                rnTF                rbGc\n\nSeventeenth International Linguistics Olympiad (2019)                                    4\nIndividual Contest Solutions\n\n          R", "item_type": "partial_variant", "prompt_chars": 4478, "record_id": "iol-2019-individual-p3-partial-through-abc-target-d", "target_subparts": ["d"], "visible_solved_subparts": ["a", "b", "c"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-sol.en.pdf", "source_problem_id": "2019-individual-3", "target_subparts": ["a", "b", "c", "d", "full"], "title": "Book Pahlavi script", "year": 2019}
+{"author": "Samuel Ahmed", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Kilivila (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 2086, "canonical_excerpt": "Problem 3.\n\n   • Word order:\n\n      – V(O)S (V = verb; O = object; S = subject)\n\n      – QVX (Q = interrogative element)\n\n      – (which | Qt)N(D)(A) (Qt = quantity, N = noun, D = demonstrative pronouns, A\n      = adjective)\n\n   • verb:\n\n      – le- past\n      – bi- future\n      – -si subject = plural (-gisi-si > -gisesi)\n\n   • noun:\n                           singular    plural\n         man        tau       tauwau\n               child        gwadi     gugwadi\n               chief       guyau     gweguyau\n              old woman  namwaya  nunumwaya\n        ∼               singular = plural\n\n   • class (Cl):\n\n      – men: te- (before Qt); to-\n      – women/animals: na-\n      – children: gudi-", "item_type": "full_problem", "prompt_chars": 2198, "record_id": "iol-2021-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4"], "canonical_answer_chars": 488, "canonical_excerpt": "(a)  14. Navila vivila biyamata tomwaya mtona?\n       How many women will this old man look after?\n       15. Bikamkwamsi kweyu vivila minasina.\n         These women will eat two things.\n       16. Amagudina gwadi lekota?\n        Which child arrived?\n       17. Tevila tauwau bigisesi gugwadi gudigasisi?\n       How many men will see the wild children?;\n       How many men will the wild children see?\n       18. Legisesi ketala waga vivila minasiwena.\n         Those women saw one canoe.", "item_type": "subproblem", "prompt_chars": 1983, "record_id": "iol-2021-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 560, "canonical_excerpt": "(b)  19. How many canoes did those old women see?\n          Kevila waga legisesi nunumwaya minasiwena?\n       20. These four white men will look after this clever child.\n          Biyamatasi gwadi magudina gudikabitam tevasi dimdim mtosina.\n       21. How many children will eat these pigs?\n          Gudivila gugwadi bikamkwamsi bunukwa minasina?\n       22. Which woman caught those beautiful fish?\n        Aminana vivila lebani yena minasiwena namanabweta?\n       23. Two wild dogs saw that old man.\n           Legisesi tomwaya mtowena nayu ka’ukwa nagasisi.", "item_type": "subproblem", "prompt_chars": 2493, "record_id": "iol-2021-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 560, "canonical_excerpt": "(b)  19. How many canoes did those old women see?\n          Kevila waga legisesi nunumwaya minasiwena?\n       20. These four white men will look after this clever child.\n          Biyamatasi gwadi magudina gudikabitam tevasi dimdim mtosina.\n       21. How many children will eat these pigs?\n          Gudivila gugwadi bikamkwamsi bunukwa minasina?\n       22. Which woman caught those beautiful fish?\n        Aminana vivila lebani yena minasiwena namanabweta?\n       23. Two wild dogs saw that old man.\n           Legisesi tomwaya mtowena nayu ka’ukwa nagasisi.", "item_type": "partial_variant", "prompt_chars": 3034, "record_id": "iol-2021-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-sol.en.pdf", "source_problem_id": "2021-individual-3", "target_subparts": ["a", "b", "full"], "title": "Kilivila", "year": 2021}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Nǀuuki (Tuu)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 10, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4", "full.5", "full.6", "full.7", "full.8", "full.9", "full.10"], "canonical_answer_chars": 2511, "canonical_excerpt": "Problem 3.\n\n   • Word order:\n                      \n      –                   S  L0                                                    O′)                                               (ŋ\n                                                              (O)                                                    (O)                                                    O′)            F(S)                                               (ŋ                                                                            (N) (T) V\n                                                    O′)          F(O)                     S                        Lf                                      ", "item_type": "full_problem", "prompt_chars": 2586, "record_id": "iol-2022-individual-p3", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-sol.en.pdf", "source_problem_id": "2022-individual-3", "target_subparts": ["full"], "title": "Nǀuuki (Nǁng)", "year": 2022}
+{"author": "David Hultman", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Coastal Marind (Anim)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 13, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4", "full.5", "full.6", "full.7", "full.8", "full.9", "full.10", "full.11", "full.12", "full.13"], "canonical_answer_chars": 1355, "canonical_excerpt": "Problem 3.\n                                                                       - root = …C↑C+\n                                                             object:\n    1. subject:\n                                                       • ana 1st person        • nak 1st person\n        • o 2nd person sg                               • aɣa 2nd person sg\n        • e 2nd person pl                                                       • a 2nd person pl\n    2.   • b ‘ really’                                                       • e 3rd person sg\n        • bat ‘Aw, ——.’\n        • bam ‘Oh, ——!’\n\n    3. a (subject = 3rd person sg)\n\n    4. um ‘ in vain’\n\n    5. e {subject, object} = 1st person pl\n\nkb", "item_type": "full_problem", "prompt_chars": 2058, "record_id": "iol-2023-individual-p3", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-sol.en.pdf", "source_problem_id": "2023-individual-3", "target_subparts": ["full"], "title": "Coastal Marind", "year": 2023}
+{"author": "Aida Davletova", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Komnzo (Yam)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "c.1"], "canonical_answer_chars": 1680, "canonical_excerpt": "Problem 3.\n          {\n                   rä  —   if A is a woman\n     • A B-ane X\n                 yé —   if A is a man\n\n       – A = B’s relative\n       – X = kinship term\n\n     •  ŋafe        father                                          B   brother\n     ŋame      mother                                  D   daughter\n      nge         child                                                    F    father\n       zath        grandfather, grandson                         H   husband\n       zath ŋare  grandmother, granddaughter                  M  mother\n      nane       elder sibling                                             S   son\n      ngth      younger sibling                      W  wi", "item_type": "full_problem", "prompt_chars": 2534, "record_id": "iol-2024-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 360, "canonical_excerpt": "(a)\n\n                             Trafe              Kurai\n\n                   Wafine                Naimr                   Nfiyam       Marua                Mea                          Skri\n\n                        Toko               Wims                  Nakre     Mabata            Tawth                   Abia               Gwam\n\n                  Maraga", "item_type": "subproblem", "prompt_chars": 1882, "record_id": "iol-2024-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 124, "canonical_excerpt": "(b)    1. bäiŋam rä\n          2. enat yé\n          3. zath ŋare rä\n          4. nge yé\n          5. Nakreane\n          6. rä", "item_type": "subproblem", "prompt_chars": 1880, "record_id": "iol-2024-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 157, "canonical_excerpt": "(c) Skri Abiaane bäiŋaf yé.\n\nTwenty-first International Linguistics Olympiad (2024)                                            4\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3200, "record_id": "iol-2024-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 124, "canonical_excerpt": "(b)    1. bäiŋam rä\n          2. enat yé\n          3. zath ŋare rä\n          4. nge yé\n          5. Nakreane\n          6. rä", "item_type": "partial_variant", "prompt_chars": 2293, "record_id": "iol-2024-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 157, "canonical_excerpt": "(c) Skri Abiaane bäiŋaf yé.\n\nTwenty-first International Linguistics Olympiad (2024)                                            4\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3765, "record_id": "iol-2024-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-sol.en.pdf", "source_problem_id": "2024-individual-3", "target_subparts": ["a", "b", "c", "full"], "title": "Komnzo", "year": 2024}
+{"author": "Eimear McKnight", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Kuria (Atlantic-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7"], "canonical_answer_chars": 1536, "canonical_excerpt": "Problem 3.  Rules:\n\n     • Word order: verb (object).\n\n     • Verb:\n       – ‘indeed’: n-\n       – subject: -n- 1st person sg, -to- 1st person pl, -a- 3rd person sg, -ꞵa- 3rd person pl\n       – tense: -ra- ‘about to’, -V- past\n       – stem\n     -n-V- > -naa-, else -V- = preceding vowel; nꞵ> mb; nr > nd\n\n     • Tone contours:\n       – ‘about to’: ◌̀◌̀◌̀◌́\n       – past: ◌̀◌̀◌́\n       – past + ‘indeed’, present: ◌́\n\n     • Tone assignment:\n          1. All vowels are initially ◌̀.\n          2. Assign tones to vowels one-to-one starting from first vowel of stem, continuing on to object\n                if it exists.\n          3.  If final vowel reached and more than one tone remains:\n          ", "item_type": "full_problem", "prompt_chars": 2877, "record_id": "iol-2025-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 18, "canonical_excerpt": "(a)   20. ahéétóka", "item_type": "subproblem", "prompt_chars": 1723, "record_id": "iol-2025-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 7, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7"], "canonical_answer_chars": 635, "canonical_excerpt": "(b)   21. ꞵaasukurá — They have rubbed (something).\n        22. toosya iʧíímbéɣo — We have ground the seeds.\n        23. ndóma — I bite (something).\n        24. naaꞵína — Indeed, he has sung (something).\n   (c)   25. We are about to eat the seeds. — torarya iʧiímbéγo\n        26.  I sing (something). — mbína\n        27. Indeed, we have measured the secretary bird. — ntooꞵíímá írítáárákímúra\n        28. We are about to burn (something). — torasaambǎ\n        29. He has remembered (something). — aaheetóka\n\nTwenty-second International Linguistics Olympiad (2025)                                          4\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 4107, "record_id": "iol-2025-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 7, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7"], "canonical_answer_chars": 635, "canonical_excerpt": "(b)   21. ꞵaasukurá — They have rubbed (something).\n        22. toosya iʧíímbéɣo — We have ground the seeds.\n        23. ndóma — I bite (something).\n        24. naaꞵína — Indeed, he has sung (something).\n   (c)   25. We are about to eat the seeds. — torarya iʧiímbéγo\n        26.  I sing (something). — mbína\n        27. Indeed, we have measured the secretary bird. — ntooꞵíímá írítáárákímúra\n        28. We are about to burn (something). — torasaambǎ\n        29. He has remembered (something). — aaheetóka\n\nTwenty-second International Linguistics Olympiad (2025)                                          4\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 4178, "record_id": "iol-2025-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-sol.en.pdf", "source_problem_id": "2025-individual-3", "target_subparts": ["a", "b", "full"], "title": "Kuria", "year": 2025}

benchmark/IOL/ioling_hf/reports/manual_verification/batch_02.jsonl

@@ -0,0 +1,22 @@
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_needs_manual_review": 4}, "languages": ["Arabic (Afroasiatic)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 2, "problem_page_end": 2, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["assignment_1.1", "assignment_2.1"], "canonical_answer_chars": 2053, "canonical_excerpt": "Solution of Problem 2\nAll Arabic words in the problem are made according to one of the patterns la2a3t, 11243, 1u23\nand 1u23én (whereby words using the first and the second pattern always come together in this\norder and words using the other two patterns occur on their own). In these patterns 1-2-3 is\none of the triples of consonants r-b-¢, s-b-¢, s-d-s, t-l-t, t-m-n, t-s-¢, z-m-s, ¢-8-r. Let us assume\nthat the consonant triples correspond to numbers between 1 and 10 and the arrangements of the\nvowels indicate certain functions, in particular, /a2a3t i1'2’a3' is either 77 or u (and in either\ncase zamast irmas = + =1), and 1u23 = i and 1u23én = i, for some as yet unknown i and j.\n\nFrom equali", "item_type": "full_problem", "prompt_chars": 862, "record_id": "iol-2003-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 62, "canonical_excerpt": "Assignment 1. (1) $+3=$, 2) $+=¥, 8) $+4=2,0 $+43=$,0) 34+35%.", "item_type": "subproblem", "prompt_chars": 713, "record_id": "iol-2003-individual-p2-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 286, "canonical_excerpt": "Assignment 2. rubs + sadart itsas = 4+ 42 = & and sabast isdas = 2. Thus either\nVrubs + sasart itsas = sabast isdas or, perhaps, rubs+ sasart itsas = (sabast isdas)* (if we don’t\nconsider brackets to be a sign).\n\n1st IOL: Borovetz 03. Solutions to the Problems of the Individual Contest", "item_type": "subproblem", "prompt_chars": 1117, "record_id": "iol-2003-individual-p2-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 286, "canonical_excerpt": "Assignment 2. rubs + sadart itsas = 4+ 42 = & and sabast isdas = 2. Thus either\nVrubs + sasart itsas = sabast isdas or, perhaps, rubs+ sasart itsas = (sabast isdas)* (if we don’t\nconsider brackets to be a sign).\n\n1st IOL: Borovetz 03. Solutions to the Problems of the Individual Contest", "item_type": "partial_variant", "prompt_chars": 1243, "record_id": "iol-2003-individual-p2-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-sol.en.pdf", "source_problem_id": "2003-individual-2", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Arabic Arithmetic", "year": 2003}
+{"author": "Boris Iomdin", "curation_statuses": {"auto_extracted_needs_manual_review": 6}, "languages": [], "problem_extraction_confidence": "ocr_heading", "problem_number": 3, "problem_page_end": 2, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-team-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_2.1", "assignment_3.1"], "canonical_answer_chars": 1535, "canonical_excerpt": "Solution of Problem 3\nAssignment 1. The left column contains what are technically known as performative verbs.\n(The concept of performativity was introduced in 1965 by the English philosopher John Austin.)\nThey are different from other verbs in that the action they name can be performed by their use,\nrather than simply described. So the words ‘I accuse you of murder’ all by themselves constitute\nan accusation; the words I denounce you as an impostor’, a denunciation; ‘I command you to\nreport to the headquarters at once’, a command; ‘I advise you not to go there’, advice; ‘I assure\nyou that this problem is not so hard’, assurance. Performativity is a rather peculiar property;\nas the statement", "item_type": "full_problem", "prompt_chars": 828, "record_id": "iol-2003-team-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 893, "canonical_excerpt": "Assignment 1. The left column contains what are technically known as performative verbs.\n(The concept of performativity was introduced in 1965 by the English philosopher John Austin.)\nThey are different from other verbs in that the action they name can be performed by their use,\nrather than simply described. So the words ‘I accuse you of murder’ all by themselves constitute\nan accusation; the words I denounce you as an impostor’, a denunciation; ‘I command you to\nreport to the headquarters at once’, a command; ‘I advise you not to go there’, advice; ‘I assure\nyou that this problem is not so hard’, assurance. Performativity is a rather peculiar property;\nas the statement of the problem shows,", "item_type": "subproblem", "prompt_chars": 562, "record_id": "iol-2003-team-p3-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 384, "canonical_excerpt": "Assignment 2. These are the verbs forbid (‘I forbid leaving the room before the class is over’),\nswear (‘I swear to cheat no more’), approve (‘I approve of your decision’), refuse (‘I refuse to try\nto solve this problem’), dedicate (‘I dedicate this book to my parents’), give up (‘I can’t do this\nproblem, I give up’), demand (‘I demand to be told how this problem is to be solved’).", "item_type": "subproblem", "prompt_chars": 826, "record_id": "iol-2003-team-p3-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 234, "canonical_excerpt": "Assignment 3. For example, thank (‘I thank you for the clarification’), congratulate (‘I con-\ngratulate you on your success’).\nEdited by Ivan Derzhanski (editor-in-chief), Boris Iomdin, Maria Rubinstein.\nTranslated by Ivan Derzhanski.", "item_type": "subproblem", "prompt_chars": 840, "record_id": "iol-2003-team-p3-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 384, "canonical_excerpt": "Assignment 2. These are the verbs forbid (‘I forbid leaving the room before the class is over’),\nswear (‘I swear to cheat no more’), approve (‘I approve of your decision’), refuse (‘I refuse to try\nto solve this problem’), dedicate (‘I dedicate this book to my parents’), give up (‘I can’t do this\nproblem, I give up’), demand (‘I demand to be told how this problem is to be solved’).", "item_type": "partial_variant", "prompt_chars": 1783, "record_id": "iol-2003-team-p3-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 234, "canonical_excerpt": "Assignment 3. For example, thank (‘I thank you for the clarification’), congratulate (‘I con-\ngratulate you on your success’).\nEdited by Ivan Derzhanski (editor-in-chief), Boris Iomdin, Maria Rubinstein.\nTranslated by Ivan Derzhanski.", "item_type": "partial_variant", "prompt_chars": 2220, "record_id": "iol-2003-team-p3-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-team-sol.en.pdf", "source_problem_id": "2003-team-3", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Verbs", "year": 2003}
+{"author": "Boris Iomdin", "curation_statuses": {"auto_extracted_needs_manual_review": 4}, "languages": ["Tzotzil (Mayan)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["assignment_1.1", "assignment_2.1"], "canonical_answer_chars": 2176, "canonical_excerpt": "Problem 1\n\nAs we analyse the given material we can see that:\n\n1. Affirmative sentences (declarative and interrogative) contain the word ‘oy ‘be, exist’.\nDeclarative sentences begin with this word.\n\n2. General questions begin with mi. Special questions begin with the interrogative words\nbu ‘where’ or k’usi ‘what’.\n\n3. General negative sentences begin with ch’abal. Particular negation is formed by the\nphrases muk’ bu ‘nowhere’ or muk’ k’usi ‘nothing’, which also begin the sentence.\n\n4. The present tense is not marked. The past and the future are marked by the word ‘ox,\nwhich comes after ‘oy, ch’abal, muk’ bu or muk’ k’usi.\n\n5. The person or thing whose (non-)existence or location is stated is ", "item_type": "full_problem", "prompt_chars": 2112, "record_id": "iol-2005-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 290, "canonical_excerpt": "Assignment 1.\n\nCh’abal alekil ‘ixim. You have no good corn.\n\nMi ‘oy ‘ox vob ta k’in? Was there / will there be music at the party?\nK’usi ‘oy ‘ox ta Mexico lavie? What will there be in Mexico today?\n\n“Oy ‘ox k’op ta batz’i k’op ta jna volje. There was a talk in Tzotzil in my house yesterday", "item_type": "subproblem", "prompt_chars": 1856, "record_id": "iol-2005-individual-p1-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 403, "canonical_excerpt": "Assignment 2.\n\nWhere is the party today? Bu ‘oy k’in lavie?\n\nThere was nothing in the pot today. Muk’ k’usi ‘ox ta p’in nax.\n\nYou have a real house. ‘Oy abatz’i na.\n\nWill Juana be in San-Cristobal tomorrow? Mi ‘oy ‘ox li Xunka e ta Jobel ‘ok’ ob?\nHe will soon have no pot. Ch/’abal ‘ox sp’in po’ot.\n\nThird International Olympiad in Linguistics. Solutions to the problems of the individual competition. 2", "item_type": "subproblem", "prompt_chars": 2474, "record_id": "iol-2005-individual-p1-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 403, "canonical_excerpt": "Assignment 2.\n\nWhere is the party today? Bu ‘oy k’in lavie?\n\nThere was nothing in the pot today. Muk’ k’usi ‘ox ta p’in nax.\n\nYou have a real house. ‘Oy abatz’i na.\n\nWill Juana be in San-Cristobal tomorrow? Mi ‘oy ‘ox li Xunka e ta Jobel ‘ok’ ob?\nHe will soon have no pot. Ch/’abal ‘ox sp’in po’ot.\n\nThird International Olympiad in Linguistics. Solutions to the problems of the individual competition. 2", "item_type": "partial_variant", "prompt_chars": 2828, "record_id": "iol-2005-individual-p1-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-sol.en.pdf", "source_problem_id": "2005-individual-1", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Tzotzil", "year": 2005}
+{"author": "Boris Iomdin", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Catalan (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_2.1", "assignment_3.1"], "canonical_answer_chars": 3320, "canonical_excerpt": "Problem #2\n         1.  Articles.  The problem features words with the articles el (pl. els) and la (pl. les).\nClearly these articles precede masculine and feminine nouns, respectively; this can be seen, among\nother things, from the words meaning ‘(male) follower’ and ‘(female) follower’.\n        2.  Stress. We can see that the mark ` is placed over e, o, a and the mark ´ over i, u.\nExamining the words ending in –s, we notice that the stress marks in them are placed on the last\nsyllable or the third one from the end. We can assume that in such words the stressed syllable is\nusually the penultimate, and the stress is explicitly marked when it is not. In all other words stress\nusually falls on", "item_type": "full_problem", "prompt_chars": 2290, "record_id": "iol-2006-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 774, "canonical_excerpt": "Assignment 1.\nel bastaix       els bastaixos\nel troleibús     els troleibusos\nla clos           les clos\nel iris           els iris\nel llaç           els llaços\nel ònix          els ònixs\nla pelvis        les pelvis\nel pis           els pisos\nel                els sequaços    sequaç\nla                les sequaçs\nel tastaolletes  els tastaolletes\nla xeix          les xeixs\n        Theoretically one might derive the forms *la clo and *el tastaollete, but those are poorly\ncompatible with the statement of the problem, which talks of nouns ending in the letters s, x or ç\nonly. The forms *el sequaços and *la sequaçs are also wrong, because it is hinted that the words\nfor ‘(male) follower’ and ‘(fe", "item_type": "subproblem", "prompt_chars": 1843, "record_id": "iol-2006-individual-p2-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 89, "canonical_excerpt": "Assignment 2. The marks ` and ´ are used when the stress falls on a non-default syllable.", "item_type": "subproblem", "prompt_chars": 1953, "record_id": "iol-2006-individual-p2-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 315, "canonical_excerpt": "Assignment 3. When there is no ` or ´ mark, the stress is penultimate if the word ends in s,\notherwise it is final.\n                                                      Problem and solution: Boris Iomdin\n\nFourth International Linguistic Olympiad. Solutions of the Problems of the Individual Contest               3", "item_type": "subproblem", "prompt_chars": 2629, "record_id": "iol-2006-individual-p2-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 89, "canonical_excerpt": "Assignment 2. The marks ` and ´ are used when the stress falls on a non-default syllable.", "item_type": "partial_variant", "prompt_chars": 2791, "record_id": "iol-2006-individual-p2-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 315, "canonical_excerpt": "Assignment 3. When there is no ` or ´ mark, the stress is penultimate if the word ends in s,\notherwise it is final.\n                                                      Problem and solution: Boris Iomdin\n\nFourth International Linguistic Olympiad. Solutions of the Problems of the Individual Contest               3", "item_type": "partial_variant", "prompt_chars": 3595, "record_id": "iol-2006-individual-p2-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-sol.en.pdf", "source_problem_id": "2006-individual-2", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Catalan", "year": 2006}
+{"author": "Yakov Testelets", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Georgian (South Caucasian)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 4949, "canonical_excerpt": "Problem #3\nWe break the Georgian words into their components. We derive         amb             -ob\nthe suffixes -e and -ob by comparing vinadire and nadirob, and    v-    i-  gor              -eb\n                                                                                     keni                          -tthe suffix -eb by contrasting visadileb and vsadilob. We can’t\n                                                                               v-      keni\ntell  if -ob  is contained in ambob, -eb in vigoreb, or -e in                                                                                                                   i-   nadir              -e     -t\n(v)itavmGdomareve, be", "item_type": "full_problem", "prompt_chars": 966, "record_id": "iol-2007-individual-p3", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-sol.en.pdf", "source_problem_id": "2007-individual-3", "target_subparts": ["full"], "title": "Georgian", "year": 2007}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Drehu (Austronesian) Cemuhî (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 1238, "canonical_excerpt": "Problem #3.  The modifier follows its head in both languages.\n\n (a)  jun            bone\n      i-jun             skeleton            (multitude of bones)\n     i-wahnawa     bunch of bananas   (multitude of bananas)\n       i-drai           calendar           (multitude of days)\n      drai-hmitrötr   Sunday             (holy day)\n      gaa-hmitrötr   sanctuary          (holy place)\n     uma-hmitrötr   church             (holy house)\n     ngöne-uma      wall               (house border)\n      ngöne-gejë      coast              (water border)\n      nyine-thin      awl                  (tool to poke)\n        tii               to write\n       bé-tii             pencil               (tool to write)\n", "item_type": "full_problem", "prompt_chars": 1945, "record_id": "iol-2008-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 1021, "canonical_excerpt": "(a)  jun            bone\n      i-jun             skeleton            (multitude of bones)\n     i-wahnawa     bunch of bananas   (multitude of bananas)\n       i-drai           calendar           (multitude of days)\n      drai-hmitrötr   Sunday             (holy day)\n      gaa-hmitrötr   sanctuary          (holy place)\n     uma-hmitrötr   church             (holy house)\n     ngöne-uma      wall               (house border)\n      ngöne-gejë      coast              (water border)\n      nyine-thin      awl                  (tool to poke)\n        tii               to write\n       bé-tii             pencil               (tool to write)\n      bé-wöli           fork                 (tool to poke)\n   ", "item_type": "subproblem", "prompt_chars": 1260, "record_id": "iol-2008-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 67, "canonical_excerpt": "(b) wahnawa ‘banana’, drai ‘day’; wöli ‘to poke’, pulut ‘to sleep’.", "item_type": "subproblem", "prompt_chars": 1374, "record_id": "iol-2008-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 81, "canonical_excerpt": "(c) i-bii ‘swarm of bees (multitude of bees)’, tusi-hmitrötr ‘Bible (holy book)’.", "item_type": "subproblem", "prompt_chars": 2508, "record_id": "iol-2008-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 67, "canonical_excerpt": "(b) wahnawa ‘banana’, drai ‘day’; wöli ‘to poke’, pulut ‘to sleep’.", "item_type": "partial_variant", "prompt_chars": 2448, "record_id": "iol-2008-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 81, "canonical_excerpt": "(c) i-bii ‘swarm of bees (multitude of bees)’, tusi-hmitrötr ‘Bible (holy book)’.", "item_type": "partial_variant", "prompt_chars": 3677, "record_id": "iol-2008-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-sol.en.pdf", "source_problem_id": "2008-individual-3", "target_subparts": ["a", "b", "c", "full"], "title": "Drehu and Cemuhî", "year": 2008}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Burmese (Sino-Tibetan)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-prob.en-gb.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 820, "canonical_excerpt": "Problem #3. We can see that the names of the children born on the same day of the week\n  have similar first sounds:\n     • Monday: kau6 myaP, khi6 le nwE, khai6 mi6 thu6, keþi thu6\n     • Tuesday: zeiya cO, su myaP so, susu wi6, sha6 thu6, shu ma6 cO\n     • Wednesday: wi6 i mu6, lwi6 koko, wi6 cO au6, yadana u, yi6yi6 myi6\n     • Thursday: pa6 we, pyesou6 au6, mimi khai6, phou6 nai6 thu6, myo khi6 wi6\n     • Saturday: thou6 u6, ne li6, ti6 mau6 laP, thEP au6, ti6 za mO\n  Answers:\n     • Nwe si6þu — 13/07/2009 (Monday);\n     • so mo cO — 16/06/2009 (Tuesday);\n     • yE au6 nai6 — 24/06/2009 (Wednesday),\n     • daliya — 18/07/2009 (Saturday),\n     • e ti6 — 14/06/2009 (Sunday: there are no Sun", "item_type": "full_problem", "prompt_chars": 2807, "record_id": "iol-2009-individual-p3", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-sol.en-gb.pdf", "source_problem_id": "2009-individual-3", "target_subparts": ["full"], "title": "Burmese Names", "year": 2009}
+{"author": "Alexander Piperski", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": [], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 1982, "canonical_excerpt": "Problem #3.       : noun,      : adjective,      : verb (if there is more than one symbol in the\nword, the mark is placed above the leftmost one).\n   Pointers (  ,   ,   ,  ) are used to refer to specific parts of the symbols.\n\n (a)\n              part of speech  composition                meaning\n\n              verb          mouth + nose                to breathe\n            noun           water + mouth                 saliva\n\n               adjective         circle (sun) + pointer        western\n\n               adjective        activity                        active\n            noun         body (torso) + 2 pointers     waist\n\n              verb          mouth + (air + outwards)    to blow\n\n", "item_type": "full_problem", "prompt_chars": 982, "record_id": "iol-2010-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 826, "canonical_excerpt": "(a)\n              part of speech  composition                meaning\n\n              verb          mouth + nose                to breathe\n            noun           water + mouth                 saliva\n\n               adjective         circle (sun) + pointer        western\n\n               adjective        activity                        active\n            noun         body (torso) + 2 pointers     waist\n\n              verb          mouth + (air + outwards)    to blow\n\n               adjective            ill, sick                                     ill, sick\n            noun         mouth + 2 pointers             lips\n\n              verb            eye + (water + downwards)   to weep\n        ", "item_type": "subproblem", "prompt_chars": 683, "record_id": "iol-2010-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 449, "canonical_excerpt": "(b)\n            part of speech  composition               meaning\n          noun           nose                         nose\n          noun           water                         water, liquid\n          noun         body (torso) + pointer       neck\n\n            verb             activity                       to act, be active\n          noun           eye with eyebrow + pointer   eyebrow\n          noun          head with neck + pointer     neck", "item_type": "subproblem", "prompt_chars": 795, "record_id": "iol-2010-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 473, "canonical_excerpt": "(c)\n             part of speech  composition          meaning\n           noun             air                        air\n           noun         body (torso)         body (torso)\n\n            verb          upwards                to rise\n           noun             circle (sun) + pointer   east\n\n              adjective        heart + downwards    sad\n\nEighth International Olympiad in Linguistics (2010).                                      3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1159, "record_id": "iol-2010-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 449, "canonical_excerpt": "(b)\n            part of speech  composition               meaning\n          noun           nose                         nose\n          noun           water                         water, liquid\n          noun         body (torso) + pointer       neck\n\n            verb             activity                       to act, be active\n          noun           eye with eyebrow + pointer   eyebrow\n          noun          head with neck + pointer     neck", "item_type": "partial_variant", "prompt_chars": 1674, "record_id": "iol-2010-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 473, "canonical_excerpt": "(c)\n             part of speech  composition          meaning\n           noun             air                        air\n           noun         body (torso)         body (torso)\n\n            verb          upwards                to rise\n           noun             circle (sun) + pointer   east\n\n              adjective        heart + downwards    sad\n\nEighth International Olympiad in Linguistics (2010).                                      3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2515, "record_id": "iol-2010-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-sol.en.pdf", "source_problem_id": "2010-individual-3", "target_subparts": ["a", "b", "c", "full"], "title": "Blissymbolics", "year": 2010}
+{"author": "Olga Kuznetsova", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Vai (Mande)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 803, "canonical_excerpt": "Problem #3.  Rules:\n\n   1. Adjectives follow their nouns.\n\n   2. A noun (or the adjective if there is one) gets the marker -ˇE, unless it is inalienably possessed\n     (body part, kinship term); in the latter case it is preceded by the possessor.\n\n   3. Alienable possession is expressed by ´a between the possessor and the possessed.\n\n   4. In compound nouns the last syllable has low tone (“`”).\n\n (a) m`us´uˇE ´a gb`om`uˇE: the woman’s fish\n       l´eN k´und´uˇE ´a ny`ım`ı`ıˇE: the short child’s snake\n      gb`om`u-l`End`E k´und´uˇE: the short boat\n\n (b) k´and`O-l`End`E l`O`OˇE: the small airplane\n\n (c) the eagle’s snake: k`O´anj`aˇE ´a ny`ım`ı`ıˇE\n     the small child’s eye: l´eN l`O`OˇE j´a\n", "item_type": "full_problem", "prompt_chars": 1668, "record_id": "iol-2011-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 152, "canonical_excerpt": "(a) m`us´uˇE ´a gb`om`uˇE: the woman’s fish\n       l´eN k´und´uˇE ´a ny`ım`ı`ıˇE: the short child’s snake\n      gb`om`u-l`End`E k´und´uˇE: the short boat", "item_type": "subproblem", "prompt_chars": 1280, "record_id": "iol-2011-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 47, "canonical_excerpt": "(b) k´and`O-l`End`E l`O`OˇE: the small airplane", "item_type": "subproblem", "prompt_chars": 1260, "record_id": "iol-2011-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 198, "canonical_excerpt": "(c) the eagle’s snake: k`O´anj`aˇE ´a ny`ım`ı`ıˇE\n     the small child’s eye: l´eN l`O`OˇE j´a\n     the tall man’s sister: k`a´ı jˇaNˇE l`O`O-m`us`u\n     the small baby-snake: ny`ım`ı`ı-l`eN l`O`OˇE", "item_type": "subproblem", "prompt_chars": 1890, "record_id": "iol-2011-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 47, "canonical_excerpt": "(b) k´and`O-l`End`E l`O`OˇE: the small airplane", "item_type": "partial_variant", "prompt_chars": 1465, "record_id": "iol-2011-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 198, "canonical_excerpt": "(c) the eagle’s snake: k`O´anj`aˇE ´a ny`ım`ı`ıˇE\n     the small child’s eye: l´eN l`O`OˇE j´a\n     the tall man’s sister: k`a´ı jˇaNˇE l`O`O-m`us`u\n     the small baby-snake: ny`ım`ı`ı-l`eN l`O`OˇE", "item_type": "partial_variant", "prompt_chars": 2170, "record_id": "iol-2011-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-sol.en-gb.pdf", "source_problem_id": "2011-individual-3", "target_subparts": ["a", "b", "c", "full"], "title": "Vai", "year": 2011}
+{"author": "Natalya Zaika", "curation_statuses": {"auto_extracted_text_high_confidence": 8}, "languages": ["Basque (Isolate)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-prob.en.pdf", "record_count": 8, "record_types": {"full_problem": 1, "partial_variant": 3, "subproblem": 4}, "records": [{"answer_unit_count": 4, "answer_unit_ids": ["a.1", "b.1", "c.1", "d.1"], "canonical_answer_chars": 1594, "canonical_excerpt": "Problem #3.\n      1st person sg   1st person pl  2nd person sg  2nd person pl  3rd person sg  3rd person pl\n A  nau-                                               du-           ditu-\n B  natzai-       gatzaizki-                                   zai-           zaizki-\n Z               -t          -gu           -zu          -zue                            -te\n        A    B        Z\n ahaztu  — about whom —  who       forgot\n hurbildu         who      whom     approached\n lagundu   whom           who       helped\n mintzatu        who        to whom  talked\n ukitu     whom           who      touched\n (a)\n    ahaztu ditut                          lagundu dute            they helped him          ", "item_type": "full_problem", "prompt_chars": 1329, "record_id": "iol-2012-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 654, "canonical_excerpt": "(a)\n    ahaztu ditut                          lagundu dute            they helped him                                   I forgot them    ahaztu zaizkit !                      lagundu nauzue         you(pl) helped me\n    ahaztu zaizu         you(sg) forgot him    mintzatu natzaizu        I talked to you(sg)\n    hurbildu natzaizue   I approached you(pl)  mintzatu gatzaizkizue  we talked to you(pl)\n    hurbildu zait        he approached me    mintzatu zaizkigu       they talked to us\n    lagundu ditugu     we helped them       ukitu ditugu          we touched them\n    lagundu dituzu       you(sg) helped them   ukitu naute             they touched me", "item_type": "subproblem", "prompt_chars": 931, "record_id": "iol-2012-individual-p3-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 76, "canonical_excerpt": "(b) you(sg) touched me — ukitu nauzu, they approached me — hurbildu zaizkit.", "item_type": "subproblem", "prompt_chars": 979, "record_id": "iol-2012-individual-p3-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 77, "canonical_excerpt": "(c) lagundu dut — I helped him, hurbildu gatzaizkizu — we approached you(sg).", "item_type": "subproblem", "prompt_chars": 971, "record_id": "iol-2012-individual-p3-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 173, "canonical_excerpt": "(d) you(sg) forgot him (ahaztu zaizu) — ahaztu duzu.\n\nTenth International Olympiad in Linguistics (2012).                                      3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1539, "record_id": "iol-2012-individual-p3-sub-d", "target_subparts": ["d"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 76, "canonical_excerpt": "(b) you(sg) touched me — ukitu nauzu, they approached me — hurbildu zaizkit.", "item_type": "partial_variant", "prompt_chars": 1686, "record_id": "iol-2012-individual-p3-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 77, "canonical_excerpt": "(c) lagundu dut — I helped him, hurbildu gatzaizkizu — we approached you(sg).", "item_type": "partial_variant", "prompt_chars": 1782, "record_id": "iol-2012-individual-p3-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 173, "canonical_excerpt": "(d) you(sg) forgot him (ahaztu zaizu) — ahaztu duzu.\n\nTenth International Olympiad in Linguistics (2012).                                      3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2455, "record_id": "iol-2012-individual-p3-partial-through-abc-target-d", "target_subparts": ["d"], "visible_solved_subparts": ["a", "b", "c"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-sol.en.pdf", "source_problem_id": "2012-individual-3", "target_subparts": ["a", "b", "c", "d", "full"], "title": "Basque", "year": 2012}
+{"author": "ArtÅ«rs Semeņuks", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Pirahã (Mura)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-prob.en-gb.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 5, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4", "full.5"], "canonical_answer_chars": 2432, "canonical_excerpt": "Problem #3.  Rules:\n\n   1. x [P].\n\n   2. A noun and a following modifier are pronounced as one word, but at the end of the first\n     word of the phrase [i] after a vowel is lost and in the beginning of the second word [P] is\n       lost.                                      xisitai xagai [Pisitai/⌣P/agai] →[Pisitaagai]\n\n   3. 1 syllable = CVV, CV or VV (C = consonant, V = vowel). The syllabification starts\n     from the end of the word.\n\n                        xiiaapisi →[Piiaapisi] →[Piiaapi.si] →[Piiaa.pi.si] →[Pii.aa.pi.si]\n                        hixi xitaixi [hiPi⌣P/itaiPi] →[hiPiitaiPi] →. . . →[hi.Pii.tai.Pi]\n                                                      (\n   4. Syllable weight", "item_type": "full_problem", "prompt_chars": 3222, "record_id": "iol-2013-individual-p3", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-sol.en-gb.pdf", "source_problem_id": "2013-individual-3", "target_subparts": ["full"], "title": "Pirahã", "year": 2013}
+{"author": "ArtÅ«rs Semeņuks", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Engenni (Niger-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 4, "answer_unit_ids": ["a.1", "a.2", "b.1", "c.1"], "canonical_answer_chars": 2207, "canonical_excerpt": "Problem #4.  Rules:\n\n   • Word order:\n\n                         in questions                   (S = subject; V = verb; O = object)\n                              ↓\n\n           S1 wei ga  S2 V O  à\n\n         ↑\n           in answers\n       N A D               (N = noun; A = adjective; D = demonstrative adjective)\n\n   • Choice of pronoun (S2):\n\n          S1 = S2 ⇒S2 = òkì; S1̸ = S2 ⇒S2 = o\n\n   • Vowel lowering:    (W. = word with lowered vowels; W = word without lowered vowels)\n\n                 . . . o W. . . . > . . .  o. W. . . .                           . . . W. âno . . . > . . . W. â. no . . .\n                 . . . o W . . . > . . . o W . . .                           . . . W âno . . . > ", "item_type": "full_problem", "prompt_chars": 3104, "record_id": "iol-2014-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 368, "canonical_excerpt": "(a)    8. Did this man frighten the thief?\n        The girl said that hethis man did frighten the thief.\n         9. Will the deceived girl kill this youth?\n        The deceived girl said that shethe deceived girl would kill this youth.\n       10. This thief said that the child did not heal the man. or\n          This thief said that the child would not heal the man.", "item_type": "subproblem", "prompt_chars": 2586, "record_id": "iol-2014-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 234, "canonical_excerpt": "(b)  11. okàa kí.nonò a. modhyòmú ko lomù ânò à?\n                a. mò wei ga ó. ki.nono a. modhyòmú ko lomù ânò.\n       12.  a. vùràmú nu. amù â. nó nwa. sese edèì à?\n                a. vùràmú nu. amù â. nò wei ga òkí nwa. sese edèì.", "item_type": "subproblem", "prompt_chars": 2262, "record_id": "iol-2014-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 483, "canonical_excerpt": "(c) Nouns are not a ected by any grammatical tone changes when placed at the very end of\n     the sentence, before à or before ânò. Because of this, the base form for `thief' is ozyi\n       (7.), and the base form for `girl' is amemùrè (2.).\n\n   1Usually has negative past interpretation, but can denote negative future if the surrounding context re-\nquires it.\n\nTwelfth International Olympiad in Linguistics (2014).                                     4\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3080, "record_id": "iol-2014-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 234, "canonical_excerpt": "(b)  11. okàa kí.nonò a. modhyòmú ko lomù ânò à?\n                a. mò wei ga ó. ki.nono a. modhyòmú ko lomù ânò.\n       12.  a. vùràmú nu. amù â. nó nwa. sese edèì à?\n                a. vùràmú nu. amù â. nò wei ga òkí nwa. sese edèì.", "item_type": "partial_variant", "prompt_chars": 2683, "record_id": "iol-2014-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 483, "canonical_excerpt": "(c) Nouns are not a ected by any grammatical tone changes when placed at the very end of\n     the sentence, before à or before ânò. Because of this, the base form for `thief' is ozyi\n       (7.), and the base form for `girl' is amemùrè (2.).\n\n   1Usually has negative past interpretation, but can denote negative future if the surrounding context re-\nquires it.\n\nTwelfth International Olympiad in Linguistics (2014).                                     4\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3763, "record_id": "iol-2014-individual-p4-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-sol.en-gb.pdf", "source_problem_id": "2014-individual-4", "target_subparts": ["a", "b", "c", "full"], "title": "Engenni", "year": 2014}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Wambaya (West Barkly)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 9, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 1903, "canonical_excerpt": "Problem 4.  Rules:\n\n   • Word order: V P (S/O); S/O P V Poss, V P Poss; S Poss.\n\n   • V = verb (past →future: -bi →-ba, ∅→-jba).\n\n   • S = subject (noun). The subject of a transitive verb gets the ending -ni.\n\n   • O = object (noun).\n\n   • P = pronouns (subject + object) + tense:\n\n      – subject:\n\n           * 1st ngV-,\n           * 2nd nyV-,\n\n          \n           intransitive verb:  gV-{\n           * 3rd                       masculine gVnV-\n           transitive verb:                                         feminine  ngVyV-;\n\n      – object: 1st -ngV, 2nd -nyV, 3rd -∅;\n\nThirteenth International Olympiad in Linguistics (2015).                                  4\nIndividual Contest Sol", "item_type": "full_problem", "prompt_chars": 2159, "record_id": "iol-2015-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5"], "canonical_answer_chars": 424, "canonical_excerpt": "(a)   1. Alayulujba nguyunyu bungmanyani.  The old woman will find you (sg.).\n        2. Yagu gininya.                    He left you (sg.).\n        3. Janji darrangguwaji.               The dog doesn’t have a stick.\n        4. Ngirra nya alanga.                You (sg.) stole the girl.\n        5. Daguma nyinga.                   You (sg.) struck me.\n        6. Dirragbi ga balamurrungunya.       She jumped with the spear.", "item_type": "subproblem", "prompt_chars": 1885, "record_id": "iol-2015-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 375, "canonical_excerpt": "(b)    7. You (sg.) will leave me.                Yagujba nyungu.\n         8. The doctor slept.                      Gulugbi ga ngunybulugi.\n         9. The man will run (away) with the money.  Juwa gu bardba gijilulunguji.\n       10. He will steal the dog.                   Ngirrajba gunu janji.\n       11. The girl saw you (sg.).                  Ngajbi ngiyinya alangani.", "item_type": "subproblem", "prompt_chars": 2513, "record_id": "iol-2015-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 375, "canonical_excerpt": "(b)    7. You (sg.) will leave me.                Yagujba nyungu.\n         8. The doctor slept.                      Gulugbi ga ngunybulugi.\n         9. The man will run (away) with the money.  Juwa gu bardba gijilulunguji.\n       10. He will steal the dog.                   Ngirrajba gunu janji.\n       11. The girl saw you (sg.).                  Ngajbi ngiyinya alangani.", "item_type": "partial_variant", "prompt_chars": 2990, "record_id": "iol-2015-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-sol.en-gb.pdf", "source_problem_id": "2015-individual-4", "target_subparts": ["a", "b", "full"], "title": "Wambaya", "year": 2015}
+{"author": "ArtÅ«rs Semeņuks", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Iatmül (Sepik)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 6, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 19, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "b.1", "b.2", "c.1", "c.2", "c.3"], "canonical_answer_chars": 1905, "canonical_excerpt": "Problem 4.\n\n (a)    1.  gu-na vaala        water- possessor canoe             c.  canoe\n         2.  ka’ik              picture/shadow                   g.   picture, shadow\n         3.  ka’ik-gu           picture+water                           j.  mirror\n         4.  kla-wun            to get- 1st person sg                       i.   I got\n         5.  laavu             banana                            a.  banana\n         6.  laavu-ga vi’       book+to see                     n.   to read\n         7.  laavu-ga          banana+leaves                  d.  book\n         8.  ni’bu              ground, land                      o.  ground, land\n         9.  ni’bu-na vaala     ground, land- pos", "item_type": "full_problem", "prompt_chars": 2239, "record_id": "iol-2016-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 14, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14"], "canonical_answer_chars": 1266, "canonical_excerpt": "(a)    1.  gu-na vaala        water- possessor canoe             c.  canoe\n         2.  ka’ik              picture/shadow                   g.   picture, shadow\n         3.  ka’ik-gu           picture+water                           j.  mirror\n         4.  kla-wun            to get- 1st person sg                       i.   I got\n         5.  laavu             banana                            a.  banana\n         6.  laavu-ga vi’       book+to see                     n.   to read\n         7.  laavu-ga          banana+leaves                  d.  book\n         8.  ni’bu              ground, land                      o.  ground, land\n         9.  ni’bu-na vaala     ground, land- possessor canoe ", "item_type": "subproblem", "prompt_chars": 1589, "record_id": "iol-2016-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 451, "canonical_excerpt": "(b)    1. vaala — canoe\n\n         2. gu — water\n\n         3. vi’ — to see\n\n    The word vaala is a synonym of guna vaala ‘canoe’. The later expression guna vaala\n        (lit. ‘water canoe’) came into use because the Iatmül people were introduced to cars\n     (and also planes), and vaala has slowly started to expand its meaning.\n\nFourteenth International Linguistics Olympiad (2016).                                    5\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1967, "record_id": "iol-2016-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 169, "canonical_excerpt": "(c)    1. pig — bâk\n\n         2. banana leaves ≡book — laavuga (because of similarity of shape and material)\n\n         3. sun — nya\n\n         4. white people — walini’ba", "item_type": "subproblem", "prompt_chars": 2501, "record_id": "iol-2016-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 451, "canonical_excerpt": "(b)    1. vaala — canoe\n\n         2. gu — water\n\n         3. vi’ — to see\n\n    The word vaala is a synonym of guna vaala ‘canoe’. The later expression guna vaala\n        (lit. ‘water canoe’) came into use because the Iatmül people were introduced to cars\n     (and also planes), and vaala has slowly started to expand its meaning.\n\nFourteenth International Linguistics Olympiad (2016).                                    5\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3286, "record_id": "iol-2016-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 169, "canonical_excerpt": "(c)    1. pig — bâk\n\n         2. banana leaves ≡book — laavuga (because of similarity of shape and material)\n\n         3. sun — nya\n\n         4. white people — walini’ba", "item_type": "partial_variant", "prompt_chars": 4299, "record_id": "iol-2016-individual-p4-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-sol.en.pdf", "source_problem_id": "2016-individual-4", "target_subparts": ["a", "b", "c", "full"], "title": "Iatmül", "year": 2016}
+{"author": "Tae Hun Lee", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Laven (Jru') (Austroasiatic)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1397, "canonical_excerpt": "Problem 4.  Word order: [[‹class›] ‹species›] [‹numeral› ‹class›].\n   The direction of writing is from left to right. One syllable = two characters:\n\n C(C)-                               -V(V)(C)\n\n     b,     bl,     c,    hn,    k,    kr,      l,       aː,    aːk,     aj,     at,    aw,     əːr,\n\n     p,     pl,    pr,     t,    tr                       ɛː,     ie,     iet,     ɨət,     oː,    uan,\n\n                                                 uat,   ʌh\n\n   Answers:\n\n  3  →         ←         taw bɛː                 to see the raft\n\n  6 →          ←  →plaj prɨət pɛː plaj←         three bananas\n\n  7                   →hnaː traw toː ←          six rhinoceros\n\n  8               →  puan lʌh ", "item_type": "full_problem", "prompt_chars": 2015, "record_id": "iol-2017-individual-p4", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-sol.en.pdf", "source_problem_id": "2017-individual-4", "target_subparts": ["full"], "title": "Khom script/Laven", "year": 2017}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Mountain Arapesh (Torricelli)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "c.1", "c.2", "c.3"], "canonical_answer_chars": 2287, "canonical_excerpt": "Problem 4.\n                      singular   plural\n                   -b       -(a)bys  bø-bys ‘betel nuts’, juhuru-bys ‘eggs’\n                   -h       -(a)ruh  ara-ruh ‘sugar canes’\n                    -g       -(a)gas   bai-gas ‘sheets of sago bark’\n                       -t       -(o)gu   nyba-t ‘4’\n\n                         atu-        1        baig(as)\n  anowip          6     bi-          2       bøb(ys)⎛                        ⎞  ⎧\n            2           =                   12                 +                                      3                                                                    1              + 2  wiwis                                                           ", "item_type": "full_problem", "prompt_chars": 1934, "record_id": "iol-2018-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 72, "canonical_excerpt": "(a) (6, 9, 11): 4 = nybat, 8 = biogu nybat, k × 4 (3 ≤k) = k-t/gu nybat.", "item_type": "subproblem", "prompt_chars": 1746, "record_id": "iol-2018-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 230, "canonical_excerpt": "(b)  12. biabys atub bøbys — 3 betel nuts\n       13. atub juhurub — 1 egg\n       14. nybat atuh araruh — 5 sugar canes\n       15. biagas atugas biagas baigas — 5 sheets of sago bark\n       16. wiwis baigas — 12 sheets of sago bark", "item_type": "subproblem", "prompt_chars": 1276, "record_id": "iol-2018-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 207, "canonical_excerpt": "(c)  17. 4 betel nuts — biabys atubys atub bøbys\n       18. 11 eggs — biogu nybat biabys atub juhurubys\n       19. 20 eggs — biogu atugu biogu juhurubys\n       20. 25 sheets of sago bark — araman atug baigas", "item_type": "subproblem", "prompt_chars": 1850, "record_id": "iol-2018-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4"], "canonical_answer_chars": 230, "canonical_excerpt": "(b)  12. biabys atub bøbys — 3 betel nuts\n       13. atub juhurub — 1 egg\n       14. nybat atuh araruh — 5 sugar canes\n       15. biagas atugas biagas baigas — 5 sheets of sago bark\n       16. wiwis baigas — 12 sheets of sago bark", "item_type": "partial_variant", "prompt_chars": 1401, "record_id": "iol-2018-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 207, "canonical_excerpt": "(c)  17. 4 betel nuts — biabys atubys atub bøbys\n       18. 11 eggs — biogu nybat biabys atub juhurubys\n       19. 20 eggs — biogu atugu biogu juhurubys\n       20. 25 sheets of sago bark — araman atug baigas", "item_type": "partial_variant", "prompt_chars": 2233, "record_id": "iol-2018-individual-p4-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-sol.en.pdf", "source_problem_id": "2018-individual-4", "target_subparts": ["a", "b", "c", "full"], "title": "Mountain Arapesh (Bukiyip)", "year": 2018}
+{"author": "Elysia Warner", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["West Tarangan (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 858, "canonical_excerpt": "Problem 4.  ↓C1VC2 (V is the stressed vowel)\n\n   • North dialect: R = C1V́C2, but\n\n      – C1 = C2 ⇒R = C1V\n      – there is a vowel before C1 ⇒R = C2\n\n   • Coast dialect: R = C1V́1C2V2, but\n\n      – ¬∃V2 ⇒R = C1V1C2\n      – ∃V2 ∧C2 ∈{C1, k, g, ŋ, w} ⇒R = C1V1\n\n                                                (C1, C2 are consonants; V, V1, V2 are vowels)\n\n    Answers:    North dialect  Coast dialect\n    ɛtaleŋa    ɛtaŋleŋa      ɛtaleleŋa\n    jaga        jagjaga       jajaga\n    gasírana   garsírana     gasirasírana\n    daramota  daratmota   daramotamota\n   pɔwna    pɔwpɔwna   pɔwpɔwna\n    ɛrtɔpa     ɛrtɔptɔpa    ɛrtɔpatɔpa\n    dabuka    dakbuka    dabubuka\n   ŋuŋim    ŋuŋuŋim    ŋuŋuŋim\n     ", "item_type": "full_problem", "prompt_chars": 2639, "record_id": "iol-2019-individual-p4", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-sol.en.pdf", "source_problem_id": "2019-individual-4", "target_subparts": ["full"], "title": "West Tarangan", "year": 2019}
+{"author": "Eimear McKnight", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Ikwerre (Atlantic-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 897, "canonical_excerpt": "Problem 4.  tV- before CV0:\n\n   •\n                       C = y                     C̸ ∈{b, m, w, y}          C ∈{b, m, w}\n                               V0 ∈{e, ɛ, i, ɪ}  V0 ∈{o, ɔ, u, ʊ, a}\n      V0 ∈{e, i, o, u}          V = i                  V = u\n      V0 ∈{ɛ, ɔ, ɪ, ʊ, a}        V = ɪ                  V = ʊ\n\n   • Ṽ↔Ṽ0\n\n       ? ?     ? ?     ? ?\n   • CV́0 > tV́-CV́0 ; CV̂0 > tV́-CV̀0 ; CV̀0 > tV̀-CV̀0\n\n   • N-CV0 (N ∈{m, n, ɲ, ŋ}) > n-tV-CV0\n\n (a)   tі̃́mі̃́tі̃́nі̃́ (*tṹmі̃́tі̃́nі̃́)\n\n (b)    1.  rʊ́tʊ̂  — tʊ́rʊ́tʊ́tʊ̀\n\n         2.  rɛ̀wʊ̃́  — tɪ̀rɛ̀tʊ̃́wʊ̃́\n\n         3. bі́kó  — túbі́túkó\n\n         4.  ɲ́yẽ́ rɪ̀yã̀ — ńtі̃́yẽ́ tɪ̀rɪ̀tɪ̃̀yã̀\n\n         5.  bɛ́kʷʊ̃́ — tʊ́bɛ́tʊ̃́kʷʊ̃́\n\n    ", "item_type": "full_problem", "prompt_chars": 2860, "record_id": "iol-2021-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 40, "canonical_excerpt": "(a)   tі̃́mі̃́tі̃́nі̃́ (*tṹmі̃́tі̃́nі̃́)", "item_type": "subproblem", "prompt_chars": 2062, "record_id": "iol-2021-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 401, "canonical_excerpt": "(b)    1.  rʊ́tʊ̂  — tʊ́rʊ́tʊ́tʊ̀\n\n         2.  rɛ̀wʊ̃́  — tɪ̀rɛ̀tʊ̃́wʊ̃́\n\n         3. bі́kó  — túbі́túkó\n\n         4.  ɲ́yẽ́ rɪ̀yã̀ — ńtі̃́yẽ́ tɪ̀rɪ̀tɪ̃̀yã̀\n\n         5.  bɛ́kʷʊ̃́ — tʊ́bɛ́tʊ̃́kʷʊ̃́\n\n         6.  rі́yõ̀rõ̀ — tі́rі́tі̃̀yõ̀tũ̀rõ̀\n\n         7. hʷɔ̀gâ — tʊ̀hʷɔ̀tʊ́gà\n\nEighteenth International Linguistics Olympiad (2021)                                     4\n Individual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3738, "record_id": "iol-2021-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 401, "canonical_excerpt": "(b)    1.  rʊ́tʊ̂  — tʊ́rʊ́tʊ́tʊ̀\n\n         2.  rɛ̀wʊ̃́  — tɪ̀rɛ̀tʊ̃́wʊ̃́\n\n         3. bі́kó  — túbі́túkó\n\n         4.  ɲ́yẽ́ rɪ̀yã̀ — ńtі̃́yẽ́ tɪ̀rɪ̀tɪ̃̀yã̀\n\n         5.  bɛ́kʷʊ̃́ — tʊ́bɛ́tʊ̃́kʷʊ̃́\n\n         6.  rі́yõ̀rõ̀ — tі́rі́tі̃̀yõ̀tũ̀rõ̀\n\n         7. hʷɔ̀gâ — tʊ̀hʷɔ̀tʊ́gà\n\nEighteenth International Linguistics Olympiad (2021)                                     4\n Individual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3831, "record_id": "iol-2021-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-sol.en.pdf", "source_problem_id": "2021-individual-4", "target_subparts": ["a", "b", "full"], "title": "Agbirigba", "year": 2021}
+{"author": "Tom McCoy", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Arabana (Pama–Nyungan)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 2044, "canonical_excerpt": "Problem 4.                 △ ⃝△ ⃝\n                                                                          apityi              lhuka\n     △ ⃝ △ ∼ ⃝ ∼  ∼  ∼\n               nhupa nhupa   nhuthi kupaka   kaku  kupaka      piyaka      wardu\n △ ⃝  △ ⃝   △ ⃝  △ ⃝\n          amanyi     kadnhini                       kadnhini     thanti\n  △ ⃝  △ ⃝ △ ⃝  △ ⃝\n   ∼    ∼   ∼    ∼\n          amanyi             kadnhini      kadnhini               thanti\n\n                                                        I ⃝Q                              (b)   (i) nhupa (a)       △\n                                1     2                                                      (ii) piyaka\n                                 ", "item_type": "full_problem", "prompt_chars": 3287, "record_id": "iol-2022-individual-p4", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-sol.en.pdf", "source_problem_id": "2022-individual-4", "target_subparts": ["full"], "title": "Arabana", "year": 2022}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Plains Cree (Algic)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 6, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4", "full.5", "full.6", "full.7"], "canonical_answer_chars": 1408, "canonical_excerpt": "Problem 4.  S = subject, O = object\n\n                                      S  O\n\n          2nd person̸ ∈{S, O}               1|2   3   -ā(w)\n        ni-:\n   •       ∧1st person ∈{S, O}  + root +  1    2    -it-in  +\n        ki-:  2nd person ∈{S, O}              2    1    -in\n\n                                          3    1|2   -ik(o|w)\n\n          1st person pl ∈{S, O} :     -(n)ān\n\n    + 2nd person pl ∈{S, O}        + 3rd person pl ∈{S, O} :  -(w)ak\n                                                       :  -(ā)wāw\n        ∧1st person pl̸ ∈{S, O}\n\n                                                  {S, O}\n                 S  O\n                ē-  as …                             {2nd person", "item_type": "full_problem", "prompt_chars": 2052, "record_id": "iol-2023-individual-p4", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-sol.en.pdf", "source_problem_id": "2023-individual-4", "target_subparts": ["full"], "title": "Plains Cree", "year": 2023}
+{"author": "João Henrique Oliveira Fontes", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Dâw (Naduhup)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 6, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 24, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "a.17", "a.18", "a.19", "b.1", "b.2", "c.1", "c.2", "c.3"], "canonical_answer_chars": 1686, "canonical_excerpt": "Problem 4.\n\n  (a)    1. çʉm ’aa’ (foot — receptacle): C. flip-flops\n          2. dâw çʉʉm (person — foot): I. foot\n          3. dâw nõr (person — mouth): B. mouth\n          4. dâw nõr keet (person — mouth — leaf/tongue): H. tongue\n          5. dâw tôog (person — daughter): F. daughter\n          6. dâw sôb pis piis ((person — hand — small) — small): D. little finger\n          7. dâw tôoj (person — nose): J. nose\n          8. dôo’ piis (to make — small): E. to decrease (something)\n          9. sôb dak (hand — to put): A. ring\n        10. suk ’aa’ (flour — receptacle): G. can of flour\n        11. be keet (tree — leaf/tongue): M. leaf\n        12. be tʉm (tree — seed/eye): Q. seed\n        13. yak", "item_type": "full_problem", "prompt_chars": 3066, "record_id": "iol-2024-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 19, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "a.17", "a.18", "a.19"], "canonical_answer_chars": 1134, "canonical_excerpt": "(a)    1. çʉm ’aa’ (foot — receptacle): C. flip-flops\n          2. dâw çʉʉm (person — foot): I. foot\n          3. dâw nõr (person — mouth): B. mouth\n          4. dâw nõr keet (person — mouth — leaf/tongue): H. tongue\n          5. dâw tôog (person — daughter): F. daughter\n          6. dâw sôb pis piis ((person — hand — small) — small): D. little finger\n          7. dâw tôoj (person — nose): J. nose\n          8. dôo’ piis (to make — small): E. to decrease (something)\n          9. sôb dak (hand — to put): A. ring\n        10. suk ’aa’ (flour — receptacle): G. can of flour\n        11. be keet (tree — leaf/tongue): M. leaf\n        12. be tʉm (tree — seed/eye): Q. seed\n        13. yak yaa’ (manioc ", "item_type": "subproblem", "prompt_chars": 1651, "record_id": "iol-2024-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 176, "canonical_excerpt": "(b)   21. dâw sôb piis (person — (hand — small)): finger\n        22. dâw sôob pis ((person — hand) — small): little hand\n        23. dâw çʉm piis (person — (foot — small)): toe", "item_type": "subproblem", "prompt_chars": 1803, "record_id": "iol-2024-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 353, "canonical_excerpt": "(c)   24. brook, stream — nâx piis (water — small)\n        25.  little tapir — taax pis (tapir) — small\n        26. eye — dâw tʉm (person — seed/eye)\n        27. granddaughter — dâw tôg tôog (person — (daughter — daughter))\n\nTwenty-first International Linguistics Olympiad (2024)                                            5\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 4311, "record_id": "iol-2024-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 176, "canonical_excerpt": "(b)   21. dâw sôb piis (person — (hand — small)): finger\n        22. dâw sôob pis ((person — hand) — small): little hand\n        23. dâw çʉm piis (person — (foot — small)): toe", "item_type": "partial_variant", "prompt_chars": 2990, "record_id": "iol-2024-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 353, "canonical_excerpt": "(c)   24. brook, stream — nâx piis (water — small)\n        25.  little tapir — taax pis (tapir) — small\n        26. eye — dâw tʉm (person — seed/eye)\n        27. granddaughter — dâw tôg tôog (person — (daughter — daughter))\n\nTwenty-first International Linguistics Olympiad (2024)                                            5\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 5702, "record_id": "iol-2024-individual-p4-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-sol.en.pdf", "source_problem_id": "2024-individual-4", "target_subparts": ["a", "b", "c", "full"], "title": "Dâw", "year": 2024}
+{"author": "Samuel Ahmed", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Kewa (Trans–New Guinea)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 13, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10", "b.11", "b.12"], "canonical_answer_chars": 858, "canonical_excerpt": "Problem 4.\n\n  (a)\n       15  D.  16  N.  17   T.   18   L.   19  W.\n       20  O.  21   Y.   22  H.  23     I.   24   G.\n       25  A.  26  Q.  27   C.   28   V.   29    P.\n       30   K.  31   U.  32    F.   33  M.  34   E.\n       35     J.   36   X.  37   S.   38   R.   39   B.\n\n  (b)   40. repena — tree; fire\n        41. agaa — mouth; speech\n        42.  iri — hair; grass\n        43. yagaa — jaw\n        44. nida dia — it is not me\n        45. yaa-iri — bird feathers\n        46. nogo-naaki — children\n\n   (c)   47. white man — bali\n        48. bone — uni\n        49. seed of a tree — repena-ini\n        50. hole — boke\n        51. very big — ora adaa\n        52. pandanus — aga\n        53. old", "item_type": "full_problem", "prompt_chars": 3692, "record_id": "iol-2025-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 266, "canonical_excerpt": "(a)\n       15  D.  16  N.  17   T.   18   L.   19  W.\n       20  O.  21   Y.   22  H.  23     I.   24   G.\n       25  A.  26  Q.  27   C.   28   V.   29    P.\n       30   K.  31   U.  32    F.   33  M.  34   E.\n       35     J.   36   X.  37   S.   38   R.   39   B.", "item_type": "subproblem", "prompt_chars": 3181, "record_id": "iol-2025-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 12, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10", "b.11", "b.12"], "canonical_answer_chars": 574, "canonical_excerpt": "(b)   40. repena — tree; fire\n        41. agaa — mouth; speech\n        42.  iri — hair; grass\n        43. yagaa — jaw\n        44. nida dia — it is not me\n        45. yaa-iri — bird feathers\n        46. nogo-naaki — children\n\n   (c)   47. white man — bali\n        48. bone — uni\n        49. seed of a tree — repena-ini\n        50. hole — boke\n        51. very big — ora adaa\n        52. pandanus — aga\n        53. old woman’s eye — pamoagaena ini\n\nTwenty-second International Linguistics Olympiad (2025)                                          5\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 4279, "record_id": "iol-2025-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 12, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10", "b.11", "b.12"], "canonical_answer_chars": 574, "canonical_excerpt": "(b)   40. repena — tree; fire\n        41. agaa — mouth; speech\n        42.  iri — hair; grass\n        43. yagaa — jaw\n        44. nida dia — it is not me\n        45. yaa-iri — bird feathers\n        46. nogo-naaki — children\n\n   (c)   47. white man — bali\n        48. bone — uni\n        49. seed of a tree — repena-ini\n        50. hole — boke\n        51. very big — ora adaa\n        52. pandanus — aga\n        53. old woman’s eye — pamoagaena ini\n\nTwenty-second International Linguistics Olympiad (2025)                                          5\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 4598, "record_id": "iol-2025-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-sol.en.pdf", "source_problem_id": "2025-individual-4", "target_subparts": ["a", "b", "full"], "title": "Kewa", "year": 2025}

benchmark/IOL/ioling_hf/reports/manual_verification/batch_03.jsonl

@@ -0,0 +1,22 @@
+{"author": "Alexandre Arkhipov", "curation_statuses": {"auto_extracted_needs_manual_review": 6}, "languages": ["Basque (Isolate)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_2.1", "assignment_3.1"], "canonical_answer_chars": 2360, "canonical_excerpt": "Solution of Problem 3\nThere are two types of English expressions in the problem: some (I) consist of a date, a month\nand a day of the week, others (II) name the number of the day of the week within the month\ninstead of the date. The word order in the Basque expressions of type (I) is (month) (date),\n(day of the week), whilst in type (II) it is (month) (number of the day) (day of the week). The\nlast word ends in -a, whereas the preceding words have no final -a (except for the word hogeita,\nwhich means ‘20’ in compound numerals). The element -garren forms ordinal numbers. The word\nastea is not a name of a day of the week (six of those we have seen in examples 1-10, the seventh\noccurs in Assign", "item_type": "full_problem", "prompt_chars": 1793, "record_id": "iol-2003-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 619, "canonical_excerpt": "Assignment 1. —_urtarrilaren hogeita hirugarrena, larunbata the 23rd of January, Saturday\nabenduaren azken astea the last week of December\notsailaren lehenengo osteguna the first Thursday of February\nekainaren bederatzigarrena, igandea the ninth of June, Sunday\nabenduaren lehena, asteazkena the first of December, Wednesday\nirailaren azken asteazkena the last Wednesday of September\nazaroaren hirugarren ostirala the third Friday of November\nurriaren azken larunbata the last Saturday of October\nirailaren lehena, astelehena the first of September, Monday\nurtarrilaren bigarrena, ostirala the second of January, Friday", "item_type": "subproblem", "prompt_chars": 1062, "record_id": "iol-2003-individual-p3-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 273, "canonical_excerpt": "Assignment 2. _ the first Monday of December abenduaren lehenengo astelehena\nthe 29th of November, Saturday azaroaren hogeita bederatzigarrena, larunbata\nthe second week of January urtarrilaren bigarren astea\nthe third of February, Monday otsailaren hirugarrena, astelehena", "item_type": "subproblem", "prompt_chars": 1224, "record_id": "iol-2003-individual-p3-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 652, "canonical_excerpt": "Assignment 3. Astelehena ‘Monday’, asteazkena ‘Wednesday’; asteartea, the only day of the\n\nweek not found in in Assignment 1, is ‘Tuesday’. All three names are formed from the word aste\n\n‘week’. Astelehena means literally ‘first (day) of the week’, asteazkena ‘last (day) of the week’.\n\nTuesday’s Basque name can be translated more or less as ‘day in the middle of the week’.\n\nNo one knows for sure why Basque calls Wednesday ‘last day of the week’. In Basque dialects\nother variants of the names of the days of the week are also found, including loans from Romance\nlanguages.\n\n1st IOL: Borovetz ’03. Solutions to the Problems of the Individual Contest", "item_type": "subproblem", "prompt_chars": 2298, "record_id": "iol-2003-individual-p3-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 273, "canonical_excerpt": "Assignment 2. _ the first Monday of December abenduaren lehenengo astelehena\nthe 29th of November, Saturday azaroaren hogeita bederatzigarrena, larunbata\nthe second week of January urtarrilaren bigarren astea\nthe third of February, Monday otsailaren hirugarrena, astelehena", "item_type": "partial_variant", "prompt_chars": 1907, "record_id": "iol-2003-individual-p3-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 652, "canonical_excerpt": "Assignment 3. Astelehena ‘Monday’, asteazkena ‘Wednesday’; asteartea, the only day of the\n\nweek not found in in Assignment 1, is ‘Tuesday’. All three names are formed from the word aste\n\n‘week’. Astelehena means literally ‘first (day) of the week’, asteazkena ‘last (day) of the week’.\n\nTuesday’s Basque name can be translated more or less as ‘day in the middle of the week’.\n\nNo one knows for sure why Basque calls Wednesday ‘last day of the week’. In Basque dialects\nother variants of the names of the days of the week are also found, including loans from Romance\nlanguages.\n\n1st IOL: Borovetz ’03. Solutions to the Problems of the Individual Contest", "item_type": "partial_variant", "prompt_chars": 3293, "record_id": "iol-2003-individual-p3-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-sol.en.pdf", "source_problem_id": "2003-individual-3", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Basque Dates", "year": 2003}
+{"author": "Peter Zubkov", "curation_statuses": {"auto_extracted_needs_manual_review": 4}, "languages": ["Kayapo (Ge)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 1, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["assignment_1.1", "assignment_2.1"], "canonical_answer_chars": 1213, "canonical_excerpt": "Problem No.1\nThe direct object is expressed by a verb prefix (i - for first person, a- for the second\nperson). There are different rules of expressing the subject in affirmative and negative\nsentences. In affirmative sentences, the subject is expressed by a separate pronoun (ba for the\nfirst person, ga for the second person). In negative sentences, the subject is expressed by the\nsame prefixes as the direct object, which are connected to the verb (if it is intransitive) or to\nthe je particle, which is positioned before the verb (if it is transitive and the prefix slot is already\ninstantiated by the object prefix). In the negative form, the verb has an additional suffix\nconsisting of r + the ", "item_type": "full_problem", "prompt_chars": 1155, "record_id": "iol-2004-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 137, "canonical_excerpt": "Assignment 1.\nAje ikuru két You are not devouring me\nBa mé aku We are devouring you or\nI am devouring you guys\nTréré két Tam not swimming", "item_type": "subproblem", "prompt_chars": 1167, "record_id": "iol-2004-individual-p1-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 165, "canonical_excerpt": "Assignment 2.\nYou guys are not devouring us Mé aje mé ikuru két\nWe are not decorating you guys Mé ije mé anhéré két\nWe are dancing Ba mé to\nI am devouring you Ba aku", "item_type": "subproblem", "prompt_chars": 1233, "record_id": "iol-2004-individual-p1-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 165, "canonical_excerpt": "Assignment 2.\nYou guys are not devouring us Mé aje mé ikuru két\nWe are not decorating you guys Mé ije mé anhéré két\nWe are dancing Ba mé to\nI am devouring you Ba aku", "item_type": "partial_variant", "prompt_chars": 1434, "record_id": "iol-2004-individual-p1-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-sol.en.pdf", "source_problem_id": "2004-individual-1", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Kayapo", "year": 2004}
+{"author": "Ksenia Guiliarova", "curation_statuses": {"auto_extracted_needs_manual_review": 6}, "languages": ["Lango (Nilo-Saharan)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 2, "problem_page_end": 2, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_2.1", "assignment_3.1"], "canonical_answer_chars": 985, "canonical_excerpt": "Problem 2\n\nWe can see from the statement of the problem that some things named by one English\nword take a two-word phrase to say in Lango. Let us try to represent the English nouns as\nphrases, too, or better, as combinations of meanings. Thus the meaning ‘house’ is contained in\nthe concepts roof, floor and restaurant, ‘top’ or ‘head’ in roof and hat, ‘bottom’ in floor and sole;\nfurthermore hat contains the meaning of garment, and eyeball, perhaps, of grain.\n\nWe also determine the order of the words in the Lango phrases: possessed+possessor (hat\n= gin wic ‘garment of the head’ but roof = wic ot ‘head of the house’).\n\nAssignment 1. dye dt — floor (bottom of house’), dye tyen — sole of foot (bo", "item_type": "full_problem", "prompt_chars": 734, "record_id": "iol-2005-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 260, "canonical_excerpt": "Assignment 1. dye dt — floor (bottom of house’), dye tyen — sole of foot (bottom\nof foot’), gin — garment, gin wic — hat (garment of head’), nig — grain, nig wan -\neyeball (grain of eye’), St cém — restaurant (house of eating’), wic ot — roof (head of\nhouse’).", "item_type": "subproblem", "prompt_chars": 534, "record_id": "iol-2005-individual-p2-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 41, "canonical_excerpt": "Assignment 2. cem-— eating, dye — bottom.", "item_type": "subproblem", "prompt_chars": 500, "record_id": "iol-2005-individual-p2-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 56, "canonical_excerpt": "Assignment 3. window — way ot (/it. “eye of the house’).", "item_type": "subproblem", "prompt_chars": 938, "record_id": "iol-2005-individual-p2-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 41, "canonical_excerpt": "Assignment 2. cem-— eating, dye — bottom.", "item_type": "partial_variant", "prompt_chars": 824, "record_id": "iol-2005-individual-p2-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 56, "canonical_excerpt": "Assignment 3. window — way ot (/it. “eye of the house’).", "item_type": "partial_variant", "prompt_chars": 1342, "record_id": "iol-2005-individual-p2-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-sol.en.pdf", "source_problem_id": "2005-individual-2", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Lango", "year": 2005}
+{"author": "Sergey Dmitrenko", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Khmer (Austro-Asiatic)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["assignment_1.1", "assignment_2.1"], "canonical_answer_chars": 2261, "canonical_excerpt": "Problem #3\n      The ‘hook’ that each word has in the middle or at the right edge (if the word is written by\ntwo characters) marks a vowel.  The  little circle on top of  it indicates brevity (accordingly, the\nabsence of a circle indicates length). From the words ‘passion’ and ‘turn around’, which contain a\ncharacter not found anywhere else, we see that the direction of the script is left to right and the ‘rare\ncharacter’ stands for k. The other characters in these words are m (the final consonant in ‘passion’)\nand ŋ (in ‘turn around’). No word but ‘have’ begins with m, so we derive the character for n. And\nso on; each consonant has its letter, except for final m after a short vowel; there i", "item_type": "full_problem", "prompt_chars": 1539, "record_id": "iol-2006-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 1078, "canonical_excerpt": "Assignment 1.\nka:m     ‘passion’                                a:c      ‘passion’\nam       ‘suck (a breast)’                cam    ‘suck (a breast)’\npiъm     ‘delta’                      poьm   ‘delta’\nnoьm    ‘lead, guide’                     aŋ       ‘lead, guide’\nŋoьm    ‘inarticulately’                    peьŋ    ‘inarticulately’\nŋiъk     ‘turn around’                   miъn   ‘turn around’\ncaŋ       ‘trim, rough-hew’\n     To write the words in the Roman script, we need to determine the rules that govern the\nchoice of the vowel: since it is always written by the same character, its quality will have to depend\non the consonants (and, possibly, its quantity). Here is a summary of what we ", "item_type": "subproblem", "prompt_chars": 1048, "record_id": "iol-2006-individual-p3-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 402, "canonical_excerpt": "Assignment 2.\n        kam    ‘spoke (in a wheel)’       ca:m      ‘Cham’\n           ka:k    ‘crow’                 niъ:m     ‘name’\n          ŋeьŋ    ‘bright red’          moьm     ‘robust’\n\n                         Problem: Sergei Dmitrenko Solution: Svetlana Burlak, ed. by Ivan Derzhanski\n\nFourth International Linguistic Olympiad. Solutions of the Problems of the Individual Contest               4", "item_type": "subproblem", "prompt_chars": 2134, "record_id": "iol-2006-individual-p3-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 402, "canonical_excerpt": "Assignment 2.\n        kam    ‘spoke (in a wheel)’       ca:m      ‘Cham’\n           ka:k    ‘crow’                 niъ:m     ‘name’\n          ŋeьŋ    ‘bright red’          moьm     ‘robust’\n\n                         Problem: Sergei Dmitrenko Solution: Svetlana Burlak, ed. by Ivan Derzhanski\n\nFourth International Linguistic Olympiad. Solutions of the Problems of the Individual Contest               4", "item_type": "partial_variant", "prompt_chars": 3276, "record_id": "iol-2006-individual-p3-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-sol.en.pdf", "source_problem_id": "2006-individual-3", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Khmer", "year": 2006}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Ndom (Trans-New Guinea)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1749, "canonical_excerpt": "Problem !4\nPrecisely half of the names of squares include the word nif.  It is reasonable to assume that these\nare the greater numbers – from 36 to 100, – with nif being the least among them. The frequently\noccurring word abo probably denotes addition. We notice the pair of squares nif abo tondor abo\nmer abo thonith and nif thef abo tondor abo mer abo thonith; this supports our guess that 36 is a\nkey notion in the formation of the number names, since in that case we would expect the names\nof 64=36+28 and 100=64+36=36%2+28 to be very similar.\n      So the base of the number system is 6. There is one peculiarity: not only 36 but also 18\nhas a special name; accordingly, instead of multiplying 6", "item_type": "full_problem", "prompt_chars": 1459, "record_id": "iol-2007-individual-p4", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-sol.en.pdf", "source_problem_id": "2007-individual-4", "target_subparts": ["full"], "title": "Ndom", "year": 2007}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Copainalá Zoque (Mixe-Zoquean)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 4, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4"], "canonical_answer_chars": 1150, "canonical_excerpt": "Problem #4.  The noun suffixes seen in this problem are:\n\n   1. -k2sm2 ‘above, overhead’, -kotoya ‘for’, -pit ‘with’;\n\n   2. -ˇseh ‘like, as if’;\n\n   3. -taPm plural;\n\n   4. -tih ‘only (just, right)’.\n\nAfter a nasal consonant (m, n, N) the stops p, t, k become voiced (b, d, g respectively).  If k\ncomes after y, the two sounds exchange places.\n   The possessive pronouns are P2s ‘my’ and mis ‘your’;  if the noun begins with a stop, this\nconsonant becomes voiced and the corresponding nasal appears before it.\n\n (a)  caky2sm2tih            right above the vine\n      k2m2Nˇseh                like a shadow\n      P2s mok           my corn\n     mis nd2ctaPm        your teeth\n      pahsunbit            ", "item_type": "full_problem", "prompt_chars": 2054, "record_id": "iol-2008-individual-p4", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-sol.en.pdf", "source_problem_id": "2008-individual-4", "target_subparts": ["full"], "title": "Copainalá Zoque", "year": 2008}
+{"author": "Alexander Piperski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Old Indic (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 794, "canonical_excerpt": "Problem #4.\n\nIf the stop consonant in the root   and the vowel in the suffix is a,  and the vowel in the suffix is i,\n                               is voiced   the stress is on the suffix.        the stress is on the root.\n                             is voiceless   the stress is on the root.          the stress is on the suffix.\n\n   (a) This rule holds if the root contains precisely one stop consonant. If there are two (bh¯ag-a-,\n       pad-a-, pat-i-), or if there are none (us-ri-), the place of the stress can’t be determined.\n\n   (b) mr.dh-rá-, phé-na-, stu-tí-, tan-tí-, bh¯ar-á-, d¯u-tá-, sváp-na-, bh´¯u-mi-, ghar-má-,\n       abh-rá-, ghan-á-, gh´r.s.-vi-.\n\nSeventh International Olympia", "item_type": "full_problem", "prompt_chars": 3790, "record_id": "iol-2009-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 198, "canonical_excerpt": "(a) This rule holds if the root contains precisely one stop consonant. If there are two (bh¯ag-a-,\n       pad-a-, pat-i-), or if there are none (us-ri-), the place of the stress can’t be determined.", "item_type": "subproblem", "prompt_chars": 1755, "record_id": "iol-2009-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 254, "canonical_excerpt": "(b) mr.dh-rá-, phé-na-, stu-tí-, tan-tí-, bh¯ar-á-, d¯u-tá-, sváp-na-, bh´¯u-mi-, ghar-má-,\n       abh-rá-, ghan-á-, gh´r.s.-vi-.\n\nSeventh International Olympiad in Linguistics (2009).                                        3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 5905, "record_id": "iol-2009-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 254, "canonical_excerpt": "(b) mr.dh-rá-, phé-na-, stu-tí-, tan-tí-, bh¯ar-á-, d¯u-tá-, sváp-na-, bh´¯u-mi-, ghar-má-,\n       abh-rá-, ghan-á-, gh´r.s.-vi-.\n\nSeventh International Olympiad in Linguistics (2009).                                        3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 6156, "record_id": "iol-2009-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-sol.en-gb.pdf", "source_problem_id": "2009-individual-4", "target_subparts": ["a", "b", "full"], "title": "Old Indic", "year": 2009}
+{"author": "Alexander Berdichevski", "curation_statuses": {"auto_extracted_text_high_confidence": 2}, "languages": [], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-prob.en.pdf", "record_count": 2, "record_types": {"full_problem": 1, "subproblem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 2691, "canonical_excerpt": "Problem #4.  The four polypeptides in the example consist of 24, 10, 3 and 25 amino acids,\nand the mRNA sequence contains 195 = ((24 + 10 + 3 + 25) + 3) × 3 nucleotides. It appearsprobable that three nucleotides (a triplet) denote one amino acid or are a separator between\npolypeptides (in reality a signal to terminate synthesis). However, since there are 43 = 64\npossible triplets (all but two of which are present in the example) and only 20 different amino\nacids, some triplets have the same meaning.\n              . . . U. . .               . . . C. . .               . . . A. . .                   . . . G. . .\n        UUU →Phe     UCU →Ser     UAU →Tyr       UGU →Cys\n        UUC           UCC ", "item_type": "full_problem", "prompt_chars": 2077, "record_id": "iol-2010-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 310, "canonical_excerpt": "(c) A root XY is strong if XYA, XYG, XYC and XYU                            encode  the same amino acid (UC, CC, CG,\n      GC). A root is weak if this is not the case (UU, CA, AG, GA).\n\nEighth International Olympiad in Linguistics (2010).                                      4\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2350, "record_id": "iol-2010-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-sol.en.pdf", "source_problem_id": "2010-individual-4", "target_subparts": ["c", "full"], "title": "mRNA Sequencing", "year": 2010}
+{"author": "Liudmila Fedorova", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Nahuatl (Uto-Aztecan)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 1607, "canonical_excerpt": "Problem #4.  In compound nouns the left-hand part modifies the right-hand one. A noun gets the\nending -tl/li unless it has one of the suffixes -capil (dimin.), -huah ‘one who has  . . . ’, -tlah/lah\n‘place of many . . . ’, or -tzintli ‘revered . . . ’ (-li and -lah after l, otherwise -tl and -tlah).\n\n (a)   a-cal-huah          canoe owner (a-cal-li canoe, “water house”)\n        a-chil-li             water pepper\n        a-tl                  water\n       cal-lah                  village\n       cal-huah             master of house\n        chil-a-tl                  chili water\n        chil-li                     chili\n        col-li                 grandfather/ancestor\n        col-tzintli       ", "item_type": "full_problem", "prompt_chars": 1383, "record_id": "iol-2011-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 849, "canonical_excerpt": "(a)   a-cal-huah          canoe owner (a-cal-li canoe, “water house”)\n        a-chil-li             water pepper\n        a-tl                  water\n       cal-lah                  village\n       cal-huah             master of house\n        chil-a-tl                  chili water\n        chil-li                     chili\n        col-li                 grandfather/ancestor\n        col-tzintli             revered grandfather/ancestor\n      cone-huah           mother, “one who has child(ren)”\n       cone-huah-capil    mum(my)\n       cone-tl                 child\n       oquich-cone-tl        boy, male child\n      oquich-huah           wife, “one who has a husband”\n       oquich-totol-tzintli   re", "item_type": "subproblem", "prompt_chars": 843, "record_id": "iol-2011-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 140, "canonical_excerpt": "(b) house: calli                        stone: tetl                       possessor of water: ahuah\n      revered man/husband: oquichtzintli", "item_type": "subproblem", "prompt_chars": 923, "record_id": "iol-2011-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 312, "canonical_excerpt": "(c) cacahua-tl: cocoa                                              cacahua-te-tl: cocoa bean\n     cacahua-a-tl: cocoa drink                           cacahua-huah: possessor of cocoa\n\nNinth International Olympiad in Linguistics (2011).                                               3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1833, "record_id": "iol-2011-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 140, "canonical_excerpt": "(b) house: calli                        stone: tetl                       possessor of water: ahuah\n      revered man/husband: oquichtzintli", "item_type": "partial_variant", "prompt_chars": 1825, "record_id": "iol-2011-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 312, "canonical_excerpt": "(c) cacahua-tl: cocoa                                              cacahua-te-tl: cocoa bean\n     cacahua-a-tl: cocoa drink                           cacahua-huah: possessor of cocoa\n\nNinth International Olympiad in Linguistics (2011).                                               3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2903, "record_id": "iol-2011-individual-p4-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-sol.en-gb.pdf", "source_problem_id": "2011-individual-4", "target_subparts": ["a", "b", "c", "full"], "title": "Nahuatl", "year": 2011}
+{"author": "Maria Konoshenko", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Teop (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 6, "answer_unit_ids": ["a.1", "a.2", "b.1", "b.2", "b.3", "c.1"], "canonical_answer_chars": 2156, "canonical_excerpt": "Problem #4.  The sentences have the following structure: S paa V O [O′]  (S: subject, V:\nverb, O: object, O′: another object).\n      give         call      strike, kill\n O   to whom  whom  whom\n O′  what     what   with what\n   Every noun is preceded by an article, which is a  if this is the first third person in the\nsentence and bona otherwise. The form of the 3rd person sg pronoun e or bona is chosen in\nthe same way.\n   Personal pronouns:\n         1st person sg   1st person pl  2nd person sg  2nd person pl  3rd person sg  3rd person pl\n S     enaa       enam       ean        eam          eove          eori\n O, O ′  anaa                   vuan       ameam        e, bona\n\n (a)  13. Eam paa an", "item_type": "full_problem", "prompt_chars": 2669, "record_id": "iol-2012-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 208, "canonical_excerpt": "(a)  13. Eam paa ani a overe. — You(pl) ate the coconut.\n       14. Ean paa tasu a oraoraa bona kae. — You (sg.) struck the sorcerer with the\n          bag.\n       15. Eove paa tara ameam. — He saw you (pl.).", "item_type": "subproblem", "prompt_chars": 1467, "record_id": "iol-2012-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 909, "canonical_excerpt": "(b)  16. We gave the food to you (sg.). — Enam paa hee vuan a taba’ani.\n       17. He called me a child. — Eove paa dao anaa bona beiko.\n       18. I killed him with it. — Enaa paa asun e bona.\n       19. The sorcerer gave the fish to the boy. — A oraoraa paa hee bona visoasi bona\n         iana.\nWhat was mentioned in the context is moved to the\nfirst place and takes the article a. If thereby a pro-   S paa V O [O′]    S paa V O [O′]                               →\nnominal subject moves after the verb,  it loses its   S paa V O [O\n                                                                                                                                                     ′                ", "item_type": "subproblem", "prompt_chars": 2967, "record_id": "iol-2012-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 353, "canonical_excerpt": "(c)  20. (Why was the sorcerer offended?) They called the sorcerer a woman.\n     — A oraoraa paa dao ori bona moon.\n       21. (Why is this axe wet?) The boy killed the fish with the axe.\n     — A toraara paa asun a visoasi bona iana.\n\nTenth International Olympiad in Linguistics (2012).                                      4\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2177, "record_id": "iol-2012-individual-p4-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 909, "canonical_excerpt": "(b)  16. We gave the food to you (sg.). — Enam paa hee vuan a taba’ani.\n       17. He called me a child. — Eove paa dao anaa bona beiko.\n       18. I killed him with it. — Enaa paa asun e bona.\n       19. The sorcerer gave the fish to the boy. — A oraoraa paa hee bona visoasi bona\n         iana.\nWhat was mentioned in the context is moved to the\nfirst place and takes the article a. If thereby a pro-   S paa V O [O′]    S paa V O [O′]                               →\nnominal subject moves after the verb,  it loses its   S paa V O [O\n                                                                                                                                                     ′                ", "item_type": "partial_variant", "prompt_chars": 3228, "record_id": "iol-2012-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 353, "canonical_excerpt": "(c)  20. (Why was the sorcerer offended?) They called the sorcerer a woman.\n     — A oraoraa paa dao ori bona moon.\n       21. (Why is this axe wet?) The boy killed the fish with the axe.\n     — A toraara paa asun a visoasi bona iana.\n\nTenth International Olympiad in Linguistics (2012).                                      4\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3375, "record_id": "iol-2012-individual-p4-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-sol.en.pdf", "source_problem_id": "2012-individual-4", "target_subparts": ["a", "b", "c", "full"], "title": "Teop", "year": 2012}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Muna (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 5, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 5, "answer_unit_ids": ["a.1", "a.2", "a.3", "b.1", "b.2"], "canonical_answer_chars": 1644, "canonical_excerpt": "Problem #4.  The word order is  subject  verb  object   .  If the subject is neither pos-\n\nsessed nor a proper name, it is preceded by the article a.                                        \n                                             possessor:\n                               -ku    1st person sg\n   Noun:  root + [-hi plural] +                                                                                                  -no   3rd person sg   .\n                            -ndo  3rd person pl\n   Proper name: a + the first root syllable + a + root  .\n    Possession:\n         possessed -no  possessor (singular)   ,  possessed -ndo  possessor (plural)   .\n      {             }  {     ", "item_type": "full_problem", "prompt_chars": 1755, "record_id": "iol-2013-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["a.1", "a.2", "a.3"], "canonical_answer_chars": 318, "canonical_excerpt": "(a)\n             2. a dhinihi dasumuli we murindo robhinehi.\n           The demons will return to the women’s pupil.\n\n             3. The Ant will climb the pupil’s stone.\n             a-la-a-laga na-moni we kontu-no muri.\n             4. The ants are going to the Demon.\n           a laga-hi do-kala we a-dhi-a-dhini.", "item_type": "subproblem", "prompt_chars": 1382, "record_id": "iol-2013-individual-p4-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 209, "canonical_excerpt": "(b)\n             5. My women’s monkeys will cut my bananas.\n            ndoke-hi-ndo robhine-hi-ku da-dumodo kalei-hi-ku.\n             6. The monkey’s mountains are far.\n            molo-hi-no ndoke no-kodoho.", "item_type": "subproblem", "prompt_chars": 2208, "record_id": "iol-2013-individual-p4-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 209, "canonical_excerpt": "(b)\n             5. My women’s monkeys will cut my bananas.\n            ndoke-hi-ndo robhine-hi-ku da-dumodo kalei-hi-ku.\n             6. The monkey’s mountains are far.\n            molo-hi-no ndoke no-kodoho.", "item_type": "partial_variant", "prompt_chars": 2579, "record_id": "iol-2013-individual-p4-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-sol.en-gb.pdf", "source_problem_id": "2013-individual-4", "target_subparts": ["a", "b", "full"], "title": "Muna", "year": 2013}
+{"author": "Boris Iomdin", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Gbaya (Niger-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 5, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 1881, "canonical_excerpt": "Problem #5.  body part −w´ı `human' = spatial preposition:\n                                              s`e`e   (liver)   at the centre\n                                    y´ık   (face)   at the surface\n                                    z`u   (head)  above, on\n   Answers:\n\n (a)   P´a´a                 (to put)                        to put\n         P´a´a n´ang n´u k`o   (to put + foot + edge + hole)  to be dying\n         P´a´a s`e`e            (to put + liver)                 to envy\n       b´um´a y´ık          (hair + eyes)                   eyelash\n       b´um´a z`u y´ık      (hair + above, on + eyes)      eyebrow\n      d´ang g`ok         (bad + snake)                 poisonous snake\n  ", "item_type": "full_problem", "prompt_chars": 1774, "record_id": "iol-2014-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 1200, "canonical_excerpt": "(a)   P´a´a                 (to put)                        to put\n         P´a´a n´ang n´u k`o   (to put + foot + edge + hole)  to be dying\n         P´a´a s`e`e            (to put + liver)                 to envy\n       b´um´a y´ık          (hair + eyes)                   eyelash\n       b´um´a z`u y´ık      (hair + above, on + eyes)      eyebrow\n      d´ang g`ok         (bad + snake)                 poisonous snake\n        d´ı f`o             (good +  eld)                good  eld\n        d´ı s`e`e            (good + liver)                 happiness\n        k`o y´ık             (hole + eyes)                   eye socket\n        k`o z`ok             (hole + nose)                     nostril\n", "item_type": "subproblem", "prompt_chars": 872, "record_id": "iol-2014-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 210, "canonical_excerpt": "(b)  b´um´a z`u   (hair + head)        hair (on the head)\n        k`o          (hole)               hole\n         l´eb´e g`ok   (tongue + snake)    snake's tongue\n         l´eb´e w´ı    (tongue + human)  tongue", "item_type": "subproblem", "prompt_chars": 926, "record_id": "iol-2014-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 187, "canonical_excerpt": "(c)   s`e`e          (liver)             at the centre\n        z`u w´ı     (head + human)  head\n      d´ang s`e`e  (bad + liver)      displeasure\n       z`ok w´ı     (nose + human)   nose", "item_type": "subproblem", "prompt_chars": 2612, "record_id": "iol-2014-individual-p5-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 210, "canonical_excerpt": "(b)  b´um´a z`u   (hair + head)        hair (on the head)\n        k`o          (hole)               hole\n         l´eb´e g`ok   (tongue + snake)    snake's tongue\n         l´eb´e w´ı    (tongue + human)  tongue", "item_type": "partial_variant", "prompt_chars": 2179, "record_id": "iol-2014-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 187, "canonical_excerpt": "(c)   s`e`e          (liver)             at the centre\n        z`u w´ı     (head + human)  head\n      d´ang s`e`e  (bad + liver)      displeasure\n       z`ok w´ı     (nose + human)   nose", "item_type": "partial_variant", "prompt_chars": 4103, "record_id": "iol-2014-individual-p5-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-sol.en-gb.pdf", "source_problem_id": "2014-individual-5", "target_subparts": ["a", "b", "c", "full"], "title": "Gbaya", "year": 2014}
+{"author": "Alexander Piperski", "curation_statuses": {"auto_extracted_text_high_confidence": 2}, "languages": ["Somali (Afroasiatic)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 5, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-prob.en-gb.pdf", "record_count": 2, "record_types": {"full_problem": 1, "subproblem": 1}, "records": [{"answer_unit_count": 9, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9"], "canonical_answer_chars": 832, "canonical_excerpt": "Problem 5.\n\n          ◦◦     ◦◦    ◦= V (a, e, i, o, u) (a) (◦)◦◦           ◦◦ ◦◦◦◦◦◦◦ ◦◦,     ◦◦= VV (aa, ee, ii, oo, uu)\n\n (b)  36.  war     is—maa—ciil       daa-  rood  ×\n                                  √\n       37.         dir mi-  yaad   wa-  daag-  taan\n                                  √\n       38.          laba-  daad    ka   duu-  diye\n       39.       ka jan-    na-daad     daa-  hiye  ×\n                                  √\n       40.            adi-   ga i-    yo  deris-  kaa\n                                  √\n       41.          diga-  xaar-    ka  mari-  yoo\n       42.      ciid i-       yo doo-     lo di-  raac  ×\n\n       43.          noo-   ma kee-    neen  darka ×\n       4", "item_type": "full_problem", "prompt_chars": 3377, "record_id": "iol-2015-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 9, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9"], "canonical_answer_chars": 707, "canonical_excerpt": "(b)  36.  war     is—maa—ciil       daa-  rood  ×\n                                  √\n       37.         dir mi-  yaad   wa-  daag-  taan\n                                  √\n       38.          laba-  daad    ka   duu-  diye\n       39.       ka jan-    na-daad     daa-  hiye  ×\n                                  √\n       40.            adi-   ga i-    yo  deris-  kaa\n                                  √\n       41.          diga-  xaar-    ka  mari-  yoo\n       42.      ciid i-       yo doo-     lo di-  raac  ×\n\n       43.          noo-   ma kee-    neen  darka ×\n       44.   ka-    la de-   yaa-    yaa mi-    yaan  ×\n                                  √\n       45.  wu-   xuun    kaa  dan-   qaa", "item_type": "subproblem", "prompt_chars": 4477, "record_id": "iol-2015-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-sol.en-gb.pdf", "source_problem_id": "2015-individual-5", "target_subparts": ["b", "full"], "title": "Somali Masafo", "year": 2015}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Jaqaru (Aymaran)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 6, "problem_page_start": 6, "problem_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["a.1", "a.2", "b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 1057, "canonical_excerpt": "Problem 5.  The word forms in the problem consist of a root and up to three suffixes. Each\nsuffix may cause the elision of the preceding vowel:\n\n                        possessor              case\n                     -p”a                -p”a                -psa\n                       -sa                   -t”a                   -txi\n oral suffixes          always consume the  consume the vowel if it belongs to another\n                      vowel               suffix, but leave it if it belongs to the root\n                     -nha, -ma          -na\n nasal suffixes        consume the vowel if they keep theirs, but\n                         leave it if they give up theirs\n\n   Answers:\n\n (a)    1. utnhan", "item_type": "full_problem", "prompt_chars": 2015, "record_id": "iol-2016-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 123, "canonical_excerpt": "(a)    1. utnhantxi — is it in my house?\n\n         2. yapp”psa — his field too\n\n         3. yapup”psa — through the field too", "item_type": "subproblem", "prompt_chars": 1100, "record_id": "iol-2016-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 247, "canonical_excerpt": "(b)    4. our1+2 cliff— qaqsa\n         5. in your corner — chushumna\n\n         6. in your corner too — chushmanpsa\n\n         7. from your corner too — chushumt”psa\n\n         8. is it from the cave? — aq”it”txi\n\n         9. is it the house? — utatxi", "item_type": "subproblem", "prompt_chars": 3010, "record_id": "iol-2016-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 247, "canonical_excerpt": "(b)    4. our1+2 cliff— qaqsa\n         5. in your corner — chushumna\n\n         6. in your corner too — chushmanpsa\n\n         7. from your corner too — chushumt”psa\n\n         8. is it from the cave? — aq”it”txi\n\n         9. is it the house? — utatxi", "item_type": "partial_variant", "prompt_chars": 3186, "record_id": "iol-2016-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-sol.en.pdf", "source_problem_id": "2016-individual-5", "target_subparts": ["a", "b", "full"], "title": "Jaqaru", "year": 2016}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Madak (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 5, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 13, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10", "b.11", "b.12"], "canonical_answer_chars": 1753, "canonical_excerpt": "Problem 5.\n            \n                                            . . .                                                 . . . . . . .                                                            . . . . . . . . . .                                                                 ..  singular                            }              {                          vut-                                   people                          V-                                plural\n                         ven-                                    otherwise   1. number: l-             \n                        xan-.                                                     . . . .                  ", "item_type": "full_problem", "prompt_chars": 2635, "record_id": "iol-2017-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 44, "canonical_excerpt": "(a) lavatbungmenemen — big group of villages", "item_type": "subproblem", "prompt_chars": 1227, "record_id": "iol-2017-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 12, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10", "b.11", "b.12"], "canonical_answer_chars": 499, "canonical_excerpt": "(b)    1. laradi — man\n\n         2. lavatkonuna — big branch\n\n         3. laxantoonan — many spirits\n\n         4. levengkatli — eyes\n\n         5. loxot — place\n\n         6. lubungkavus ��� group of white men\n\n         7. luvaroos — two big vines\n\n (c)    8. heart — leveve\n\n         9. gardens — leventaamang\n\n       10. many things — laxampanga\n\n       11. grandchildren — luvuttubuno\n\n       12. two parts of day — luxonaleng\n\n       13. branches — lengkonuna\n\n       14. big branches — lempatkonuna", "item_type": "subproblem", "prompt_chars": 4123, "record_id": "iol-2017-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 12, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10", "b.11", "b.12"], "canonical_answer_chars": 499, "canonical_excerpt": "(b)    1. laradi — man\n\n         2. lavatkonuna — big branch\n\n         3. laxantoonan — many spirits\n\n         4. levengkatli — eyes\n\n         5. loxot — place\n\n         6. lubungkavus — group of white men\n\n         7. luvaroos — two big vines\n\n (c)    8. heart — leveve\n\n         9. gardens — leventaamang\n\n       10. many things — laxampanga\n\n       11. grandchildren — luvuttubuno\n\n       12. two parts of day — luxonaleng\n\n       13. branches — lengkonuna\n\n       14. big branches — lempatkonuna", "item_type": "partial_variant", "prompt_chars": 4220, "record_id": "iol-2017-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-sol.en.pdf", "source_problem_id": "2017-individual-5", "target_subparts": ["a", "b", "full"], "title": "Madak", "year": 2017}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Akan (Atlantic-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 5, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 1239, "canonical_excerpt": "Problem 5.\nYɛfrɛ me L: ‘My name is L’                                   ne …: ‘and …’\nYɛfrɛ me _ L: ‘My _’s name is L’                                -nom: plural\n     Terms for a man’s kin:\nɛna ‘mother; mother’s sister’\nagya ‘father; father’s brother’\nnua ‘sibling; father’s brother’s son/daughter; mother’s sister’s son/daughter’\nba ‘son/daughter; fraternal nephew/niece (brother’s son/daughter)’\nsewaa ‘father’s sister’\nsewaaba ‘father’s sister’s son/daughter’\nwɔfa ‘mother’s brother’\nwɔfaba ‘mother’s brother’s son/daughter’\nwɔfaase ‘sororal nephew/niece (sister’s son/daughter)’\nyere ‘wife’\n\n             Ofori        Dubaku      Kunto      Enu          Nsia      [Awotwi]   ✗     ✔   ✗     ✔\n  ", "item_type": "full_problem", "prompt_chars": 2273, "record_id": "iol-2018-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 33, "canonical_excerpt": "(a)  ⃝  △  △  ⃝  △  ⃝      ⃝    △", "item_type": "subproblem", "prompt_chars": 1092, "record_id": "iol-2018-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 386, "canonical_excerpt": "(b)    1. Yɛfrɛ me Yaw. Yɛfrɛ me ɛnanom Dubaku ne Kunto. Yɛfrɛ me agyanom\n        Nsia ne Enu. Yɛfrɛ me nuanom Thema ne Ama ne Kobina. Yɛfrɛ me\n        sewaa Awotwi. Yɛfrɛ me wɔfa Ofori. Yɛfrɛ me wɔfabanom Esi ne Kofi.\n         Yɛfrɛ me sewaaba Berko.\n         2. Yɛfrɛ me Ofori. Yɛfrɛ me banom Kofine Esi. Yɛfrɛ me wɔfaasenom\n       Yaw ne Ama ne Kobina. Yɛfrɛ me nuanom Kunto ne Dubako.", "item_type": "subproblem", "prompt_chars": 3536, "record_id": "iol-2018-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 386, "canonical_excerpt": "(b)    1. Yɛfrɛ me Yaw. Yɛfrɛ me ɛnanom Dubaku ne Kunto. Yɛfrɛ me agyanom\n        Nsia ne Enu. Yɛfrɛ me nuanom Thema ne Ama ne Kobina. Yɛfrɛ me\n        sewaa Awotwi. Yɛfrɛ me wɔfa Ofori. Yɛfrɛ me wɔfabanom Esi ne Kofi.\n         Yɛfrɛ me sewaaba Berko.\n         2. Yɛfrɛ me Ofori. Yɛfrɛ me banom Kofine Esi. Yɛfrɛ me wɔfaasenom\n       Yaw ne Ama ne Kobina. Yɛfrɛ me nuanom Kunto ne Dubako.", "item_type": "partial_variant", "prompt_chars": 3622, "record_id": "iol-2018-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-sol.en.pdf", "source_problem_id": "2018-individual-5", "target_subparts": ["a", "b", "full"], "title": "Akan", "year": 2018}
+{"author": "Samuel Ahmed", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Nooni (Atlantic-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 6, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "a.2", "a.3", "b.1", "c.1", "c.2", "c.3", "c.4"], "canonical_answer_chars": 1652, "canonical_excerpt": "Problem 5.  Rules:\n\n   • word order:\n               (                  {        )\n                                                                  cí  (past)\n      – A (T) nú B   ‘A is B’/‘A was B’, T = tense:\n                                                    ∅   (future)\n\n      – S T V O (lé) L D\n\n          (S = subject, V = predicate, O = object, L = place, D = day)\n\n          ∗‘to see’ = yɛn (O) lé\n           ∗after a consonant: lé > é\n\n              tense           ‘just now’/‘be about to …’  today  < 3 days > 3 days\n   • tense:  past     (V᷅)   nɔ̂ː                                  bɛ̂ː       cí nɔ̂ː         tɔ̀ nɔ̂ː\n              future   (V́)  nú                    nú géː  nú bóy  ", "item_type": "full_problem", "prompt_chars": 3514, "record_id": "iol-2019-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["a.1", "a.2", "a.3"], "canonical_answer_chars": 316, "canonical_excerpt": "(a)  16. bvúsɔ̀w nǔ fɔ̀ shě ntfǔː.          Bvusɔw is the day after tomorrow.\n       17. me nú ŋgéː nyɛ́n kèŋkfǔ lé ɛ̀bɛ̀n.    Today, I will see the yam.\n       18. wvù tɔ̀ nɔ̂ː yɛ᷅n bèsèn ě bvúmbòn.   On Bvumbon, he saw us.\n       19. bèsèn nǔ bóy tɔ́ fɔ̀wǎy bvúzhı᷇ːde᷆n.  On Bvuzhiden, we will come to the market.", "item_type": "subproblem", "prompt_chars": 2322, "record_id": "iol-2019-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 76, "canonical_excerpt": "(b)\n Bvutfu, Bvuŋka, Bvuzhi, Bvukema, Bvuŋkaden, Bvuzhiden, Bvusɔw, Bvumbon.", "item_type": "subproblem", "prompt_chars": 2014, "record_id": "iol-2019-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4"], "canonical_answer_chars": 415, "canonical_excerpt": "(c)  20. On Bvusɔw, I helped the man.       me ncí nɔ̂ː nfı᷅dìɛ̀msɛ̌n bvúsɔ̀w.\n       21. The thief stole the yam just now.           cɔ̌ŋ nɔ̂ː cɔ᷅ŋ kèŋkfǔ.\n       22. On Bvuŋkaden, I will hear the car.    me nú nyúː nyów mutù bvûŋka᷆ːde᷆n.\n       23. Today, the woman will kill the man.      kwɔ᷆ːn nǔ géː yó dìɛ̀msɛ̌n ɛ̀bɛ̀n.\n       24. Today, the man saw you(sg).               dìɛ̀msɛ̌n bɛ̂ː yɛ᷅n wɔ̀ lě ɛ̀bɛ̀n.", "item_type": "subproblem", "prompt_chars": 4616, "record_id": "iol-2019-individual-p5-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 76, "canonical_excerpt": "(b)\n Bvutfu, Bvuŋka, Bvuzhi, Bvukema, Bvuŋkaden, Bvuzhiden, Bvusɔw, Bvumbon.", "item_type": "partial_variant", "prompt_chars": 2383, "record_id": "iol-2019-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 4, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4"], "canonical_answer_chars": 415, "canonical_excerpt": "(c)  20. On Bvusɔw, I helped the man.       me ncí nɔ̂ː nfı᷅dìɛ̀msɛ̌n bvúsɔ̀w.\n       21. The thief stole the yam just now.           cɔ̌ŋ nɔ̂ː cɔ᷅ŋ kèŋkfǔ.\n       22. On Bvuŋkaden, I will hear the car.    me nú nyúː nyów mutù bvûŋka᷆ːde᷆n.\n       23. Today, the woman will kill the man.      kwɔ᷆ːn nǔ géː yó dìɛ̀msɛ̌n ɛ̀bɛ̀n.\n       24. Today, the man saw you(sg).               dìɛ̀msɛ̌n bɛ̂ː yɛ᷅n wɔ̀ lě ɛ̀bɛ̀n.", "item_type": "partial_variant", "prompt_chars": 5089, "record_id": "iol-2019-individual-p5-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-sol.en.pdf", "source_problem_id": "2019-individual-5", "target_subparts": ["a", "b", "c", "full"], "title": "Nooni", "year": 2019}
+{"author": "André Nikulin", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Rikbaktsa (Macro-Jê)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 5, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1531, "canonical_excerpt": "Problem 5.  Verb structure:\n\n        subject                 tense              (object-)       root        (-subject)  [                   }\n     1st person sg         my-C               ik- 1st person sg         -naha 1st person pl\n∅-                                    future\n    3rd person sgfuture        p-V              a- 2nd person sg   } [                }\n                                                                i- 3rd person sgtsi-C     2nd person sg       ik-C\n                                    past    my- 1st person pl ts-V      1st person pl        ∅-V\n         }                           aha- 2nd person pl\n          ni-C\n                3rd person sgpast                 ", "item_type": "full_problem", "prompt_chars": 3506, "record_id": "iol-2021-individual-p5", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-sol.en.pdf", "source_problem_id": "2021-individual-5", "target_subparts": ["full"], "title": "Rikbaktsa", "year": 2021}
+{"author": "Tung-Le Pan", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Proto-Chamic (Chamic) Phan Rang Cham (Chamic) Tsat (Chamic)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 6, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 2959, "canonical_excerpt": "Problem 5.\n\nPhan Rang Cham:                             Tsat:\n (…) D (V L)  V  (C)             … h    … ay     … aN                                                                                           ;                    ;                     V̀                                ∅⁵⁵        aːiʔ⁴²        aːnʔ⁴²\n                                          Proto-Chamic … T    …({N, w, y})\n   • both syllables                 +D               ʔ⁴²               ¹¹\n                             −D                                                                ʔ²⁴               ³³      – {*b, *d, *g, *j} > {p, t, k, c}\n      – NV > Nɨ                            • {*b, *p, *g} (V) {*l, *r} > {ph, p, ", "item_type": "full_problem", "prompt_chars": 6944, "record_id": "iol-2022-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 17, "canonical_excerpt": "(a)      *phia¹¹.", "item_type": "subproblem", "prompt_chars": 3843, "record_id": "iol-2022-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 1536, "canonical_excerpt": "(b)\n      Proto-Chamic   Phan    Rang  Tsat          meaning\n                 Cham\n                                                             skin       *kulit            kaliʔ                (1)   liʔ²⁴\n                                                             black      *hitam        hatam              (2)   taːnʔ⁴²\n                                                 mouth      *bubah        papàh              (3)  pha⁵⁵\n                                                              to tie       *ʔikat               (4)   ikaʔ          (5)   kaʔ²⁴\n                                                     fathom      *dəpa               (6)   tapa          (7)   pa¹¹\n                           ", "item_type": "subproblem", "prompt_chars": 10125, "record_id": "iol-2022-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 1536, "canonical_excerpt": "(b)\n      Proto-Chamic   Phan    Rang  Tsat          meaning\n                 Cham\n                                                             skin       *kulit            kaliʔ                (1)   liʔ²⁴\n                                                             black      *hitam        hatam              (2)   taːnʔ⁴²\n                                                 mouth      *bubah        papàh              (3)  pha⁵⁵\n                                                              to tie       *ʔikat               (4)   ikaʔ          (5)   kaʔ²⁴\n                                                     fathom      *dəpa               (6)   tapa          (7)   pa¹¹\n                           ", "item_type": "partial_variant", "prompt_chars": 10195, "record_id": "iol-2022-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-sol.en.pdf", "source_problem_id": "2022-individual-5", "target_subparts": ["a", "b", "full"], "title": "Proto-Chamic, Phan Rang Cham and Tsat", "year": 2022}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Supyire (Atlantic-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 6, "problem_page_start": 6, "problem_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 956, "canonical_excerpt": "Problem 5.\n                                                           \n            400 :  kampwoo               80 :  ŋkuu               20 :  beɲjaaga\n   •                         +                    +                +\n       α1 × 400 :  kampwɔhii α1       α2 × 80 :  ŋkwuu α2       α3 × 20 :  be-α3\n      [10 : kɛ] + [5] + [β], 2 ≤α1,2,3 ≤4, 1 ≤β ≤4\n\n   • +: na\n\n   •   1:  niŋkin    -niŋkin →-ni\n\n        2:  shuunni\n\n        3:  taanre\n\n        4:   sicyɛɛre   -sicyɛɛre →-ricyɛɛre\n\n        5:  kaŋkuro  kaŋkuro na →baa-\n\nTwentieth International Linguistics Olympiad (2023)                                        6\n\nIndividual Contest Solutions\n\n (a)  kampwɔhii shuunni na kɛ  ", "item_type": "full_problem", "prompt_chars": 1720, "record_id": "iol-2023-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 69, "canonical_excerpt": "(a)  kampwɔhii shuunni na kɛ  810\n\n      ŋkuu na baataanre         88", "item_type": "subproblem", "prompt_chars": 697, "record_id": "iol-2023-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 214, "canonical_excerpt": "(b)    15  kɛ na kaŋkuro\n\n       109  ŋkuu na beɲjaaga na baaricyɛɛre\n\n       152  ŋkuu na beetaanre na kɛ na shuunni\n\n       403  kampwoo na taanre\n\n      1534  kampwɔhii taanre na ŋkwuu sicyɛɛre na kɛ na sicyɛɛre", "item_type": "subproblem", "prompt_chars": 2823, "record_id": "iol-2023-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 214, "canonical_excerpt": "(b)    15  kɛ na kaŋkuro\n\n       109  ŋkuu na beɲjaaga na baaricyɛɛre\n\n       152  ŋkuu na beetaanre na kɛ na shuunni\n\n       403  kampwoo na taanre\n\n      1534  kampwɔhii taanre na ŋkwuu sicyɛɛre na kɛ na sicyɛɛre", "item_type": "partial_variant", "prompt_chars": 2945, "record_id": "iol-2023-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 6, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-sol.en.pdf", "source_problem_id": "2023-individual-5", "target_subparts": ["a", "b", "full"], "title": "Supyire", "year": 2023}
+{"author": "Dan-Mircea Mirea", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Yanyuwa (Pama–Nyungan)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 7, "problem_page_start": 6, "problem_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 9, "answer_unit_ids": ["a.1", "a.2", "a.3", "b.1", "b.2", "b.3", "c.1", "c.2", "c.3"], "canonical_answer_chars": 2319, "canonical_excerpt": "Problem 5.  Word order:\n\n     • S V  (S = subject of a intransitive verb; V = verb)\n\n     • A-lu V P  (A = subject of a transitive verb; P = direct object)\n\nNoun:                                           Verb:\n\n       noun classes:  P = S   A                     • tense (j- present continuous, k- future/past)\n        women       rra-    rru-\n                                                                • (direct object) (♀)     men       nya-   nyu-\n (♂)  men = animals    ∅-      ki-                    • subject\n (♀)      animals      ∅-        ji-\n         edible things   ma-                           noun classes:   P    S    A\n               pl              li-     linji-              ", "item_type": "full_problem", "prompt_chars": 3508, "record_id": "iol-2024-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["a.1", "a.2", "a.3"], "canonical_answer_chars": 314, "canonical_excerpt": "(a)   15. mirningiya ka-wulumala — The man will run.\n        16. nyu-munangalu kalilu-wudurruma li-wakuku — The white man fed the dogs.\n        17. linji-ardulu jalu-wubanji ma-burlurlu — The boys/girls (children) are baking the bread.\n        18. rru-nhanawalu janda-athamanji ardu — The woman is chasing the boy.", "item_type": "subproblem", "prompt_chars": 2106, "record_id": "iol-2024-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 54, "canonical_excerpt": "(b)   15. ♂\n        16. ♀\n        17. ?⃝\n        18. ♂", "item_type": "subproblem", "prompt_chars": 1990, "record_id": "iol-2024-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 470, "canonical_excerpt": "(c)   19. The dog chased the girl.\n         ♀ji-wakukulu karrinju-athama rra-ardu\n         ♂ki-wakukulu karrilu-athama rra-ardu\n        20. The pregnant woman is feeding the old women.\n      ♀=♂rru-walkurulu jalanda-wudurrumanji li-bardibardi\n        21. The old man healed the Yanyuwa man.\n         ♀nyu-malbulu kanyilu-yabima nya-yanyuwa\n         ♂ki-malbulu kilu-yabima yanyuwa\n        22. The kangaroo will return.\n        ♀wunala ki-wanila\n        ♂wunala ka-wanila", "item_type": "subproblem", "prompt_chars": 4668, "record_id": "iol-2024-individual-p5-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 54, "canonical_excerpt": "(b)   15. ♂\n        16. ♀\n        17. ?⃝\n        18. ♂", "item_type": "partial_variant", "prompt_chars": 2357, "record_id": "iol-2024-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 470, "canonical_excerpt": "(c)   19. The dog chased the girl.\n         ♀ji-wakukulu karrinju-athama rra-ardu\n         ♂ki-wakukulu karrilu-athama rra-ardu\n        20. The pregnant woman is feeding the old women.\n      ♀=♂rru-walkurulu jalanda-wudurrumanji li-bardibardi\n        21. The old man healed the Yanyuwa man.\n         ♀nyu-malbulu kanyilu-yabima nya-yanyuwa\n         ♂ki-malbulu kilu-yabima yanyuwa\n        22. The kangaroo will return.\n        ♀wunala ki-wanila\n        ♂wunala ka-wanila", "item_type": "partial_variant", "prompt_chars": 5117, "record_id": "iol-2024-individual-p5-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-sol.en.pdf", "source_problem_id": "2024-individual-5", "target_subparts": ["a", "b", "c", "full"], "title": "Yanyuwa", "year": 2024}
+{"author": "Dan-Mircea Mirea", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Kaqchikel (Mayan)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 8, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 9, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "b.1", "c.1"], "canonical_answer_chars": 1487, "canonical_excerpt": "Problem 5.  Rules:\n\n     • Word order: VXY, XVY ({X, Y} = {S, O}; V = verb, S = subject, O = object)\n\n     • Noun: ri ( pl: taq ) N\n\n           O  S\n     • Verb: x- sg  —  ru  - stem ; before a vowel ru- > r-, ki- > k-\n                pl   e    ki\n     •\n                                    Auditory Cortex (activity)\n                                        higher       lower\n                           VXY      XVY\n       Frontal Cortex   higher  VS    VSO     OVS\n            (activity)     lower  VO    VOS     SVO\n\nAnswers:\n\n  (a)    1. Xerunïm ri taq säq ri xar\n          2. Ri taq [q’ëq]A xkich’äy [ri]B käq\n          3. [Xeroyoj]C ri q’ëq ri taq käq\n          4. Ri q’ëq xkoyoj [ri taq xar]D\n", "item_type": "full_problem", "prompt_chars": 4281, "record_id": "iol-2025-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 7, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7"], "canonical_answer_chars": 341, "canonical_excerpt": "(a)    1. Xerunïm ri taq säq ri xar\n          2. Ri taq [q’ëq]A xkich’äy [ri]B käq\n          3. [Xeroyoj]C ri q’ëq ri taq käq\n          4. Ri q’ëq xkoyoj [ri taq xar]D\n          5. [Xkinïm ri taq käq]E ri säq\n          6. Xkich’äy [ri xar ri taq säq]F\n          7. [Ri taq xar]G xeroyoj [ri käq]H\n          8. [Ri säq xeruch’äy ri taq q’ëq]I", "item_type": "subproblem", "prompt_chars": 2377, "record_id": "iol-2025-individual-p5-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 72, "canonical_excerpt": "(b)   9. Ri taq säq xkinïm ri q’ëq   10. Xekich’äy ri taq xar ri taq käq", "item_type": "subproblem", "prompt_chars": 2741, "record_id": "iol-2025-individual-p5-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 533, "canonical_excerpt": "(c)     • Xroyoj ri säq ri xar\n           • Xroyoj ri xar ri säq\n           • Ri xar xroyoj ri säq\n           • Ri säq xroyoj ri xar\n\n  (d)\n                                                    Frontal Cortex  Auditory Cortex\n                                                           (activity)         (activity)\n             11  Xeruq’etey ri k��q ri taq q’ëq    higher          higher\n             12   Xerachik’aj ri taq säq ri xar    lower           higher\n             13   Ri taq q’ëq xekitz’ët ri taq säq   ?              lower", "item_type": "subproblem", "prompt_chars": 5831, "record_id": "iol-2025-individual-p5-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 72, "canonical_excerpt": "(b)   9. Ri taq säq xkinïm ri q’ëq   10. Xekich’äy ri taq xar ri taq käq", "item_type": "partial_variant", "prompt_chars": 3135, "record_id": "iol-2025-individual-p5-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 533, "canonical_excerpt": "(c)     • Xroyoj ri säq ri xar\n           • Xroyoj ri xar ri säq\n           • Ri xar xroyoj ri säq\n           • Ri säq xroyoj ri xar\n\n  (d)\n                                                    Frontal Cortex  Auditory Cortex\n                                                           (activity)         (activity)\n             11  Xeruq’etey ri käq ri taq q’ëq    higher          higher\n             12   Xerachik’aj ri taq säq ri xar    lower           higher\n             13   Ri taq q’ëq xekitz’ët ri taq säq   ?              lower", "item_type": "partial_variant", "prompt_chars": 6325, "record_id": "iol-2025-individual-p5-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-sol.en.pdf", "source_problem_id": "2025-individual-5", "target_subparts": ["a", "b", "c", "full"], "title": "Kaqchikel", "year": 2025}

benchmark/IOL/ioling_hf/reports/manual_verification/batch_05.jsonl

@@ -0,0 +1,21 @@
+{"author": "Boris Iomdin", "curation_statuses": {"auto_extracted_needs_manual_review": 1}, "languages": ["French (Indo-European)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 5, "problem_page_end": 4, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1227, "canonical_excerpt": "Solution of Problem 5\nréassortir pick again assortir pick\nrécurer clean curer clean\nréformer reform IVNTIUITUVIQQ00NQU00IUI1\nreformer form again former form\nréfuter refute ITQTIUIUUVIQQ00000000111\nrelancer throw again lancer throw\nrémunérer remunerate —_||I|I{I|[|IIIIIIIIII\nrépartir distribute —_| |[[IIIIIIIIIIIIIIIIIIII\nThe table features verbs with two different prefixes: re- and ré-. All verbs with re- indicate a\nrepetition or a renewal of the action named by the verb without a prefix. Contrariwise, if the\nprefix is ré-, then the corresponding prefixless verb either doesn’t exist or means the same thing\nas the prefixed one does. The verbs whose stems begin with vowels are an exception: th", "item_type": "full_problem", "prompt_chars": 1731, "record_id": "iol-2003-individual-p5", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 4, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-indiv-sol.en.pdf", "source_problem_id": "2003-individual-5", "target_subparts": ["full"], "title": "French", "year": 2003}
+{"author": null, "curation_statuses": {"auto_extracted_needs_manual_review": 4}, "languages": ["Latin (Indo-European) English (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 3, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["assignment_1.1", "assignment_2.1"], "canonical_answer_chars": 1719, "canonical_excerpt": "Problem No.3\n\nConsider the four English words: beard, widow, word, red, and the four corresponding Latin\nwords: barba, vidua, verbum, rubrica. Note that all of the English words contain the d sound.\nSince this sound is the only one occurring in all four words, and since it sounds closest to dh, it\nwould be natural to assume that it is this very sound that replaced dh in the Latin words\ncognate to the English ones. At the same time, three of the four Latin words have 6 in that\nposition, and only one of them has d there. As for the rest of the words, one of them contains a\nb (glabra), and four of them contain a d (dividit, mandere, mordax, mediocris). Five words\nremain unexamined: fiimus, frac", "item_type": "full_problem", "prompt_chars": 1570, "record_id": "iol-2004-individual-p3", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 196, "canonical_excerpt": "Assignment 1. The dh sound could not have occurred in mordax: there is neither an f\nnor a b in this word, and the d cannot be the result of dh transition, since next to r, dh would\nhave yielded b.", "item_type": "subproblem", "prompt_chars": 1322, "record_id": "iol-2004-individual-p3-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 369, "canonical_excerpt": "Assignment 2. The dh sound could not have been there in brevis (at the beginning of the\nword dh would have yielded f even before r, as in fracés); in gurdus (after r, dh would have\nyielded b); and in débeo (at the beginning of the word dh would have yielded f, and in the - :\n\n| middle, in the absence of r, it would have yielded d)., | Orhopwatupopano: pycckwit\nOCccuA", "item_type": "subproblem", "prompt_chars": 1910, "record_id": "iol-2004-individual-p3-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 369, "canonical_excerpt": "Assignment 2. The dh sound could not have been there in brevis (at the beginning of the\nword dh would have yielded f even before r, as in fracés); in gurdus (after r, dh would have\nyielded b); and in débeo (at the beginning of the word dh would have yielded f, and in the - :\n\n| middle, in the absence of r, it would have yielded d)., | Orhopwatupopano: pycckwit\nOCccuA", "item_type": "partial_variant", "prompt_chars": 2170, "record_id": "iol-2004-individual-p3-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-sol.en.pdf", "source_problem_id": "2004-individual-3", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Latin", "year": 2004}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_needs_manual_review": 4}, "languages": ["Yoruba (Niger-Congo)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["assignment_1.1", "assignment_2.1"], "canonical_answer_chars": 933, "canonical_excerpt": "Problem 4\nThe modifier (the possessor) follows the head (the possessed) in the Yoruba phrases. If\nthe second word begins with i, this sound assimilates to the final vowel of the first word,\nwhatever it is; if the second word does not begin with i but rather with another vowel\n(a, €, @, 0, 9), the final vowel of the first word assimilates to this sound. All tones\nremain intact.\n(No word ever begins with u in Standard Yoruba; in those dialects where initial u does\noccur, however, it behaves exactly as i.)\nAssignment 1.\n[owa ake] the money of the axe [oko sya] the mother’s\n(i.e., the price of the axe) husband\n[eba alu] the vicinity of the city laze elu] the stranger’s dog\n(i.e., near the city)\n", "item_type": "full_problem", "prompt_chars": 1693, "record_id": "iol-2005-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 190, "canonical_excerpt": "Assignment 1.\n[owa ake] the money of the axe [oko sya] the mother’s\n(i.e., the price of the axe) husband\n[eba alu] the vicinity of the city laze elu] the stranger’s dog\n(i.e., near the city)", "item_type": "subproblem", "prompt_chars": 1098, "record_id": "iol-2005-individual-p4-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 233, "canonical_excerpt": "Assignment 2.\nthe head of the tree [ori igi] the witch’s city [ila a3é]\nthe house oflove [ile efe] the husband’s axe = [ako oko]\n\nThird International Olympiad in Linguistics. Solutions to the problems of the individual competition. 4", "item_type": "subproblem", "prompt_chars": 2390, "record_id": "iol-2005-individual-p4-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 233, "canonical_excerpt": "Assignment 2.\nthe head of the tree [ori igi] the witch’s city [ila a3é]\nthe house oflove [ile efe] the husband’s axe = [ako oko]\n\nThird International Olympiad in Linguistics. Solutions to the problems of the individual competition. 4", "item_type": "partial_variant", "prompt_chars": 2644, "record_id": "iol-2005-individual-p4-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 4, "solution_page_start": 3, "solution_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-sol.en.pdf", "source_problem_id": "2005-individual-4", "target_subparts": ["assignment_1", "assignment_2", "full"], "title": "Yoruba", "year": 2005}
+{"author": "Olga Fyodorova", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Ngoni (Niger-Congo)"], "problem_extraction_confidence": "heading", "problem_number": 5, "problem_page_end": 5, "problem_page_start": 5, "problem_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1862, "canonical_excerpt": "Problem #5\n\n         It is easy to tell what all nouns and the only verb mean. All common nouns denoting people\nhave the prefix m- in the singular and the prefix va- in the plural.\n\n       There are two possible orders of the Subject, Verb, Direct Object and Indirect Object:\n         (1) S V DO IO\n         (2) S V IO DO\n\n      The verb buy has the structure (a/va)(m/va)guli, with the  first prefix denoting the\nnumber of S and the second, of IO; DO’s number is not marked at all in the verb. We have no\nexamples where both S and IO are plural.\n        Clearly one or both prefixes may be omitted; however,\n         (*) If IO is marked in the verb, so shall S.\n\n         If the word order is (1), b", "item_type": "full_problem", "prompt_chars": 2211, "record_id": "iol-2006-individual-p5", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 5, "solution_page_start": 5, "solution_pdf_local_path": "data/raw/pdfs/iol-2006-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2006-indiv-sol.en.pdf", "source_problem_id": "2006-individual-5", "target_subparts": ["full"], "title": "Ngoni", "year": 2006}
+{"author": "Olga Fyodorova", "curation_statuses": {"manual_reviewed_source_text": 1}, "languages": ["Hawaiian (Austronesian)"], "problem_extraction_confidence": "team_full_document", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2007-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2007-team-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 304, "canonical_excerpt": "♀ Mihil\n\n            ♂ Akamu       ♀ Lola\n\n♂ Kimo         ♀ Halia\n\n♂ Aukai       ♂ Etana                      ♀ Kalena\n\n ♀ Loni        ♂ Akela       ♂ Elta          ♀ Nani      ♂ Makani\n\n            ♂ Keoki        ♀ Ola        ♂ Alika      ♂ Abia\n\n♂ Mano         ♀ Malia        ♀ Kani\n\n             ♂ Kai", "item_type": "full_problem", "prompt_chars": 3823, "record_id": "iol-2007-team-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "team_full_document", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2007-team-sol.m.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2007-team-sol.m.pdf", "source_problem_id": "2007-team-1", "target_subparts": ["full"], "title": "Hawaiian", "year": 2007}
+{"author": "Todor Tchervenkov", "curation_statuses": {"manual_reviewed_source_text": 20}, "languages": ["Mandarin (Sino-Tibetan) Cantonese (Sino-Tibetan)"], "problem_extraction_confidence": "team_full_document", "problem_number": 1, "problem_page_end": 3, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2008-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2008-team-prob.en.pdf", "record_count": 20, "record_types": {"full_problem": 1, "partial_variant": 9, "subproblem": 10}, "records": [{"answer_unit_count": 10, "answer_unit_ids": ["a.1", "b.1", "c.1", "d.1", "e.1", "f.1", "g.1", "h.1", "i.1", "j.1"], "canonical_answer_chars": 5557, "canonical_excerpt": "Sixth International Olympiad in Theoretical, Mathematical\n               and Applied Linguistics\n               Bulgaria, Sunny Beach, 4–9 August 2008\n                  Solution of the Problem of the Team Contest\n\n   The syllables of Chinese consist of three parts: onset (initial consonant, which may be missing\nas in 3B), rhyme (all following sounds) and tone. Cantonese tones can be thought of as having\ntwo distinct qualities: height (high or low) and contour (rising, level or falling).\n                                      rising        level        falling\n                 !!*    -    HHj\n                          high  35         3         53\n                         low   13 !!*   2  -  ", "item_type": "full_problem", "prompt_chars": 8520, "record_id": "iol-2008-team-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 306, "canonical_excerpt": "(a) To use a fanqie transcription in Cantonese, A’s onset and tone height are combined with B’s\n     rhyme and tone contour. But if A’s (and X’s) tone is low, X’s onset, if a stop, must always\n     be aspirated if B’s (and X’s) tone is rising (13) or falling (21), and unaspirated if it is level\n      (2).", "item_type": "subproblem", "prompt_chars": 1763, "record_id": "iol-2008-team-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 489, "canonical_excerpt": "(b) Certainly the onset was from the A character, and the rhyme from B. But the aspiration\n      rule is strange. Probably it was not part of the original fanqie system. Maybe the tone came\n     from only one of the two characters? That has to be B, because the old rule should give\n      correct results in only one transcription.\n     Thus the original simple rule for fanqie was: A’s onset is combined with B’s rhyme and tone.\n     Only transcription 11 can be read now using this rule.", "item_type": "subproblem", "prompt_chars": 3209, "record_id": "iol-2008-team-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 676, "canonical_excerpt": "(c) Looking at the syllables with a sonorant onset, we see that they are always in a low tone (13,\n     2 or 21). Assuming that all voiced consonants evolved alike in Cantonese, we may conclude\n     that what is in a low tone now, had a voiced onset earlier. This is also true of the character\n      of the example from Wu. What is said in (d) supports this idea.\n     Thus the characters whose onsets were voiced are: 1X and 1A, 2X (=6B) and 2A, 3X and\n     3A, 3B (if it had an onset at all), 4X and 4A, 5X and 5A, 7B (=14A), 9X and 9A, 14X, 15X\n     and 15A, 16B.\n     Voiced stops became aspirated if the tone was rising or falling, and unaspirated if it was\n       level.", "item_type": "subproblem", "prompt_chars": 2093, "record_id": "iol-2008-team-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 639, "canonical_excerpt": "(d) The contours of the Cantonese tones correspond to the three tones of Classical Chinese; tone\n      height is an innovation brought about by the evolution of the voiced consonants.\n\nNow we can explain why fanqie transcriptions should be read in Cantonese the way they are. The\nX character has the same tone height as A because it got its onset from A, and height in Cantonese\nis determined by the voicing of the onset in Classical Chinese. But if the onset was a voiced stop,\nit could evolve in different ways in X and A, because its aspiration was determined by the tone\ncontour, which X got from B, and it could differ from A’s contour.", "item_type": "subproblem", "prompt_chars": 3901, "record_id": "iol-2008-team-p1-sub-d", "target_subparts": ["d"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["e.1"], "canonical_answer_chars": 996, "canonical_excerpt": "(e) In Mandarin onsets and rhymes are not combined in such a straightforward way as in Can-\n      tonese. It can be noted that after ´x (k´, k´h) we always find i or y, whereas x (k, kh), s (c,\n     ch) and s‘ (c‘, c‘ h) are never followed by these vowels.\n    We already know that the onset came from A and the rhyme from B. When the constraint\n     above came into being,\n\nSixth International Olympiad in Linguistics (2008).                                          2\nSolution of the Problem of the Team Contest\n\n       •  i was lost and y became u after s‘ (c‘, c‘ h);\n       • x (k, kh) and s (c, ch) became ´x (k´, k´h) before i or y.\n     These are also the rules that we must apply when using a", "item_type": "subproblem", "prompt_chars": 3587, "record_id": "iol-2008-team-p1-sub-e", "target_subparts": ["e"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["f.1"], "canonical_answer_chars": 1313, "canonical_excerpt": "(f) On the basis of the tone of the Cantonese syllable we can determine whether the onset was\n      voiced or not in Classical Chinese. In Mandarin the tones developed as follows:\n       • rising: 51 if the onset was voiced but not a sonorant, 214 otherwise;\n       • level: 51 (always);\n       • falling: 5 if the onset was voiceless, 35 otherwise.\n    We see that the contour is not preserved here. Voiced stops became aspirated if the tone\n     was falling, and unaspirated if it was level or rising.\n     In fanqie transcriptions read in Mandarin the tones work as follows:\n                        5, 35    214     (F, H−)51  (H+, L)51\n               5    5      214      214, 51       51\n       ", "item_type": "subproblem", "prompt_chars": 1951, "record_id": "iol-2008-team-p1-sub-f", "target_subparts": ["f"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["g.1"], "canonical_answer_chars": 176, "canonical_excerpt": "(g) Syllables with a sonorant onset and tone 5 or with an unaspirated onset and tone 35 should\n     not exist in Mandarin (if they do, then the rules must have had exceptions).", "item_type": "subproblem", "prompt_chars": 2777, "record_id": "iol-2008-team-p1-sub-g", "target_subparts": ["g"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["h.1"], "canonical_answer_chars": 50, "canonical_excerpt": "(h) 46: 21, 47: 51, 48: 13, 49: 35, 50: 53, 51: 2.", "item_type": "subproblem", "prompt_chars": 1685, "record_id": "iol-2008-team-p1-sub-h", "target_subparts": ["h"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["i.1"], "canonical_answer_chars": 42, "canonical_excerpt": "(i) 52 thai53, 53 siu3, 54 loN2, 55 paai2.", "item_type": "subproblem", "prompt_chars": 1979, "record_id": "iol-2008-team-p1-sub-i", "target_subparts": ["i"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["j.1"], "canonical_answer_chars": 132, "canonical_excerpt": "(j) 56 sai51, 57 k´ian214, 58 s‘aN214, 59 phai214, 60 ´xyan51, 61 k´hyan214, 62 ´xia51, 63 xan51,\n     64 c‘ou51, 65 naN35, 66 sai5.", "item_type": "subproblem", "prompt_chars": 6947, "record_id": "iol-2008-team-p1-sub-j", "target_subparts": ["j"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 489, "canonical_excerpt": "(b) Certainly the onset was from the A character, and the rhyme from B. But the aspiration\n      rule is strange. Probably it was not part of the original fanqie system. Maybe the tone came\n     from only one of the two characters? That has to be B, because the old rule should give\n      correct results in only one transcription.\n     Thus the original simple rule for fanqie was: A’s onset is combined with B’s rhyme and tone.\n     Only transcription 11 can be read now using this rule.", "item_type": "partial_variant", "prompt_chars": 3568, "record_id": "iol-2008-team-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 676, "canonical_excerpt": "(c) Looking at the syllables with a sonorant onset, we see that they are always in a low tone (13,\n     2 or 21). Assuming that all voiced consonants evolved alike in Cantonese, we may conclude\n     that what is in a low tone now, had a voiced onset earlier. This is also true of the character\n      of the example from Wu. What is said in (d) supports this idea.\n     Thus the characters whose onsets were voiced are: 1X and 1A, 2X (=6B) and 2A, 3X and\n     3A, 3B (if it had an onset at all), 4X and 4A, 5X and 5A, 7B (=14A), 9X and 9A, 14X, 15X\n     and 15A, 16B.\n     Voiced stops became aspirated if the tone was rising or falling, and unaspirated if it was\n       level.", "item_type": "partial_variant", "prompt_chars": 2969, "record_id": "iol-2008-team-p1-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 639, "canonical_excerpt": "(d) The contours of the Cantonese tones correspond to the three tones of Classical Chinese; tone\n      height is an innovation brought about by the evolution of the voiced consonants.\n\nNow we can explain why fanqie transcriptions should be read in Cantonese the way they are. The\nX character has the same tone height as A because it got its onset from A, and height in Cantonese\nis determined by the voicing of the onset in Classical Chinese. But if the onset was a voiced stop,\nit could evolve in different ways in X and A, because its aspiration was determined by the tone\ncontour, which X got from B, and it could differ from A’s contour.", "item_type": "partial_variant", "prompt_chars": 5481, "record_id": "iol-2008-team-p1-partial-through-abc-target-d", "target_subparts": ["d"], "visible_solved_subparts": ["a", "b", "c"]}, {"answer_unit_count": 1, "answer_unit_ids": ["e.1"], "canonical_answer_chars": 996, "canonical_excerpt": "(e) In Mandarin onsets and rhymes are not combined in such a straightforward way as in Can-\n      tonese. It can be noted that after ´x (k´, k´h) we always find i or y, whereas x (k, kh), s (c,\n     ch) and s‘ (c‘, c‘ h) are never followed by these vowels.\n    We already know that the onset came from A and the rhyme from B. When the constraint\n     above came into being,\n\nSixth International Olympiad in Linguistics (2008).                                          2\nSolution of the Problem of the Team Contest\n\n       •  i was lost and y became u after s‘ (c‘, c‘ h);\n       • x (k, kh) and s (c, ch) became ´x (k´, k´h) before i or y.\n     These are also the rules that we must apply when using a", "item_type": "partial_variant", "prompt_chars": 5834, "record_id": "iol-2008-team-p1-partial-through-abcd-target-e", "target_subparts": ["e"], "visible_solved_subparts": ["a", "b", "c", "d"]}, {"answer_unit_count": 1, "answer_unit_ids": ["f.1"], "canonical_answer_chars": 1313, "canonical_excerpt": "(f) On the basis of the tone of the Cantonese syllable we can determine whether the onset was\n      voiced or not in Classical Chinese. In Mandarin the tones developed as follows:\n       • rising: 51 if the onset was voiced but not a sonorant, 214 otherwise;\n       • level: 51 (always);\n       • falling: 5 if the onset was voiceless, 35 otherwise.\n    We see that the contour is not preserved here. Voiced stops became aspirated if the tone\n     was falling, and unaspirated if it was level or rising.\n     In fanqie transcriptions read in Mandarin the tones work as follows:\n                        5, 35    214     (F, H−)51  (H+, L)51\n               5    5      214      214, 51       51\n       ", "item_type": "partial_variant", "prompt_chars": 5222, "record_id": "iol-2008-team-p1-partial-through-abcde-target-f", "target_subparts": ["f"], "visible_solved_subparts": ["a", "b", "c", "d", "e"]}, {"answer_unit_count": 1, "answer_unit_ids": ["g.1"], "canonical_answer_chars": 176, "canonical_excerpt": "(g) Syllables with a sonorant onset and tone 5 or with an unaspirated onset and tone 35 should\n     not exist in Mandarin (if they do, then the rules must have had exceptions).", "item_type": "partial_variant", "prompt_chars": 7389, "record_id": "iol-2008-team-p1-partial-through-abcdef-target-g", "target_subparts": ["g"], "visible_solved_subparts": ["a", "b", "c", "d", "e", "f"]}, {"answer_unit_count": 1, "answer_unit_ids": ["h.1"], "canonical_answer_chars": 50, "canonical_excerpt": "(h) 46: 21, 47: 51, 48: 13, 49: 35, 50: 53, 51: 2.", "item_type": "partial_variant", "prompt_chars": 6501, "record_id": "iol-2008-team-p1-partial-through-abcdefg-target-h", "target_subparts": ["h"], "visible_solved_subparts": ["a", "b", "c", "d", "e", "f", "g"]}, {"answer_unit_count": 1, "answer_unit_ids": ["i.1"], "canonical_answer_chars": 42, "canonical_excerpt": "(i) 52 thai53, 53 siu3, 54 loN2, 55 paai2.", "item_type": "partial_variant", "prompt_chars": 6873, "record_id": "iol-2008-team-p1-partial-through-abcdefgh-target-i", "target_subparts": ["i"], "visible_solved_subparts": ["a", "b", "c", "d", "e", "f", "g", "h"]}, {"answer_unit_count": 1, "answer_unit_ids": ["j.1"], "canonical_answer_chars": 132, "canonical_excerpt": "(j) 56 sai51, 57 k´ian214, 58 s‘aN214, 59 phai214, 60 ´xyan51, 61 k´hyan214, 62 ´xia51, 63 xan51,\n     64 c‘ou51, 65 naN35, 66 sai5.", "item_type": "partial_variant", "prompt_chars": 11911, "record_id": "iol-2008-team-p1-partial-through-abcdefghi-target-j", "target_subparts": ["j"], "visible_solved_subparts": ["a", "b", "c", "d", "e", "f", "g", "h", "i"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "team_full_document", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2008-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2008-team-sol.en.pdf", "source_problem_id": "2008-team-1", "target_subparts": ["a", "b", "c", "d", "e", "f", "full", "g", "h", "i", "j"], "title": "Fanqie", "year": 2008}
+{"author": "Boris Iomdin", "curation_statuses": {"manual_reviewed_source_text": 1}, "languages": ["Vietnamese (Austro-Asiatic)"], "problem_extraction_confidence": "team_full_document", "problem_number": 1, "problem_page_end": 4, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2009-team-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2009-team-prob.en-gb.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 18412, "canonical_excerpt": "7th International Linguistics Olympiad (2009)\n                                                                    Solution to the Team Contest Problem\n\nTEXT 1. MY ROOM\n\nĐây là phòng của tôi (This is my room).\n\nTrong phòng có nhiều đồ đạc (In this room, there are a lot of things).\n\nĐây là bàn và ghế (These are tables and chairs).\n\nTrên bàn có một cái máy vi tính, một vài đĩa CD, một vài quyển sách, một cuốn từ\nđiển Anh - Việt và rất nhiều bút (On the table there are a computer, some CDs,\nsome books, a dictionary and a lot of pens).\n\nĐây là giường của tôi (This is my bed).\n\nTrên giường có gối, chăn và một cái điều khiển ti vi (On the bed, there are pillows,\nblankets, and a TV remote control).\n", "item_type": "full_problem", "prompt_chars": 11726, "record_id": "iol-2009-team-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "team_full_document", "solution_page_end": 10, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2009-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2009-team-sol.en.pdf", "source_problem_id": "2009-team-1", "target_subparts": ["full"], "title": "Vietnamese", "year": 2009}
+{"author": "Boris Iomdin", "curation_statuses": {"manual_reviewed_source_text": 1}, "languages": ["Mongolian (Mongolic)"], "problem_extraction_confidence": "team_full_document", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2010-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2010-team-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 9907, "canonical_excerpt": "agaar           14, 35, 35, 35       air                         gal               1, 9, 21              fire\nagaart         6                                           gald            20, 25\nagaaryn       6                                         garaq            1, 23, 29, 30       appear, exude\namt          8                    taste                    garsan        9                 ascending\namtlag           7, 37                 tasty, sweet              gazryn        35                earthen\namttaj           2, 8                 tasty                  gašunduu      8                    bitter\nam’tny          10, 23              edible                  gedeg         13                ", "item_type": "full_problem", "prompt_chars": 3558, "record_id": "iol-2010-team-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "team_full_document", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2010-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2010-team-sol.en.pdf", "source_problem_id": "2010-team-1", "target_subparts": ["full"], "title": "Mongolian", "year": 2010}
+{"author": "Adam Hesterberg", "curation_statuses": {"manual_reviewed_source_text": 6}, "languages": ["Sanskrit (Indo-Aryan)"], "problem_extraction_confidence": "team_full_document", "problem_number": 1, "problem_page_end": 7, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2011-team-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2011-team-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "c.1"], "canonical_answer_chars": 2198, "canonical_excerpt": "Ninth International Olympiad in Linguistics\n            Pittsburgh (United States of America), 24–31 July 2011\n\n                          Team Contest Solution\n\nDistribute at 180 minutes\n\nA syllable is guru  if and only if it has a long vowel or a diphthong or ends in a consonant. Division\ninto syllables ignores word divisions. A sequence of type VCV is divided as V-CV; of type VCCV, as\nVC-CV.\n   Each line describes the metre in which it’s written.\n   Each of the first 8 syllables of yam¯at¯ar¯ajabh¯anasalagam stands for the pattern of laghu and guru\nsyllables in that and the next two syllables. The syllables la and ga stand for 1 laghu and 1 guru.\n   To make a metre’s mnemonic, group syllabl", "item_type": "full_problem", "prompt_chars": 13916, "record_id": "iol-2011-team-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 126, "canonical_excerpt": "(a) What is the additional rule of transliteration?\n   — The vowels e and o are long, although they are written without macra.", "item_type": "subproblem", "prompt_chars": 2777, "record_id": "iol-2011-team-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 1224, "canonical_excerpt": "(b)   • Deleted words: jarau, lagau.\n       • Changed letters:\n              1. Line 1: bhujanga-pray¯atam caturbhirgakaraih. < bhujanga-pray¯atam caturbhiryak¯araih.\n              2. Line 4: mabhalag¯a gajagatih. < nabhalag¯a gajagatih.\n              3. Line 5: mo go go go vidyunm¯al¯a < mo mo go go vidyunm¯al¯a\n       • Words with added macra:\n              1. Line 2: gurunidhanam¯anulaghur iha ś¯aśikal¯a < gurunidhanamanulaghur iha śaśikal¯a\n              2. Line 2: gurunidhanam¯anulaghur iha ś¯aśikal¯a < gurunidhanamanulaghur iha śaśikal¯a\n              3. Line 9: syad indrav¯ajr¯a yadi tau jagau gah. < sy¯ad indravajr¯a yadi tau jagau gah.\n              4. Line 10: ¯upendravajr¯a pratha", "item_type": "subproblem", "prompt_chars": 2949, "record_id": "iol-2011-team-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 70, "canonical_excerpt": "(c) Syllables 2, 3, 4, 6, and 10 are guru: yam¯at¯ar¯ajabh¯anasalagam.", "item_type": "subproblem", "prompt_chars": 24855, "record_id": "iol-2011-team-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 1224, "canonical_excerpt": "(b)   • Deleted words: jarau, lagau.\n       • Changed letters:\n              1. Line 1: bhujanga-pray¯atam caturbhirgakaraih. < bhujanga-pray¯atam caturbhiryak¯araih.\n              2. Line 4: mabhalag¯a gajagatih. < nabhalag¯a gajagatih.\n              3. Line 5: mo go go go vidyunm¯al¯a < mo mo go go vidyunm¯al¯a\n       • Words with added macra:\n              1. Line 2: gurunidhanam¯anulaghur iha ś¯aśikal¯a < gurunidhanamanulaghur iha śaśikal¯a\n              2. Line 2: gurunidhanam¯anulaghur iha ś¯aśikal¯a < gurunidhanamanulaghur iha śaśikal¯a\n              3. Line 9: syad indrav¯ajr¯a yadi tau jagau gah. < sy¯ad indravajr¯a yadi tau jagau gah.\n              4. Line 10: ¯upendravajr¯a pratha", "item_type": "partial_variant", "prompt_chars": 3128, "record_id": "iol-2011-team-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 70, "canonical_excerpt": "(c) Syllables 2, 3, 4, 6, and 10 are guru: yam¯at¯ar¯ajabh¯anasalagam.", "item_type": "partial_variant", "prompt_chars": 26286, "record_id": "iol-2011-team-p1-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "team_full_document", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2011-team-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2011-team-sol.en-gb.pdf", "source_problem_id": "2011-team-1", "target_subparts": ["a", "b", "c", "full"], "title": "Sanskrit Poetry", "year": 2011}
+{"author": null, "curation_statuses": {"manual_reviewed_source_text": 1}, "languages": ["Lao (Tai-Kadai)"], "problem_extraction_confidence": "team_full_document", "problem_number": 1, "problem_page_end": 1, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2012-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2012-team-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 4, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4"], "canonical_answer_chars": 4514, "canonical_excerpt": "Tenth International Olympiad in Linguistics\n\n               Ljubljana (Slovenia), 30 July – 3 August 2012\n\n                        Team Contest Solution\n\n 1.  Indonesia                        [ind¯on¯esiya]                                                             30.  South Korea   [kawhl¯ı t´ai]\n 2.  North Korea                [kawhl¯ı hn¯0a]                                                            ˙                                                                ˙                   31.   Ireland            [a˙yklˆa˙n] 3.  Nigeria                             [n¯ık¯eliya]                                                             32.  Venezuela        [v¯en¯es¯u’¯el¯a] 4.  Afghanistan  ", "item_type": "full_problem", "prompt_chars": 492, "record_id": "iol-2012-team-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "team_full_document", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2012-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2012-team-sol.en.pdf", "source_problem_id": "2012-team-1", "target_subparts": ["full"], "title": "Lao", "year": 2012}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Benabena (Trans-New Guinea)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 1262, "canonical_excerpt": "Problem #1.  Rules:\n\n   1. present: no-\n\n                     singular   plural\n   2. object =  1    nV-    lV-   , V  is the vowel from the following syllable\n              2     kV-  lenV-\n\n                                      `strike'   `pierce'\n   3.   present, subject = 1st  ho-     fu-\n                   otherwise  ha-        -\n              {\n                           1st :   lu-\n   4. future, subject =\n                      2nd :  la-\n\n   5. subject = dual: -'i\n\n                                                     indicative   `because . . . '\n   6.  subject = 1st person pl or 2nd person sg  -ne        -tagihe\n                                     otherwise  -be        -nagihe\n\n   A", "item_type": "full_problem", "prompt_chars": 1473, "record_id": "iol-2014-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 231, "canonical_excerpt": "(a)  noni be        youpl are piercing me\n      halu'ibe       we both will strike him\n         li latagihe     because yousg will pierce us\n     nokufune      we are piercing yousg\n      nolahanagihe  because youpl are striking us", "item_type": "subproblem", "prompt_chars": 1135, "record_id": "iol-2014-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 318, "canonical_excerpt": "(b)  noha'ibe         you both are striking him\n       ki lune         we will pierce yousg\n      nolenohotagihe  because we are striking youpl\n       lanagihe        because youpl will pierce him\n\nTwelfth International Olympiad in Linguistics (2014).                                     2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1891, "record_id": "iol-2014-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 318, "canonical_excerpt": "(b)  noha'ibe         you both are striking him\n       ki lune         we will pierce yousg\n      nolenohotagihe  because we are striking youpl\n       lanagihe        because youpl will pierce him\n\nTwelfth International Olympiad in Linguistics (2014).                                     2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2175, "record_id": "iol-2014-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-sol.en-gb.pdf", "source_problem_id": "2014-individual-1", "target_subparts": ["a", "b", "full"], "title": "Benabena", "year": 2014}
+{"author": "Milena Veneva", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Nahuatl (Uto-Aztecan) Arammba (South Central Papuan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["b.1", "c.1"], "canonical_answer_chars": 3111, "canonical_excerpt": "Problem 1.   Nahuatl:\n\n   • 1: cë, 2: öme, 3: ëyi, 4: nähui;\n\n   • 5: mäcuïlli, 10: mahtlactli, 15: caxtölli;\n\n                          α\n                                           1:    ceM      20β\n                                           2:    öm       20:   pöhualli\n   • α × 20β, 1 ≤α ≤5, 1 ≤β ≤3:                    -                               ;\n                                           3:      yë     400:   tzontli\n                                           4:    näuh    8000:   xiquipilli\n                                           5:  mäcuïl\n\n   • 7: chicöme;\n       {                }\n             γ ∈{10, 15}, 1 ≤δ ≤4\n   • γ + δ,                                                 ", "item_type": "full_problem", "prompt_chars": 2298, "record_id": "iol-2015-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 130, "canonical_excerpt": "(b)   • 42 = 2 × 20 + 2: öm-pöhualli-om-öme;\n\n       • 494 = 1×400+4×20+10+4: cen-tzontli-on-näuh-pöhualli-om-mahtlactli-on-nähui.", "item_type": "subproblem", "prompt_chars": 1792, "record_id": "iol-2015-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 126, "canonical_excerpt": "(c)   • 43 = 36 + 6 + 1: fete nimbo ngámbi;\n\n       • 569 = 2×216+3×36+4×6+5: yànparo tarumba yenówe fete asàr tàxwo tambaroy.", "item_type": "subproblem", "prompt_chars": 2736, "record_id": "iol-2015-individual-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 126, "canonical_excerpt": "(c)   • 43 = 36 + 6 + 1: fete nimbo ngámbi;\n\n       • 569 = 2×216+3×36+4×6+5: yànparo tarumba yenówe fete asàr tàxwo tambaroy.", "item_type": "partial_variant", "prompt_chars": 2919, "record_id": "iol-2015-individual-p1-partial-through-b-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["b"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-sol.en-gb.pdf", "source_problem_id": "2015-individual-1", "target_subparts": ["b", "c", "full"], "title": "Nahuatl and Arammba", "year": 2015}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Aralle-Tabulahan (Austronesian)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 3, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 15, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4", "full.5", "full.6", "full.7", "full.8", "full.9", "full.10", "full.11", "full.12", "full.13", "full.14", "full.15"], "canonical_answer_chars": 1187, "canonical_excerpt": "Problem 1.  The words bete’, dai’, naung, pano, sau and tama reflect the direction at\nthe last stage of the journey:\n\n   • bete’ — across a river;\n\n   • dai’ — upwards;\n\n   • naung — downwards;\n\n   • pano — on a flat road;\n\n   • sau — downstream;\n\n   • tama — upstream.\n\n   Answers:\n\n   1. In Kahangang hamlet:  Lamaoä’ dai’ di Palempang.\n\n   2. In Kombeng hamlet:  Lamaoä’ dai’ di Palempang.\n\n   3. In Kota hamlet:  Lamaoä’ bete’ di Bulung.\n\n   4. In Kota hamlet:  Lamaoä’ sau di Sohongang.\n\n   5. In Palempang hamlet:  Lamaoä’ naung di Bulung.\n\n   6. In Palempang hamlet:  Lamaoä’ sau di Kota.\n\n   7. In Palempang hamlet:  Lamaoä’ naung di Pahihuang.\n\n   8. In Bulung hamlet:  Lamaoä’ tama di Kota.\n\n", "item_type": "full_problem", "prompt_chars": 3570, "record_id": "iol-2016-individual-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-sol.en.pdf", "source_problem_id": "2016-individual-1", "target_subparts": ["full"], "title": "Aralle-Tabulahan", "year": 2016}
+{"author": "Milena Veneva", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Birom (Atlantic Congo)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 10, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "b.1", "c.1"], "canonical_answer_chars": 1062, "canonical_excerpt": "Problem 1.  The number system is duodecimal.\n\n   • gwīnìŋ — 1, bà — 2, tàt — 3, nààs — 4, tùŋūn — 5, tìīmìn — 6, tàāmà — 7, rwīīt — 8\n\n   • ʃāā-α = 12 −α (1 ≤α ≤3): ʃāātàt — 9, ʃāābà — 10, ʃāāgwīnìŋ — 11\n\n   • kūrū — 12\n\n   • bā-kūrū bī-¯γ = γ · 12 (2 ≤γ ≤8), bā-kūrū ʃāā-bī-¯γ = (12 −γ) · 12 (1 ≤γ ≤3)\n      (the tone in the first syllable of γ becomes middle)\n       {             }\n              gwɛ̄ gwīnìŋ (δ = 1)\n   • β ná               = β + δ (β = k · 12)\n                   vɛ̀  δ (2 ≤δ ≤11)\n\nAnswers:\n\n (a)    1. 52 + 3 + 4 = 32\n         2. 34 = 81\n         3. 72 + 9 + 1 = 59\n         4. 91 = 9\n         5. 82 + 2 + 5 = 71\n         6. 25 = 32\n         7. 92 + 4 + 3 = 88\n         8. 43 = 64", "item_type": "full_problem", "prompt_chars": 1719, "record_id": "iol-2017-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 8, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8"], "canonical_answer_chars": 223, "canonical_excerpt": "(a)    1. 52 + 3 + 4 = 32\n         2. 34 = 81\n         3. 72 + 9 + 1 = 59\n         4. 91 = 9\n         5. 82 + 2 + 5 = 71\n         6. 25 = 32\n         7. 92 + 4 + 3 = 88\n         8. 43 = 64\n\n         9. 16 + 21 = 18 + 2 + 17", "item_type": "subproblem", "prompt_chars": 932, "record_id": "iol-2017-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 102, "canonical_excerpt": "(b) bākūrū bītāt — 36, ʃāāgwīnìŋ — 11, kūrū — 12.\n\n      A. 108 −3 −13 = 92\n\n       B. 49 −14 −15 = 20", "item_type": "subproblem", "prompt_chars": 1394, "record_id": "iol-2017-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 219, "canonical_excerpt": "(c) 6 — tìīmìn, 22 — kūrū ná vɛ̀ ʃāābà, 97 — bākūrū bīrwīīt ná gwɛ̄ gwīnìŋ, 120 — bākūrū ʃāābībā.\n\nFifteenth International Linguistics Olympiad (2017).                                      2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2032, "record_id": "iol-2017-individual-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 102, "canonical_excerpt": "(b) bākūrū bītāt — 36, ʃāāgwīnìŋ — 11, kūrū — 12.\n\n      A. 108 −3 −13 = 92\n\n       B. 49 −14 −15 = 20", "item_type": "partial_variant", "prompt_chars": 1670, "record_id": "iol-2017-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 219, "canonical_excerpt": "(c) 6 — tìīmìn, 22 — kūrū ná vɛ̀ ʃāābà, 97 — bākūrū bīrwīīt ná gwɛ̄ gwīnìŋ, 120 — bākūrū ʃāābībā.\n\nFifteenth International Linguistics Olympiad (2017).                                      2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2438, "record_id": "iol-2017-individual-p1-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-sol.en.pdf", "source_problem_id": "2017-individual-1", "target_subparts": ["a", "b", "c", "full"], "title": "Birom", "year": 2017}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Creek (Muskogean)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1674, "canonical_excerpt": "Problem 1.\n\n                   0 or more times\n                                                        ´↓       ◦ =  U*Vo                                                                   are                                                                  consonants                                               U*Vo,                                                            *,                                                                    *′                  ⎫                           $ ⎧                           ◦◦       word =   !⎧                      \"# ⎫                                                           a                                                               vowe", "item_type": "full_problem", "prompt_chars": 1921, "record_id": "iol-2018-individual-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-sol.en.pdf", "source_problem_id": "2018-individual-1", "target_subparts": ["full"], "title": "Creek", "year": 2018}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Yonggom (Trans-New Guinea)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "b.1", "b.2", "b.3"], "canonical_answer_chars": 1918, "canonical_excerpt": "Problem 1.  Word order:\n\n   • (O) V-IS (S = subject; O = object; V = predicate)\n\n   • (O1) V1-I=S (O2) V2-IS\n\n   • (O1) V1-IS1-a (O2) V2-IS2 (S1̸ = S2)\n\n   • N1-a N2-a ‘N1 and N2’  (N = noun)\n\nbefore a vowel: b > w\n   In some verbs the vowel in the last syllable of the root assimilates to the vowel of the first\nsyllable of the ending.\n               S                     I=S                       IS\n                                     (S1 = S2)  past   present   future\n                   1st person sg      -en-i       -aan   -iin       -an-iin\n               2nd person sg     -en-eb            -eeb     -an-eeb\n                 3rd person sg m   -en-e       -oon  -een     -an-een\n            ", "item_type": "full_problem", "prompt_chars": 1817, "record_id": "iol-2019-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5"], "canonical_answer_chars": 524, "canonical_excerpt": "(a)  12. Om benu aneen.                   She took the bread and ate it.\n       13. Munuuna wunuub.                 She comes and we go away.\n       14. Wingkiiwa wengamburuun.         They sing and she hears.\n       15. Anon  ye  weng  wengamberenib  They will hear the dog and will get up.\n           yeedaraniib.\n       16. Oon wedmaniina karub wedmaneeb.   I will see the fish and you(sg) will see the\n                                       man.\n       17. Ok wedmeena aniin.             He sees the water and I drink it.", "item_type": "subproblem", "prompt_chars": 1586, "record_id": "iol-2019-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 482, "canonical_excerpt": "(b)  18. We will go away and they will come.      Wananuuwa mananiib.\n       19. I take the pig and he takes the fish.       Awon biina oon been.\n       20. He saw the water and the house and heard  Oka ambiwa wedmene anon ye weng\n           the dog.                              wengamboroon.\n       21. You(sg) eat the bread and sing.       Om aneneb wingkeeb.\n\nSeventeenth International Linguistics Olympiad (2019)                                    2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2128, "record_id": "iol-2019-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 482, "canonical_excerpt": "(b)  18. We will go away and they will come.      Wananuuwa mananiib.\n       19. I take the pig and he takes the fish.       Awon biina oon been.\n       20. He saw the water and the house and heard  Oka ambiwa wedmene anon ye weng\n           the dog.                              wengamboroon.\n       21. You(sg) eat the bread and sing.       Om aneneb wingkeeb.\n\nSeventeenth International Linguistics Olympiad (2019)                                    2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2705, "record_id": "iol-2019-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-sol.en.pdf", "source_problem_id": "2019-individual-1", "target_subparts": ["a", "b", "full"], "title": "Yonggom", "year": 2019}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Ekari (Trans-New Guinea)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 627, "canonical_excerpt": "Problem 1.\n\n   • ena 1, wija 2, wido 3, wi 4, idibi 5, benomi 6, waroewo 8, ije 9\n   • rati 10, mepina 20, joka rati 30, mepija 40, rati beo 50, moeto 60\n   • α-ma 10β = 10β + α  (1 ≤α ≤9, 1 ≤β ≤5)\n   • ije 10(β + 1) = ijema 10β  (1 ≤β ≤5)\n   • γ daemita moeto = 60 + γ  (1 ≤γ ≤59)\n   • moeto δ [ma γ ka (δ + 1)] = 60δ [+ γ]  (1 ≤γ ≤59, 2 ≤δ ≤5)\n\n (a)  moeto waroewo                480\n      ijema joka rati                  39\n     moeto wido ma benomi ka wi  186\n\n (b)    1  ena\n       19  ijema rati = ije mepina\n       26  benomima mepina\n      104  wima mepija daemita moeto\n      292  moeto wi ma wijama rati beo ka idibi", "item_type": "full_problem", "prompt_chars": 1303, "record_id": "iol-2021-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 116, "canonical_excerpt": "(a)  moeto waroewo                480\n      ijema joka rati                  39\n     moeto wido ma benomi ka wi  186", "item_type": "subproblem", "prompt_chars": 1111, "record_id": "iol-2021-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 160, "canonical_excerpt": "(b)    1  ena\n       19  ijema rati = ije mepina\n       26  benomima mepina\n      104  wima mepija daemita moeto\n      292  moeto wi ma wijama rati beo ka idibi", "item_type": "subproblem", "prompt_chars": 1575, "record_id": "iol-2021-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 160, "canonical_excerpt": "(b)    1  ena\n       19  ijema rati = ije mepina\n       26  benomima mepina\n      104  wima mepija daemita moeto\n      292  moeto wi ma wijama rati beo ka idibi", "item_type": "partial_variant", "prompt_chars": 1744, "record_id": "iol-2021-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-sol.en.pdf", "source_problem_id": "2021-individual-1", "target_subparts": ["a", "b", "full"], "title": "Ekari", "year": 2021}
+{"author": "Peter Arkadiev", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Ubykh (Abkhaz–Adyghe)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 7, "answer_unit_ids": ["a.1", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 1307, "canonical_excerpt": "Problem 1.  Sb — Obdir →Obobl:\n                 Obdir   Obobl  Sb\n 1st person sg             s-\n 1st person pl               š’-           [                              ]   \n                            Obdir = pl                 2nd person sg         w-                                                                                              :  a                     + tʷ +     2nd person pl ∈{Obdir, Obobl, Sb}      + n2nd person pl           ŝʷ-                                                                                                                             ə3rd person sg    a-      ∅-     n-                                          otherwise\n3rd person pl        a-  ", "item_type": "full_problem", "prompt_chars": 1714, "record_id": "iol-2022-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 37, "canonical_excerpt": "(a) aŝʷəstʷan — I give them to youpl.", "item_type": "subproblem", "prompt_chars": 918, "record_id": "iol-2022-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 506, "canonical_excerpt": "(b)   1.   ašʼəntʷən   —  he gives him to us\n         2.   səŝʷtʷan   —  youpl give me to him\n         3.  šʼəwənatʷan —  they give us to yousg\n  (c)   4.  they give youpl to me  —  ŝʷəsənatʷan\n         5.  youpl give him to me  —  asəŝʷtʷan\n         6.  yousg give us to him  —  šʼəwtʷan\n         7.  we give yousg to them —  wašʼtʷən\n         8.  he gives them to us   —  ašʼəntʷan\n\nNineteenth International Linguistics Olympiad (2022)                                       2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2590, "record_id": "iol-2022-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 506, "canonical_excerpt": "(b)   1.   ašʼəntʷən   —  he gives him to us\n         2.   səŝʷtʷan   —  youpl give me to him\n         3.  šʼəwənatʷan —  they give us to yousg\n  (c)   4.  they give youpl to me  —  ŝʷəsənatʷan\n         5.  youpl give him to me  —  asəŝʷtʷan\n         6.  yousg give us to him  —  šʼəwtʷan\n         7.  we give yousg to them —  wašʼtʷən\n         8.  he gives them to us   —  ašʼəntʷan\n\nNineteenth International Linguistics Olympiad (2022)                                       2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 2680, "record_id": "iol-2022-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-sol.en.pdf", "source_problem_id": "2022-individual-1", "target_subparts": ["a", "b", "full"], "title": "Ubykh", "year": 2022}
+{"author": "Samuel Ahmed", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Guazacapán Xinka (Xincan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 3, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 14, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10"], "canonical_answer_chars": 1638, "canonical_excerpt": "Problem 1.\n             ⃝       −→\n\n         w- w                             I saw (it)       I see (it)\n\n         6\n         ?            w                             I was seen   I am being seen\n\n     1st person sg ⃝  −→   3rd person sg ⃝  −→\n     w- w      -n’   ɨn-    w- w      -y’ ——\n\n     6             6\n     ?             ?\n       w        ɨn-   ɨn-      w        -∅    a-\n\n   1. −→:\n\n       • -CV →-C’V      (ʂʼ →ʦʼ);\n       • -V1CC’V2- →-V1CVXC’V2-, VX = (V1 = V2) | a\n\n                                                         4.   • n →m |    {p, p’}   6\n   ?    w                                                 • n →ŋ |    {k, k’}   2.            :\n                                   ", "item_type": "full_problem", "prompt_chars": 3853, "record_id": "iol-2023-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 4, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4"], "canonical_answer_chars": 349, "canonical_excerpt": "(a)\n          Form 1   Form 2      Translation\n       15.  netkayʼ    ɨnnetakʼa    push\n       16.   kɨrɨyʼ       ɨŋɡɨrʼɨ        pull\n       17.  pʼuhruyʼ  ɨmpʼuhurʼu  make holes\n       18.  herʼoy     ɨnherʼo      smooth out\n\nTwentieth International Linguistics Olympiad (2023)                                        2\n\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 4583, "record_id": "iol-2023-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 10, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10"], "canonical_answer_chars": 424, "canonical_excerpt": "(b)\n        39.  herʼoŋ        I smoothed (it) out\n        40.   ɨniʂpaʔ       I left\n        41.  aplaʔ            it was opened\n        42.  ʂurʼuy          it chose (it)\n        43.   pirinʼ         I saw (it)\n        44.  ɨnʂawaʦʼa    I plant (it)\n        45.   kʼaaniʔ         it was trapped\n        46.  ɨmpʼɨhanaʔ   I leap\n        47.  ahɨkaʔ          it is being sewn, woven\n        48.   ɨniplaʔ        I was bathed", "item_type": "subproblem", "prompt_chars": 3203, "record_id": "iol-2023-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 10, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10"], "canonical_answer_chars": 424, "canonical_excerpt": "(b)\n        39.  herʼoŋ        I smoothed (it) out\n        40.   ɨniʂpaʔ       I left\n        41.  aplaʔ            it was opened\n        42.  ʂurʼuy          it chose (it)\n        43.   pirinʼ         I saw (it)\n        44.  ɨnʂawaʦʼa    I plant (it)\n        45.   kʼaaniʔ         it was trapped\n        46.  ɨmpʼɨhanaʔ   I leap\n        47.  ahɨkaʔ          it is being sewn, woven\n        48.   ɨniplaʔ        I was bathed", "item_type": "partial_variant", "prompt_chars": 3605, "record_id": "iol-2023-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-sol.en.pdf", "source_problem_id": "2023-individual-1", "target_subparts": ["a", "b", "full"], "title": "Guazacapán Xinka", "year": 2023}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Koryak (Chukotko–Kamchatkan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 17, "answer_unit_ids": ["full.1", "full.2", "full.3", "full.4", "full.5", "full.6", "full.7", "full.8", "full.9", "full.10", "full.11", "full.12", "full.13", "full.14", "full.15", "full.16", "full.17"], "canonical_answer_chars": 1835, "canonical_excerpt": "Problem 1.  Structure of the verb form:\n\n    1. subject (t- 1st person sg, mət- 1st person du/1st person pl, ∅- 2nd person sg, ne- 3rd person\n     du=3rd person pl)\n\n    2. ə materialises here if and only if needed to avoid a cluster of two consonants word-initially or\n      word-finally, or a cluster of three consonants in any position.\n\n    3. tense (ku- present, je- future, ∅- past)\n\n    4. ine-: direct object = 1st person sg (subject = 2nd person du/2nd person pl)\n\n    5. verb root\n\n    6. -la: subject/direct object = 1st person pl/2nd person pl\n\n    7. tense (-ŋ present/future, -∅past)\n\n    8. ə materialises here if and only if needed to avoid a cluster of two consonants word-initially ", "item_type": "full_problem", "prompt_chars": 2248, "record_id": "iol-2024-individual-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-sol.en.pdf", "source_problem_id": "2024-individual-1", "target_subparts": ["full"], "title": "Koryak", "year": 2024}
+{"author": "Vlad A. Neacșu", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Dzongkha (Sino-Tibetan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 4, "answer_unit_ids": ["a.1", "a.2", "b.1", "c.1"], "canonical_answer_chars": 1509, "canonical_excerpt": "Problem 1.\n\n                         1       2     3      4     5      6      7      8     9\n     k   k                  ci       ɲi   sum    ʑi    ŋa    ɖu    dyn   ge    gu\n          10( + k)        cu-     cu-   cu-    cy-   ce-    cu-    cup-   cop-   cy-\n   k k × 10        cutãm   ɲi-   sum-   ʑip-  ŋap-  ɖuk-  dyn-  gep-   (gup-)     k\n   k     k   k × 10( + . . .)            tsa-   so-     ʑe-   ŋa-     (re-)   døn-   ɟa-     (go-)\n     • System A:\n\n       – β × 20 ( + α) = ke β (da α )\n\n        ∗β × 20 + 10 →ke pɟe-da β + 1\n\n        ∗β × 20 + 15 →ke ko-da β + 1\n\n       – γ × 400 ( + β × 20 + α) = ɲiɕu γ (da ke β da α )\n\n        ∗γ × 400 + 200 →ɲiɕu pɟe-da γ + 1\n        ∗γ × 400 + 300", "item_type": "full_problem", "prompt_chars": 2632, "record_id": "iol-2025-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 106, "canonical_excerpt": "(a)      [X] = 60 = ɖukcu\n           [Y] = 62 = ke sum da ɲi\n           [Z] = 885 = ɲiɕu ɲi da ke ʑi da ŋa", "item_type": "subproblem", "prompt_chars": 2473, "record_id": "iol-2025-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 309, "canonical_excerpt": "(b)   (1) 13 + 70 = 83\n         (2) 800 = 20 × 40\n         (3) 469 = 50 × 9 + 19\n         (4) 600 + 110 = 500 + 210\n         (5) 2 × 43 + 21 = 2\n         (6) 1100 × 12 + 50 = 600\n         (7) 736 = 84 × 4 + 400\n         (8) 2 × 609 = 60X × 20 + 18\n         (9) 62Y + 24 = 86\n       (10) 885Z + 115 = 700 + 300", "item_type": "subproblem", "prompt_chars": 2235, "record_id": "iol-2025-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 220, "canonical_excerpt": "(c)     • 75 = ke ko-da ʑi = dønŋa\n           • 570 = ɲiɕu ci da ke pɟe-da gu = ŋapɟa dyncu\n\nTwenty-second International Linguistics Olympiad (2025)                                          2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2587, "record_id": "iol-2025-individual-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 309, "canonical_excerpt": "(b)   (1) 13 + 70 = 83\n         (2) 800 = 20 × 40\n         (3) 469 = 50 × 9 + 19\n         (4) 600 + 110 = 500 + 210\n         (5) 2 × 43 + 21 = 2\n         (6) 1100 × 12 + 50 = 600\n         (7) 736 = 84 × 4 + 400\n         (8) 2 × 609 = 60X × 20 + 18\n         (9) 62Y + 24 = 86\n       (10) 885Z + 115 = 700 + 300", "item_type": "partial_variant", "prompt_chars": 2394, "record_id": "iol-2025-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 220, "canonical_excerpt": "(c)     • 75 = ke ko-da ʑi = dønŋa\n           • 570 = ɲiɕu ci da ke pɟe-da gu = ŋapɟa dyncu\n\nTwenty-second International Linguistics Olympiad (2025)                                          2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3083, "record_id": "iol-2025-individual-p1-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-sol.en.pdf", "source_problem_id": "2025-individual-1", "target_subparts": ["a", "b", "c", "full"], "title": "Dzongkha", "year": 2025}

benchmark/IOL/ioling_hf/reports/manual_verification/batch_06.jsonl

@@ -0,0 +1,21 @@
+{"author": "Svetlana Burlak", "curation_statuses": {"auto_extracted_needs_manual_review": 1}, "languages": ["Tocharian A (Indo-European) Tocharian B (Indo-Eurorean)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2003-team-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2003-team-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1383, "canonical_excerpt": "Solution of Problem 1\nAssignment 1. A B A B A B\nstank stank aknats aknatsa pratsak — pratsako\nastar astare, astre | kramarts kramartse | raskdr —_ raskare\nwal walo, wlo sakar sakre sam sam\nasar —asare tsdnkar tsankar ysar ysar, yasar\nThe first pair gives the correspondence st—st. This determines unambiguously the second pair\n(or triple, rather), whence we learn that Tocharian B has kept the final vowels (except for the\n‘specific’ one) and Tocharian A has lost them. Consequently all words with retained final vowels\nare Tocharian B and their counterparts with lost final vowels are Tocharian A. This allows the\nfollowing conclusions to be made: In Tocharian A the ‘specific’ vowel falls out befor", "item_type": "full_problem", "prompt_chars": 1612, "record_id": "iol-2003-team-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "team", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2003-team-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2003-team-sol.en.pdf", "source_problem_id": "2003-team-1", "target_subparts": ["full"], "title": "Tocharian", "year": 2003}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_needs_manual_review": 6}, "languages": ["Lakhota (Siouxan)"], "problem_extraction_confidence": "heading", "problem_number": 4, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_2.1", "assignment_3.1"], "canonical_answer_chars": 2122, "canonical_excerpt": "Problem No.4 : :\n\nIn the Lakhota words that appear in the problem and the assignments there are pairs and\ntriples of words differing in fricative consonants, which may be sibilants (s, z), hushes (5, 2), or\nvelars (x, y), with voiceless consonants correspoding to their voiceless variants, and voiced\nconsonants correspoding to their voiced variants. These data may be summarized in a table:\n\nwords with sibilant fricatives | words with hushing fricatives words with velar fricatives\na <n\n‘a single high-pitched tone ‘a blending high-pitched tone\nsounds’ sounds’\n2\n‘it (e.g. a road) forks into two\nparts’\npo itis badly bruised”\n\n|____ ‘itis yellow™ itis brown?\n1 7\n‘it is curled but can be ‘it is shr", "item_type": "full_problem", "prompt_chars": 2108, "record_id": "iol-2004-individual-p4", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 414, "canonical_excerpt": "Assignment 1.\n\n¢ ‘it has a slight bruise’ and ‘it is fractured’ are similar to ‘it is badly bruised’ (8uza), but\n\nthe damage is less serious in the former and more serious in the latter (suza and xuya,\nrespectively).\n\n¢ ‘surface is in a scratched condition’ and ‘the surface is in a scraped condition’ make a\n\npair (k'eza and k'eya, respectively).\n\n¢ ‘it sparks’ is translated by process of elimination (p\"eSniza).", "item_type": "subproblem", "prompt_chars": 1595, "record_id": "iol-2004-individual-p4-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 355, "canonical_excerpt": "Assignment 2. ‘a thin liquid is being squeezed out’ is s/i; ‘it is soft and movable (e.g. an\nenlarged gland under the skin)’ and ‘it is semi-hard and movable (e.g. a cartilage)’ are nuza\nand nuza, respectively; ‘it is red hot’ is p*exniya, ‘it is branching into several directions’ is yata\n(there is more branching than if something forks into two parts).", "item_type": "subproblem", "prompt_chars": 1711, "record_id": "iol-2004-individual-p4-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 60, "canonical_excerpt": "Assignment 3. ‘it is brown and yellow (dark yellow, tawny)’.", "item_type": "subproblem", "prompt_chars": 2071, "record_id": "iol-2004-individual-p4-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 355, "canonical_excerpt": "Assignment 2. ‘a thin liquid is being squeezed out’ is s/i; ‘it is soft and movable (e.g. an\nenlarged gland under the skin)’ and ‘it is semi-hard and movable (e.g. a cartilage)’ are nuza\nand nuza, respectively; ‘it is red hot’ is p*exniya, ‘it is branching into several directions’ is yata\n(there is more branching than if something forks into two parts).", "item_type": "partial_variant", "prompt_chars": 2189, "record_id": "iol-2004-individual-p4-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 60, "canonical_excerpt": "Assignment 3. ‘it is brown and yellow (dark yellow, tawny)’.", "item_type": "partial_variant", "prompt_chars": 2943, "record_id": "iol-2004-individual-p4-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2004-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2004-indiv-sol.en.pdf", "source_problem_id": "2004-individual-4", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Lakhota", "year": 2004}
+{"author": "Alexander Lubotsky", "curation_statuses": {"auto_extracted_needs_manual_review": 6}, "languages": ["Lithuanian (Indo-European)"], "problem_extraction_confidence": "ocr_heading", "problem_number": 5, "problem_page_end": 4, "problem_page_start": 4, "problem_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["assignment_1.1", "assignment_2.1", "assignment_3.1"], "canonical_answer_chars": 2592, "canonical_excerpt": "Problem 5\nFirst of all, for every syllable we must determine to which of the two types it belongs. This\nis easy to do, since, according to the problem statement, the syllable type always remains the\nsame within the same root or the same ending. Hence, the root has falling intonation in\nparadigms 1 and 3, and rising intonation in paradigms 2 and 4. The endings in Nom.Sg. and\nAcc.PI. always have falling intonation, whereas in Gen.Sg. and Nom.PI. they always have rising\nintonation (the last one is explicated in the problem).\n\nLet us represent these data in a table (where a designates any vowel, and stressed\nsyllables are set in boldface):\n\nParadigm 1 2 3 4\nNom.Sg. da aa aa aa\nGen.Sg. aa aa aa a", "item_type": "full_problem", "prompt_chars": 2604, "record_id": "iol-2005-individual-p5", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_1.1"], "canonical_answer_chars": 454, "canonical_excerpt": "Assignment 1. Paradigms 1 and 2, on the one hand, and paradigms 3 and 4, on the\nother hand, originally belonged together.\n\nIn order to determine what the two initial paradigms looked like, we have to answer the\nquestion why dd and dd have different places of accent. Maybe da changed to ad? But we can\nsee the sequence dda in 3 (Nom.Sg.), and ad does not occur in any of the paradigms 1-4.\nTherefore ad always changed to aa, and not the other way around.", "item_type": "subproblem", "prompt_chars": 1904, "record_id": "iol-2005-individual-p5-sub-assignment_1", "target_subparts": ["assignment_1"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 212, "canonical_excerpt": "Assignment 2. Paradigm 1 + 2 looked as 1 looks now (the root was always accented),\nand the paradigm 3 + 4 looked as 3 looks now (the endings were accented in the Singular and\nthe root was accented in the Plural).", "item_type": "subproblem", "prompt_chars": 1892, "record_id": "iol-2005-individual-p5-sub-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 193, "canonical_excerpt": "Assignment 3. Saussure’s Law says that in the sequence syllable with rising intonation —\nsyllable with falling intonation (ad) the accent shifted from the first syllable to the second one\n(aa).", "item_type": "subproblem", "prompt_chars": 3262, "record_id": "iol-2005-individual-p5-sub-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_2.1"], "canonical_answer_chars": 212, "canonical_excerpt": "Assignment 2. Paradigm 1 + 2 looked as 1 looks now (the root was always accented),\nand the paradigm 3 + 4 looked as 3 looks now (the endings were accented in the Singular and\nthe root was accented in the Plural).", "item_type": "partial_variant", "prompt_chars": 2410, "record_id": "iol-2005-individual-p5-partial-through-assignment_1-target-assignment_2", "target_subparts": ["assignment_2"], "visible_solved_subparts": ["assignment_1"]}, {"answer_unit_count": 1, "answer_unit_ids": ["assignment_3.1"], "canonical_answer_chars": 193, "canonical_excerpt": "Assignment 3. Saussure’s Law says that in the sequence syllable with rising intonation —\nsyllable with falling intonation (ad) the accent shifted from the first syllable to the second one\n(aa).", "item_type": "partial_variant", "prompt_chars": 4031, "record_id": "iol-2005-individual-p5-partial-through-assignment_1assignment_2-target-assignment_3", "target_subparts": ["assignment_3"], "visible_solved_subparts": ["assignment_1", "assignment_2"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "ocr_heading", "solution_page_end": 4, "solution_page_start": 4, "solution_pdf_local_path": "data/raw/pdfs/iol-2005-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2005-indiv-sol.en.pdf", "source_problem_id": "2005-individual-5", "target_subparts": ["assignment_1", "assignment_2", "assignment_3", "full"], "title": "Lithuanian", "year": 2005}
+{"author": "Alexander Berdichevsky", "curation_statuses": {"auto_extracted_needs_manual_review": 1}, "languages": ["English (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 2334, "canonical_excerpt": "Fifth International Olympiad in Theoretical, Mathematical\n                  and Applied Linguistics\n                  Russia, St Petersburg, 31 July–4 August 2007\n               Solutions of the Problems of the Individual Contest\n\n                                 Problem #2\n\nThe negative forms are composed of a particle kas followed by a modification of the original\nform which contains the marker -(k)a’- in one form or another. The rules for insertion of this\nmarker are as follows:\n\n    (1) The marker is inserted after the first syllable of the word if this syllable is either closed\n          (i.e., ends in a consonant) or long (i.e., contains a long vowel); otherwise the marker is\n        i", "item_type": "full_problem", "prompt_chars": 1267, "record_id": "iol-2007-individual-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "page_index", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2007-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2007-indiv-sol.en.pdf", "source_problem_id": "2007-individual-1", "target_subparts": ["full"], "title": "Braille", "year": 2007}
+{"author": "Bozhidar Bozhanov", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Micmac (Algonquian)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 824, "canonical_excerpt": "Problem #1.  Rules:\n\n   1. The apostrophe indicates length if it follows a vowel, and is read as [@] if it follows a consonant.\n\n   2. The letter w stands for a rounding of the lips after a consonant and for the sound [w]\n      otherwise.\n\n   3. [@] is pronounced, though not written, between any consonant and a following sonorant\n     consonant ([l m n]).\n\n   4. [@] is also pronounced before a consonant cluster at the beginning of a word.\n\n   5. p t j g gw q qw are pronounced as voiced consonants ([b d j g gw G Gw]) at the beginning\n      of a word or between vowels and as voiceless consonants ([p t c k kw x xw]) at the end of a\n     word or next to another consonant.\n\nAnswers:\n\n (a) 12 [@ks", "item_type": "full_problem", "prompt_chars": 2148, "record_id": "iol-2008-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 82, "canonical_excerpt": "(a) 12 [@ks@nx¯oGon], 13 [@tk@box], 14 [g@m¯uj@min], 15 [emtoGwatk], 16 [d¯eb@lc];", "item_type": "subproblem", "prompt_chars": 1733, "record_id": "iol-2008-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 50, "canonical_excerpt": "(b) 17 tp’te’sn, 18 mtesgm, 19 alapt’g, 20 glamen.", "item_type": "subproblem", "prompt_chars": 2643, "record_id": "iol-2008-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 50, "canonical_excerpt": "(b) 17 tp’te’sn, 18 mtesgm, 19 alapt’g, 20 glamen.", "item_type": "partial_variant", "prompt_chars": 2778, "record_id": "iol-2008-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2008-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2008-indiv-sol.en.pdf", "source_problem_id": "2008-individual-1", "target_subparts": ["a", "b", "full"], "title": "Micmac", "year": 2008}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Sulka (East Papuan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-prob.en-gb.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 853, "canonical_excerpt": "Problem #1.  Here are the words from which the Sulka language constructs its numerals:\n   • tgiang 1, lomin 2, korlotge 3, korlolo 4, kti¨ek 5, mhelom 20;\n   • hori orom addition, lo doubling;\n   • a singular, o plural (from 3 on).\nNouns have different forms for the two numbers (tu, sngu; vhoi, vuo). There are separate words\nfor a foursome of coconuts, for a twosome and foursome of breadfruit (ngausmia, moulang,\nngaitegaap).\n   Answers:\n (a)   • a ksie a tgiang: 1 coconut\n       • o ngaitegaap a korlotge: 12 breadfruits\n       • o ngausmia a kti¨ek: 20 coconuts\n       • o vuo a lo kti¨ek hori orom a tgiang: 11 betel nuts\n (b)   • 2 yams: a lo tu a lomin\n       • 14 yams: o sngu a lo kti¨ek h", "item_type": "full_problem", "prompt_chars": 1888, "record_id": "iol-2009-individual-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2009-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2009-indiv-sol.en-gb.pdf", "source_problem_id": "2009-individual-1", "target_subparts": ["full"], "title": "Sulka", "year": 2009}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Budukh (Northeast Causasian)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-prob.en.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1134, "canonical_excerpt": "Problem #1.  Rules:\n   • form 1: -mV- after the first vowel, whereby V depends on the vowel in the following      syllable (a before a, o before o or u, e before i, ö before ü);\n   • form 2:\n      – -a, if the stem ends in -aR or -oR,\n      – -Ra, if the stem ends in -i, -u or -ü,\n\n     where R is l or n if one of these consonants is found in the root, or r otherwise;\n   • form 3: form 2 with -r- after the first vowel, unless R follows immediately.\n   Answers:\n                                                 form 1      form 2    form 3\n form 1       form 2     form 3                       amol˙qol      al˙qola      al˙qola\n h˛amerki      h˛arkira                             emensi     ensina\n", "item_type": "full_problem", "prompt_chars": 2224, "record_id": "iol-2010-individual-p1", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2010-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2010-indiv-sol.en.pdf", "source_problem_id": "2010-individual-1", "target_subparts": ["full"], "title": "Budukh", "year": 2010}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Menominee (Algonquin)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-prob.en-gb.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 2, "answer_unit_ids": ["a.1", "b.1"], "canonical_answer_chars": 1692, "canonical_excerpt": "Problem #1.  The Menominee verb forms have the following structure:\n\n                                                   intransitive verb:\n             kaw   down                                   -¯ahpe      laughing                                        -m      I               ket    out                              -¯anæhkæ  digging                                                       -q   we1+2                  k¯esk   through                                 -eqta   ——                                             -w    he ne-        I    pahk  off                           -ohnæ     walking\n ke-    we1+2   p¯ahk  open\n                                                    transitive verb:——", "item_type": "full_problem", "prompt_chars": 2191, "record_id": "iol-2011-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 280, "canonical_excerpt": "(a)   • kek¯eskahæq: we1+2 chop it through, break it through by tool\n       • nep¯ahkenan:\n          – I open, uncover it by hand (  p¯ahk),                     !\n          – I break it off, tear it offby hand (  pahk)                       !\n       • w¯æp¯ahpew: he begins laughing", "item_type": "subproblem", "prompt_chars": 1676, "record_id": "iol-2011-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 154, "canonical_excerpt": "(b)   • I begin to eat it: new¯æpahtan\n       • we1+2 lay it flat by hand: kek¯awenæq\n       • he digs a hole: taw¯anæhkæw\n       • he walks out: ket¯ohnæw", "item_type": "subproblem", "prompt_chars": 2786, "record_id": "iol-2011-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 154, "canonical_excerpt": "(b)   • I begin to eat it: new¯æpahtan\n       • we1+2 lay it flat by hand: kek¯awenæq\n       • he digs a hole: taw¯anæhkæw\n       • he walks out: ket¯ohnæw", "item_type": "partial_variant", "prompt_chars": 3119, "record_id": "iol-2011-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2011-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2011-indiv-sol.en-gb.pdf", "source_problem_id": "2011-individual-1", "target_subparts": ["a", "b", "full"], "title": "Menominee", "year": 2011}
+{"author": "ArtÅ«rs Semeņuks", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Dyirbal (Pama-Nyungan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 3, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 6, "answer_unit_ids": ["a.1", "b.1", "b.2", "c.1", "c.2", "c.3"], "canonical_answer_chars": 2068, "canonical_excerpt": "Problem #1.  The word order is OSV (O: object, S: subject, V: verb), NA (N: noun, A:\nadjective).\n  A →V (‘make A’): A-man.\n  V →A: V     ‘that is always being Ved’   ‘that is always Ving Ns’\n           -n   -l-muNa                    N-. . . -l-Nay-muNa\n          -ñu  -y-muNa                   N-. . . -nay-muNa\n   Every noun is preceded by an article:\n O      S\n balan  baNgun  women, dangerous animals and objects\n bayi   baNgul   men, animals\n bala   baNgu     all other things\n\n   The subject and its attributes get the ending\n   • -Ngu, if the word ends in a vowel and has two syllables;\n   • -gu, if the word ends in a vowel and has more than two syllables;\n   • -Du, if the word ends in a con", "item_type": "full_problem", "prompt_chars": 3859, "record_id": "iol-2012-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 217, "canonical_excerpt": "(a) The grasshopper is neither a woman nor a dangerous animal, but takes the same article,\n      so it must be the “old woman” from the myth. The linguist thought that baNgun\n    bundiñéu in example (14) was an error.", "item_type": "subproblem", "prompt_chars": 2922, "record_id": "iol-2012-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 417, "canonical_excerpt": "(b)  17.  balan ñalNga baNgul NumaNgu guniymuNagu bambunman.\n         The father that is always being searched for is healing the girl.\n       18.  bala diban bilmbalmuNa baNgun biñéiriñéu guniñu.\n         The lizard is searching for the stone that is always being pushed.\n       19.  bayi bargan baNgul yaóaNgu gubimbuluNunéanaymuNagu banéan.\n         The man that is always blaming doctors is following the wallaby.", "item_type": "subproblem", "prompt_chars": 2680, "record_id": "iol-2012-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 629, "canonical_excerpt": "(c)  20.  The little wallaby is looking at the dragonfly.\n          bayi yiriñéila baNgul bargandu wuraNgu buóan.\n       21.  The aunt that is always being followed is bending the feather.\n          bala yila baNgun mugunanéagu banéalmuNagu waóuman.\n       22.  The sleeping possum is ignoring the loud noise.\n          bala muNga baNgul midindu éagundu Naéin.\n       23.  The caterpillar is searching for the man that is always carrying stones.\n          bayi yaóa dibandimbanaymuNa baNgul bayimbambu guniñu.\n\nTenth International Olympiad in Linguistics (2012).                                      2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 4430, "record_id": "iol-2012-individual-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["b.1", "b.2"], "canonical_answer_chars": 417, "canonical_excerpt": "(b)  17.  balan ñalNga baNgul NumaNgu guniymuNagu bambunman.\n         The father that is always being searched for is healing the girl.\n       18.  bala diban bilmbalmuNa baNgun biñéiriñéu guniñu.\n         The lizard is searching for the stone that is always being pushed.\n       19.  bayi bargan baNgul yaóaNgu gubimbuluNunéanaymuNagu banéan.\n         The man that is always blaming doctors is following the wallaby.", "item_type": "partial_variant", "prompt_chars": 2950, "record_id": "iol-2012-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 3, "answer_unit_ids": ["c.1", "c.2", "c.3"], "canonical_answer_chars": 629, "canonical_excerpt": "(c)  20.  The little wallaby is looking at the dragonfly.\n          bayi yiriñéila baNgul bargandu wuraNgu buóan.\n       21.  The aunt that is always being followed is bending the feather.\n          bala yila baNgun mugunanéagu banéalmuNagu waóuman.\n       22.  The sleeping possum is ignoring the loud noise.\n          bala muNga baNgul midindu éagundu Naéin.\n       23.  The caterpillar is searching for the man that is always carrying stones.\n          bayi yaóa dibandimbanaymuNa baNgul bayimbambu guniñu.\n\nTenth International Olympiad in Linguistics (2012).                                      2\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 5145, "record_id": "iol-2012-individual-p1-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2012-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2012-indiv-sol.en.pdf", "source_problem_id": "2012-individual-1", "target_subparts": ["a", "b", "c", "full"], "title": "Dyirbal", "year": 2012}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Yidiny (Pama-Nyungan)"], "problem_extraction_confidence": "heading", "problem_number": 1, "problem_page_end": 2, "problem_page_start": 1, "problem_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 997, "canonical_excerpt": "Problem #1.  Rules:\n\n   1.  If the number of syllables in the word (= stem + ending) is even, all syllables are short.\n        If the number of syllables in the word is odd, the last even-numbered syllable of the\n     stem is lengthened.\n      2 = 2 + 0:         bu1ña2-∅          3 = 2 + 1:        ba1ma:2   -gu3\n      4 = 2 + 2:         bu1ña2-gi3mbal4   3 = 3 + 0:         gu1da:2ga3-∅\n      4 = 3 + 1:      gu1da2ga3-bi4         5 = 3 + 2:          éu1du:2lu3 -mu4éay5\n      4 = 4 + 0:  mu1yu2ba3ra4-∅          5 = 4 + 1:  ga1éa2gi3mba:4   -gu5\n\n   2.  If the ending -ni or -mu immediately follows a long vowel, it loses its own vowel.\n\n (a) mugaóumu, waNa:lgu.\n\n (b) bama:n — of a person, buña:b", "item_type": "full_problem", "prompt_chars": 2110, "record_id": "iol-2013-individual-p1", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 23, "canonical_excerpt": "(a) mugaóumu, waNa:lgu.", "item_type": "subproblem", "prompt_chars": 1586, "record_id": "iol-2013-individual-p1-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 104, "canonical_excerpt": "(b) bama:n — of a person, buña:bi — another woman, maéurmuéay — with a frog,\n    muéa:mni — of a mother.", "item_type": "subproblem", "prompt_chars": 1632, "record_id": "iol-2013-individual-p1-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 222, "canonical_excerpt": "(c) of a stranger — muyubara:n, for a fishing net — mugaóugu, father — bimbi, from\n     a frog — maéu:rmu, without a man — wagu:éagimbal, of a pigeon — éuduluni,\n      tortoise — baéi:gal, without a boomerang — waNalgimbal.", "item_type": "subproblem", "prompt_chars": 2482, "record_id": "iol-2013-individual-p1-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 104, "canonical_excerpt": "(b) bama:n — of a person, buña:bi — another woman, maéurmuéay — with a frog,\n    muéa:mni — of a mother.", "item_type": "partial_variant", "prompt_chars": 1708, "record_id": "iol-2013-individual-p1-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 222, "canonical_excerpt": "(c) of a stranger — muyubara:n, for a fishing net — mugaóugu, father — bimbi, from\n     a frog — maéu:rmu, without a man — wagu:éagimbal, of a pigeon — éuduluni,\n      tortoise — baéi:gal, without a boomerang — waNalgimbal.", "item_type": "partial_variant", "prompt_chars": 2690, "record_id": "iol-2013-individual-p1-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 1, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2013-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2013-indiv-sol.en-gb.pdf", "source_problem_id": "2013-individual-1", "target_subparts": ["a", "b", "c", "full"], "title": "Yidiny", "year": 2013}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 1}, "languages": ["Kiowa (Tanoan)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-prob.en-gb.pdf", "record_count": 1, "record_types": {"full_problem": 1}, "records": [{"answer_unit_count": 1, "answer_unit_ids": ["full.1"], "canonical_answer_chars": 1251, "canonical_excerpt": "Problem #2.  Rules:\n                                                   singular  dual   plural\n                         people, animals, tools         -∅         -gO\n                                                -n         -dO\n                             fruit                   -gO      -∅    -gO\n                                            -dO     -n    -dO\n                                all other things         -dO        -(n)\n   Answers:\n                      singular       dual       plural\n                      alOsOhjegO     alOsOhje   alOsOhjegO    plum\n                 tsegun       tsegun     tsegudO      dog\n                    alOguk’ogO     alOguk’o   alOguk’ogO  lemon\n        ", "item_type": "full_problem", "prompt_chars": 2224, "record_id": "iol-2014-individual-p2", "target_subparts": [], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2014-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2014-indiv-sol.en-gb.pdf", "source_problem_id": "2014-individual-2", "target_subparts": ["full"], "title": "Kiowa", "year": 2014}
+{"author": "Peter Arkadiev", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Kabardian (North Caucasian)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-prob.en-gb.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-prob.en-gb.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["a.1", "b.1", "c.1"], "canonical_answer_chars": 1530, "canonical_excerpt": "Problem 2.  Structure of the verb form:\n\n    I.    – me-: affirmative form, present, indicative mood,\n\n       – ,\n\n       – -pe ‘really’, -fe ‘pretend to’, -f ‘be able to’, -n — infinitive.\n\n      In this part of the word:\n\n          1. C + -C > CəC (de + -f + -n > de-f-ə-n, me- + bəb + -pe > me-bəb-ə-pe).\n\n          2. The last syllable receives the stress if it is closed, otherwise the penultimate is\n            stressed (defən > defə́n, mešxepe > mešxépe).\n\n          3. CéC(C)e > CáC(C)e (méšxe > mášxe, mešxépe > mešxápe).\n\n   II.  -xe — plural, -t — past, -me — conditional mood, -q̇əm — negative form.\n\n   Answers:\n\n (a)  ʒeq̇én         to bite\n     medéf        (he/she) is able to sew\n ", "item_type": "full_problem", "prompt_chars": 2223, "record_id": "iol-2015-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 236, "canonical_excerpt": "(a)  ʒeq̇én         to bite\n     medéf        (he/she) is able to sew\n     medáfe       (he/she) is pretending to sew\n       səfə́n          to be able to burn\n      megʷəš’ə́ʔe  (he/she) is speaking\n      mebə́b       (he/she) is flying", "item_type": "subproblem", "prompt_chars": 2089, "record_id": "iol-2015-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 386, "canonical_excerpt": "(b)  c̣entʷéfme           if (he/she) is able to slide\n      šxáfexeq̇əm       (they) aren’t pretending to eat\n       bəbə́ft             (he/she) was able to fly\n      šxet               (he/she) was eating\n      ṭəgʷərəgʷə́peme    if (he/she) really is trembling\n\nThirteenth International Olympiad in Linguistics (2015).                                  3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 981, "record_id": "iol-2015-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 273, "canonical_excerpt": "(c)  mádexe         (they) are sewing\n      mebəbə́fexe    (they) are pretending to fly\n       sə́pet           (he/she) really was burning\n      šxéfq̇əm        (he/she) isn’t able to eat\n      gʷəš’ə́ʔexeme    if (they) are speaking\n      meʒáq̇exe       (they) are biting", "item_type": "subproblem", "prompt_chars": 2239, "record_id": "iol-2015-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 386, "canonical_excerpt": "(b)  c̣entʷéfme           if (he/she) is able to slide\n      šxáfexeq̇əm       (they) aren’t pretending to eat\n       bəbə́ft             (he/she) was able to fly\n      šxet               (he/she) was eating\n      ṭəgʷərəgʷə́peme    if (he/she) really is trembling\n\nThirteenth International Olympiad in Linguistics (2015).                                  3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 1270, "record_id": "iol-2015-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["c.1"], "canonical_answer_chars": 273, "canonical_excerpt": "(c)  mádexe         (they) are sewing\n      mebəbə́fexe    (they) are pretending to fly\n       sə́pet           (he/she) really was burning\n      šxéfq̇əm        (he/she) isn’t able to eat\n      gʷəš’ə́ʔexeme    if (they) are speaking\n      meʒáq̇exe       (they) are biting", "item_type": "partial_variant", "prompt_chars": 2942, "record_id": "iol-2015-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": false, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2015-indiv-sol.en-gb.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2015-indiv-sol.en-gb.pdf", "source_problem_id": "2015-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Kabardian", "year": 2015}
+{"author": "Tae Hun Lee", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Luwian (Indo-European)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 20, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "b.1", "c.1", "c.2", "c.3", "c.4", "c.5", "c.6"], "canonical_answer_chars": 2272, "canonical_excerpt": "Problem 2.  The direction of writing is from left to right. Words can be written in three\nways:\n\n   • with a logogram;\n\n   • with syllabograms;\n\n   • with a logogram and syllabograms which partially or fully reflect the sound of the word.\n\n  n before a consonant is never expressed by a syllabogram.\n\n (a)\n         1.  runtiyas     B. +                  H. „0s\n              ‘deer’                                -ya-s(a)\n         2.  patis        E. Ts              U. T\n              ‘foot’             -s(a)              \n         3.  harnisas      J. H!             O. !fss\n               ‘fortress’          ha-               -ni-s(a)-s(a)\n         4.  iziyanta   ", "item_type": "full_problem", "prompt_chars": 2010, "record_id": "iol-2016-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 13, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13"], "canonical_answer_chars": 1600, "canonical_excerpt": "(a)\n         1.  runtiyas     B. +                  H. „0s\n              ‘deer’                                -ya-s(a)\n         2.  patis        E. Ts              U. T\n              ‘foot’             -s(a)              \n         3.  harnisas      J. H!             O. !fss\n               ‘fortress’          ha-               -ni-s(a)-s(a)\n         4.  iziyanta      F. A50´          T. A50t\n            ‘they made’         i-zi-ya-ta                      i-zi-ya-ta\n         5.  turpis        L. v              W. vås\n             ‘bread’                              -pi-s(a)\n         6.  tarhunzas    S. y2is        V. ´ys\n            ‘thunder’        ", "item_type": "subproblem", "prompt_chars": 1638, "record_id": "iol-2016-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 185, "canonical_excerpt": "(b) 1. tyis — tarhunzas ‘thunder’;     2. f4_ — nimuwizas ‘son’.\n\nFourteenth International Linguistics Olympiad (2016).                                    3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1614, "record_id": "iol-2016-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5", "c.6"], "canonical_answer_chars": 182, "canonical_excerpt": "(c)    1. ´ — ta;\n         2. % — (hantawatis);\n         3. W — wa, wi;\n         4. t — ta;\n         5. s — s(a);\n         6. T — ti, (patis);\n         7. 5 — zi, (zitis).", "item_type": "subproblem", "prompt_chars": 2156, "record_id": "iol-2016-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 185, "canonical_excerpt": "(b) 1. tyis — tarhunzas ‘thunder’;     2. f4_ — nimuwizas ‘son’.\n\nFourteenth International Linguistics Olympiad (2016).                                    3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 3267, "record_id": "iol-2016-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 6, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5", "c.6"], "canonical_answer_chars": 182, "canonical_excerpt": "(c)    1. ´ — ta;\n         2. % — (hantawatis);\n         3. W — wa, wi;\n         4. t — ta;\n         5. s — s(a);\n         6. T — ti, (patis);\n         7. 5 — zi, (zitis).", "item_type": "partial_variant", "prompt_chars": 4022, "record_id": "iol-2016-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2016-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2016-indiv-sol.en.pdf", "source_problem_id": "2016-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Luwian", "year": 2016}
+{"author": "Aleksejs PeguÅ¡evs", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Abui (Timor-Alor-Pantar)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 22, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "b.1", "c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 1957, "canonical_excerpt": "Problem 2.  Word order: ‹possessor› ‹possessed›.\n    Possessive prefixes:\n\n         possessor             possessed\n   1st person sg     n-\n  2nd person sg     ∅-\n                           a-  body part\n  3rd person sg     h-\n                            e-  otherwise\n  ‘… of each of us’   t-\n   ‘one’s own …’     d-\n    2nd person pl ri-\n\nAnswers:\n\n (a)    1.  a-bang                    e.  your (sg.) shoulder\n         2.  a-táng he-ya          p.  your (sg.) thumb [‘your (sg.) hand’s mother’]\n         3.  bataa ha-wata         o.  upper part of a tree [‘tree’s neck’]\n         4.  de-kafi                 d.  one’s own rope\n         5.  e-bataa ha-táng      b.  your (sg.) branch [‘your (sg.) tre", "item_type": "full_problem", "prompt_chars": 2535, "record_id": "iol-2017-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 16, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16"], "canonical_answer_chars": 1175, "canonical_excerpt": "(a)    1.  a-bang                    e.  your (sg.) shoulder\n         2.  a-táng he-ya          p.  your (sg.) thumb [‘your (sg.) hand’s mother’]\n         3.  bataa ha-wata         o.  upper part of a tree [‘tree’s neck’]\n         4.  de-kafi                 d.  one’s own rope\n         5.  e-bataa ha-táng      b.  your (sg.) branch [‘your (sg.) tree’s hand’]\n         6.  e-kuda ha-wata           i.  your (sg.) horse’s neck\n         7.  falepak ha-wei            j.   trigger [‘pistol’s ear’]\n         8.  ha-táng ha-min       a.   his fingertip [‘his hand’s nose’]\n         9.  he-lui             m.   his knife\n       10.  maama he-falepak   h.   father’s pistol\n       11.  na-pong               ", "item_type": "subproblem", "prompt_chars": 2064, "record_id": "iol-2017-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 75, "canonical_excerpt": "(b) 1. a-min — your (sg.) nose; 2. de-ya he-bataa — one’s own mother’s tree", "item_type": "subproblem", "prompt_chars": 2086, "record_id": "iol-2017-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 350, "canonical_excerpt": "(c)    1. pig — fe\n\n         2. your (pl.) knife — ri-lui\n\n         3. your (sg.) mother’s father — e-ya he-maama\n\n         4. my father’s face — ne-maama ha-pong\n\n         5. one’s own ear — da-wei\n\n         6. my sea — ne-tama\n\nFifteenth International Linguistics Olympiad (2017).                                      3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2974, "record_id": "iol-2017-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 75, "canonical_excerpt": "(b) 1. a-min — your (sg.) nose; 2. de-ya he-bataa — one’s own mother’s tree", "item_type": "partial_variant", "prompt_chars": 3314, "record_id": "iol-2017-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 5, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 350, "canonical_excerpt": "(c)    1. pig — fe\n\n         2. your (pl.) knife — ri-lui\n\n         3. your (sg.) mother’s father — e-ya he-maama\n\n         4. my father’s face — ne-maama ha-pong\n\n         5. one’s own ear — da-wei\n\n         6. my sea — ne-tama\n\nFifteenth International Linguistics Olympiad (2017).                                      3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 4305, "record_id": "iol-2017-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2017-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2017-indiv-sol.en.pdf", "source_problem_id": "2017-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Abui", "year": 2017}
+{"author": "Peter Arkadiev", "curation_statuses": {"auto_extracted_text_high_confidence": 2}, "languages": ["Hakhun (Sino-Tibetan)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-prob.en.pdf", "record_count": 2, "record_types": {"full_problem": 1, "subproblem": 1}, "records": [{"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 2109, "canonical_excerpt": "Problem 2.  The sentences have the following structure:\n            a                     —                -bə    (a = 1st person sg, 2nd person sg)    ⎫ o s ne,            a                             P                                     ⎪⎬          , kəmə  (a = 3rd person sg, plural)\n    a: subject, P: object, o: predicate,                                     ⎪⎭                                                                                                                  :  -ɤ                                                  1st person sg ∈{a, P}        1st > 2nd > 3rd  past   present        1st person pl               :   -i                         ⎧           ∈{a, P}\n                ", "item_type": "full_problem", "prompt_chars": 1458, "record_id": "iol-2018-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["b.1", "b.2", "b.3"], "canonical_answer_chars": 339, "canonical_excerpt": "(b)    7. Did I beat yousg? — ŋabə nɤ lan tɤʔ ne\n         8. Did they see me? — tarum kəmə ŋa lapkʰi tʰɤ ne\n         9. Does he know yousg? — ati kəmə nɤ cʰam ru ne\n       10. Do youpl sleep? — nuʔrum ʒip kan ne\n\nSixteenth International Linguistics Olympiad (2018).                                      2\n      Individual Contest Solutions", "item_type": "subproblem", "prompt_chars": 1773, "record_id": "iol-2018-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2018-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2018-indiv-sol.en.pdf", "source_problem_id": "2018-individual-2", "target_subparts": ["b", "full"], "title": "Hakhun", "year": 2018}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Yurok (Algic)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 29, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "a.17", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "c.1", "c.2", "c.3", "c.4", "c.5", "c.6"], "canonical_answer_chars": 2191, "canonical_excerpt": "Problem 2.  Word order: A N (A = adjective, N = noun).\n\n          Colours:     #FFFFFF   #FF0000    #964B00     #000000\n          animals       muent-er’ery                  s’erkt-er’ery   ler’erg-ery\n            trees                         pekoy-ar’                      lo’og-ar’\n         round things  muench-erh    perkery-erh\n          everything    muench-ey                    s’okt-oy       lo’og-ey\n\n (a)    1.  muencherh rohkuen         G.   white ball\n         2.  perkeryerh holeehl ʼwernerh  M.   red hazelnut\n         3.  muencherh nerhpery        D.   white berry\n         4.   sʼerkterʼery chʼeeshah        A.  brown dog\n         5.  muencharʼ pyaapʼ           H.   white manzanita", "item_type": "full_problem", "prompt_chars": 2825, "record_id": "iol-2019-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 17, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "a.17"], "canonical_answer_chars": 1050, "canonical_excerpt": "(a)    1.  muencherh rohkuen         G.   white ball\n         2.  perkeryerh holeehl ʼwernerh  M.   red hazelnut\n         3.  muencherh nerhpery        D.   white berry\n         4.   sʼerkterʼery chʼeeshah        A.  brown dog\n         5.  muencharʼ pyaapʼ           H.   white manzanita bush\n         6.  pekoyarʼ tepoo             K.   red tree\n         7.  luuehlsonʼ nerhpery         N.   purple berry\n         8.  muenchey cheekʼwar              I.    white chair\n         9.  muenterʼery chʼeeshah        F.   white dog\n       10.   lerʼergery cherʼery           C.   black bear\n       11.  ʼerrwerhsonʼ slekwoh          L.   grass-green shirt\n       12.  muenterʼery puuek           P.   white", "item_type": "subproblem", "prompt_chars": 1471, "record_id": "iol-2019-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 273, "canonical_excerpt": "(b)  19.  muencherhl    V.   milk\n       20.  ʼwerʼerrgerch  W.   alder tree\n       21.  luuehl        X.   wild iris\n       22.   lerʼergerh     T.   coffee bean\n       23.  pekoyek       U.   blood\n       24.  skoyon        Y.   sky\n       25.  tegeeʼn         S.   canary", "item_type": "subproblem", "prompt_chars": 2305, "record_id": "iol-2019-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5", "c.6"], "canonical_answer_chars": 449, "canonical_excerpt": "(c)  26. ʼerrwerh — grass\n       27. lerʼergerh rohkuen — black ball\n       28. perkeryerʼery chʼeeshah — red dog\n       29. pyerrpʼtʼery chʼeeshah — reddish-brown dog\n\n (d)  30. purple deer — luuehlson’ puuek\n       31. white shoe — muenchey no’oy\n       32. yellow ball — tegee’n rohkuen\n       33. black tree — lo’ogar’ tepoo\n\nSeventeenth International Linguistics Olympiad (2019)                                    3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3335, "record_id": "iol-2019-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 6, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6"], "canonical_answer_chars": 273, "canonical_excerpt": "(b)  19.  muencherhl    V.   milk\n       20.  ʼwerʼerrgerch  W.   alder tree\n       21.  luuehl        X.   wild iris\n       22.   lerʼergerh     T.   coffee bean\n       23.  pekoyek       U.   blood\n       24.  skoyon        Y.   sky\n       25.  tegeeʼn         S.   canary", "item_type": "partial_variant", "prompt_chars": 3408, "record_id": "iol-2019-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 6, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5", "c.6"], "canonical_answer_chars": 449, "canonical_excerpt": "(c)  26. ʼerrwerh — grass\n       27. lerʼergerh rohkuen — black ball\n       28. perkeryerʼery chʼeeshah — red dog\n       29. pyerrpʼtʼery chʼeeshah — reddish-brown dog\n\n (d)  30. purple deer — luuehlson’ puuek\n       31. white shoe — muenchey no’oy\n       32. yellow ball — tegee’n rohkuen\n       33. black tree — lo’ogar’ tepoo\n\nSeventeenth International Linguistics Olympiad (2019)                                    3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 4739, "record_id": "iol-2019-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2019-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2019-indiv-sol.en.pdf", "source_problem_id": "2019-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Yurok", "year": 2019}
+{"author": null, "curation_statuses": {"auto_extracted_text_high_confidence": 8}, "languages": ["Zuni (isolate)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-prob.en.pdf", "record_count": 8, "record_types": {"full_problem": 1, "partial_variant": 3, "subproblem": 4}, "records": [{"answer_unit_count": 23, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "c.1", "c.2", "c.3", "c.4", "c.5", "e.1"], "canonical_answer_chars": 1787, "canonical_excerpt": "Problem 2.\n\n   • -ya collectivity: homa- ‘juniper leaf’ + -ya →homa-ya- ‘juniper branch’\n\n   • N-:-N (N = noun):\n     ha- ‘leaves’ + k’yawe ‘water’ →ha-:-k’yawe ‘tea’\n\n   • N-:-A (A = adjective):\n    mu- ‘bread’ + chikwa ‘sweet (adj.)’ →mu-:-chikwa ‘chocolate bar’\n\n   •\n                            singular   plural, substance\n       monosyllabic stem   -’le     -we\n       polysyllabic stem   -nne     -:we\n\n (a)    1. mo:chikwa datda:we — I. peach trees\n         2. haya:we — G. lit. piles of leaves (actually planted fields [fields of plants])\n         3. doko:ko mo’le — H. chicken egg\n         4. ma:chikwa — C. sugar\n         5. mu:chikwa — J. chocolate bar\n         6. mulonne, mu’le — D. loa", "item_type": "full_problem", "prompt_chars": 2545, "record_id": "iol-2021-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 9, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9"], "canonical_answer_chars": 571, "canonical_excerpt": "(a)    1. mo:chikwa datda:we — I. peach trees\n         2. haya:we — G. lit. piles of leaves (actually planted fields [fields of plants])\n         3. doko:ko mo’le — H. chicken egg\n         4. ma:chikwa — C. sugar\n         5. mu:chikwa — J. chocolate bar\n         6. mulonne, mu’le — D. loaf of bread\n         7. mowe — E. beads\n         8. ha:k’yawe — B. tea\n         9. doko:ko shiwe — A. chicken meat\n       10. mo:chikwa mulo:chikwa — F. peach pie\n\nEighteenth International Linguistics Olympiad (2021)                                     2\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2312, "record_id": "iol-2021-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 8, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8"], "canonical_answer_chars": 335, "canonical_excerpt": "(b)  11. ade muwe — O. blood clots\n       12. beyanne — P. field of grass\n       13. homa:we — M. juniper leaves\n       14. ma:k’yawe — N. salt water\n       15. ma:k’ose — S. salt\n       16. ma’le — K. granule, grain\n       17. homayanne — R. juniper branch\n       18. mi’le — L. ear of corn [corn cob]\n       19. ade k’yawe — Q. blood", "item_type": "subproblem", "prompt_chars": 940, "record_id": "iol-2021-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 447, "canonical_excerpt": "(c)  20. be’le — blade of grass            (d)  26. salty — k’ose\n       21. chikwa — sweet (adj.)                    27. juniper leaf — homanne\n       22. datdanne — tree                         28. piece of meat — shi’le\n       23. ade = ade k’yawe — blood               29. water — k’yawe\n       24. mulo:we — bread                         30. corn, maize — miwe\n       25. hawe — leaves                            31. fields of grass — beya:we", "item_type": "subproblem", "prompt_chars": 1244, "record_id": "iol-2021-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["e.1"], "canonical_answer_chars": 14, "canonical_excerpt": "(e) salt lake.", "item_type": "subproblem", "prompt_chars": 2124, "record_id": "iol-2021-individual-p2-sub-e", "target_subparts": ["e"], "visible_solved_subparts": []}, {"answer_unit_count": 8, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8"], "canonical_answer_chars": 335, "canonical_excerpt": "(b)  11. ade muwe — O. blood clots\n       12. beyanne — P. field of grass\n       13. homa:we — M. juniper leaves\n       14. ma:k’yawe — N. salt water\n       15. ma:k’ose — S. salt\n       16. ma’le — K. granule, grain\n       17. homayanne — R. juniper branch\n       18. mi’le — L. ear of corn [corn cob]\n       19. ade k’yawe — Q. blood", "item_type": "partial_variant", "prompt_chars": 1564, "record_id": "iol-2021-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 5, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 447, "canonical_excerpt": "(c)  20. be’le — blade of grass            (d)  26. salty — k’ose\n       21. chikwa — sweet (adj.)                    27. juniper leaf — homanne\n       22. datdanne — tree                         28. piece of meat — shi’le\n       23. ade = ade k’yawe — blood               29. water — k’yawe\n       24. mulo:we — bread                         30. corn, maize — miwe\n       25. hawe — leaves                            31. fields of grass — beya:we", "item_type": "partial_variant", "prompt_chars": 2231, "record_id": "iol-2021-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}, {"answer_unit_count": 1, "answer_unit_ids": ["e.1"], "canonical_answer_chars": 14, "canonical_excerpt": "(e) salt lake.", "item_type": "partial_variant", "prompt_chars": 3586, "record_id": "iol-2021-individual-p2-partial-through-abc-target-e", "target_subparts": ["e"], "visible_solved_subparts": ["a", "b", "c"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 2, "solution_page_start": 1, "solution_pdf_local_path": "data/raw/pdfs/iol-2021-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2021-indiv-sol.en.pdf", "source_problem_id": "2021-individual-2", "target_subparts": ["a", "b", "c", "e", "full"], "title": "Zuni", "year": 2021}
+{"author": "Ivan Derzhanski", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Alabama (Muskogean)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 31, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "a.17", "a.18", "a.19", "a.20", "a.21", "a.22", "a.23", "a.24", "a.25", "b.1", "b.2", "b.3", "b.4", "b.5", "d.1"], "canonical_answer_chars": 1933, "canonical_excerpt": "Problem 2.\n\n   •\n                    1        many\n       state    xéxx‹ex|èe›-ka  xèex-ka\n       action    xex‹ex|èe›-li     xex-li\n   •  l →C | C   (C ∈{b, f, ɬ, n})\n   • b →m |   C (C denotes any consonant)\n\n (a)    1. achánnàaka — G. it leans against something\n         2. achanni — N. he leans them against something\n         3. ahamatli — R. he scatters it, he scatters them\n         4. ahámmatka — M. it is scattered, they are scattered\n         5. apòotka — S. they are next to something\n         6. apotòoli — D. he puts it next to something\n         7. apóttòoka — I. it is next to something\n         8. atanayli — Z. he wraps it\n         9. atánnayka — K. it is wrapped\n       10. atanni —", "item_type": "full_problem", "prompt_chars": 2754, "record_id": "iol-2022-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 25, "answer_unit_ids": ["a.1", "a.2", "a.3", "a.4", "a.5", "a.6", "a.7", "a.8", "a.9", "a.10", "a.11", "a.12", "a.13", "a.14", "a.15", "a.16", "a.17", "a.18", "a.19", "a.20", "a.21", "a.22", "a.23", "a.24", "a.25"], "canonical_answer_chars": 1233, "canonical_excerpt": "(a)    1. achánnàaka — G. it leans against something\n         2. achanni — N. he leans them against something\n         3. ahamatli — R. he scatters it, he scatters them\n         4. ahámmatka — M. it is scattered, they are scattered\n         5. apòotka — S. they are next to something\n         6. apotòoli — D. he puts it next to something\n         7. apóttòoka — I. it is next to something\n         8. atanayli — Z. he wraps it\n         9. atánnayka — K. it is wrapped\n       10. atanni — C. he wraps them\n       11. bàalka — L. they lie, are lying down\n       12. balàali — V. he lays it down\n       13. balli — B. he lays them down\n       14. incháffàaka — U. he has one\n       15. kawaɬɬi — A. he ", "item_type": "subproblem", "prompt_chars": 1838, "record_id": "iol-2022-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 334, "canonical_excerpt": "(b)  27. bállàaka — he/it lies, is lying down\n       28. inchafàali — he gets one\n       29. kawwi — he breaks it in several places, he breaks them\n (c)  30. he tears it in several places — ɬilli\n       31.  it is torn in two — ɬíllafka\n       32. he turns it upside down — ɬipìili\n       33. he leans it against something — achanàali", "item_type": "subproblem", "prompt_chars": 2616, "record_id": "iol-2022-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 150, "canonical_excerpt": "(d)  ɬilli — he tears them.\n\nNineteenth International Linguistics Olympiad (2022)                                       3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 2882, "record_id": "iol-2022-individual-p2-sub-d", "target_subparts": ["d"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5"], "canonical_answer_chars": 334, "canonical_excerpt": "(b)  27. bállàaka — he/it lies, is lying down\n       28. inchafàali — he gets one\n       29. kawwi — he breaks it in several places, he breaks them\n (c)  30. he tears it in several places — ɬilli\n       31.  it is torn in two — ɬíllafka\n       32. he turns it upside down — ɬipìili\n       33. he leans it against something — achanàali", "item_type": "partial_variant", "prompt_chars": 3902, "record_id": "iol-2022-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 1, "answer_unit_ids": ["d.1"], "canonical_answer_chars": 150, "canonical_excerpt": "(d)  ɬilli — he tears them.\n\nNineteenth International Linguistics Olympiad (2022)                                       3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 4530, "record_id": "iol-2022-individual-p2-partial-through-ab-target-d", "target_subparts": ["d"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2022-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2022-indiv-sol.en.pdf", "source_problem_id": "2022-individual-2", "target_subparts": ["a", "b", "d", "full"], "title": "Alabama", "year": 2022}
+{"author": "Ksenia Gilyarova", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Apurinã (Arawakan)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 4, "problem_page_start": 3, "problem_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 8, "answer_unit_ids": ["a.1", "a.2", "b.1", "c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 2797, "canonical_excerpt": "Problem 2.\n\n   • Word order:\n\n      – S V DO (IO) ≡DO SV (IO) (S(noun))\n          (S = subject; V = predicate (verb);\n      DO = direct object; IO = indirect object [-mukary])\n\n      –  possessor  possessed\n\n   •\n                              1st person sg  2nd person sg  3rd person sg\n\n                                                  (m)     (f)\n\n       possessor = subject        n(I)-              p(I)-          (I)-     u-\n\n       IO; DO (¬∃IO)           -nu                        -ry     -ru\n      \n        i     {nh, tx, x} i\n        (I)- =   y    C                   ; n > nh |      i\n                   ∅   V\n\n   •\n      body part  ¬ possessed   -txi (-y > -ĩ |    -txi)\n\n    ", "item_type": "full_problem", "prompt_chars": 2435, "record_id": "iol-2023-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 2, "answer_unit_ids": ["a.1", "a.2"], "canonical_answer_chars": 279, "canonical_excerpt": "(a)  17. uwa xinhikaru sytu — She remembered the woman.\n       18. txipary nhinhikary — I ate the banana.\n       19. kyky aikute pysykary ywamukary — Yousg gave the man’s house to him.\n                          \n                                  uwa mynary kanawa ximakymukary", "item_type": "subproblem", "prompt_chars": 1616, "record_id": "iol-2023-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 378, "canonical_excerpt": "(b)  20.                  — She brought the canoe to the fish (sg).\n                                   kanawa umynary ximakymukary\n\nTwentieth International Linguistics Olympiad (2023)                                        3\n\nIndividual Contest Solutions\n                           \n                                                                             aiku papukary", "item_type": "subproblem", "prompt_chars": 1730, "record_id": "iol-2023-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 5, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 1402, "canonical_excerpt": "(c)  21. Yousg found the house.      —\n                            pita apukary aiku\n                           \n                                                                        pykanawate nymynaru hãtakurumukary\n       22. I brought yoursg canoe to the girl. —\n                            nuta mynaru pykanawate hãtakurumukary\n                           \n                                                                               pĩpurããte ymynanu nutamukary\n       23. He brought yoursg water to me.  —\n                            ywa mynanu pĩpurããte nutamukary\n                           \n                                                                           hãtaku", "item_type": "subproblem", "prompt_chars": 2954, "record_id": "iol-2023-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 378, "canonical_excerpt": "(b)  20.                  — She brought the canoe to the fish (sg).\n                                   kanawa umynary ximakymukary\n\nTwentieth International Linguistics Olympiad (2023)                                        3\n\nIndividual Contest Solutions\n                           \n                                                                             aiku papukary", "item_type": "partial_variant", "prompt_chars": 2062, "record_id": "iol-2023-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 5, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5"], "canonical_answer_chars": 1402, "canonical_excerpt": "(c)  21. Yousg found the house.      —\n                            pita apukary aiku\n                           \n                                                                        pykanawate nymynaru hãtakurumukary\n       22. I brought yoursg canoe to the girl. —\n                            nuta mynaru pykanawate hãtakurumukary\n                           \n                                                                               pĩpurããte ymynanu nutamukary\n       23. He brought yoursg water to me.  —\n                            ywa mynanu pĩpurããte nutamukary\n                           \n                                                                           hãtaku", "item_type": "partial_variant", "prompt_chars": 3692, "record_id": "iol-2023-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 4, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2023-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2023-indiv-sol.en.pdf", "source_problem_id": "2023-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Apurinã", "year": 2023}
+{"author": "Samuel Ahmed", "curation_statuses": {"auto_extracted_text_high_confidence": 4}, "languages": ["Hadza (Isolate)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 4, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-prob.en.pdf", "record_count": 4, "record_types": {"full_problem": 1, "partial_variant": 1, "subproblem": 2}, "records": [{"answer_unit_count": 13, "answer_unit_ids": ["a.1", "a.2", "a.3", "b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10"], "canonical_answer_chars": 1800, "canonical_excerpt": "Problem 2.  possesseda-sfxa-POSSb possessorb-sfxb\n                                                             fm\n                                                           • animals:  zebra, impala, elephant, baboon, gi-      • animals: lion, leopard, flamingo\n                                                           raffe\n      • objects, body parts: long/thin\n                                                           • objects, body parts: short/thick\n      •\n               sg     pl                             •\n                                                         sg    pl         sfx    -∅     -bii *\n                                                        sfx     -ko   -bee      PO", "item_type": "full_problem", "prompt_chars": 3463, "record_id": "iol-2024-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 3, "answer_unit_ids": ["a.1", "a.2", "a.3"], "canonical_answer_chars": 173, "canonical_excerpt": "(a)   19. wiriko — D. short, thick root\n        20. roobee — C. horns (e.g., of dik-diks)\n        21. mu’a — B. thin twig\n        22. zzahubii — A. tails (e.g., of leopards)", "item_type": "subproblem", "prompt_chars": 2095, "record_id": "iol-2024-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 10, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10"], "canonical_answer_chars": 716, "canonical_excerpt": "(b)   23. gariko — car (= short lorry)\n        24. makubii — tall cooking pots\n        25. hazzakeko — female thief\n        26. dongobee — zebras\n        27. zzahoko — (short, thick) tail\n        28. nqibii — leopards\n\n   (c)   29. the flamingo’s leg — uphukwama gogogogo\n        30. the zebra’s V. pseudolablab tuber — shumukosa dongoko\n        31. the male impalas’ horns — ruubiitcha puphubii\n        32. the female strangers’ thick stick — mu’akoeta uhuyitibee\n        33. the boys’ tall cooking pot — makuitcha eratibii\n        34. the impala’s long, thin roots — wiribiisa pophoko\n\nTwenty-first International Linguistics Olympiad (2024)                                            3\nIndividual C", "item_type": "subproblem", "prompt_chars": 4907, "record_id": "iol-2024-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 10, "answer_unit_ids": ["b.1", "b.2", "b.3", "b.4", "b.5", "b.6", "b.7", "b.8", "b.9", "b.10"], "canonical_answer_chars": 716, "canonical_excerpt": "(b)   23. gariko — car (= short lorry)\n        24. makubii — tall cooking pots\n        25. hazzakeko — female thief\n        26. dongobee — zebras\n        27. zzahoko — (short, thick) tail\n        28. nqibii — leopards\n\n   (c)   29. the flamingo’s leg — uphukwama gogogogo\n        30. the zebra’s V. pseudolablab tuber — shumukosa dongoko\n        31. the male impalas’ horns — ruubiitcha puphubii\n        32. the female strangers’ thick stick — mu’akoeta uhuyitibee\n        33. the boys’ tall cooking pot — makuitcha eratibii\n        34. the impala’s long, thin roots — wiribiisa pophoko\n\nTwenty-first International Linguistics Olympiad (2024)                                            3\nIndividual C", "item_type": "partial_variant", "prompt_chars": 5133, "record_id": "iol-2024-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}], "requires_images": true, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2024-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2024-indiv-sol.en.pdf", "source_problem_id": "2024-individual-2", "target_subparts": ["a", "b", "full"], "title": "Hadza", "year": 2024}
+{"author": "David Hultman", "curation_statuses": {"auto_extracted_text_high_confidence": 6}, "languages": ["Gaahmg (Eastern Sudanic)"], "problem_extraction_confidence": "heading", "problem_number": 2, "problem_page_end": 3, "problem_page_start": 2, "problem_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-prob.en.pdf", "problem_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-prob.en.pdf", "record_count": 6, "record_types": {"full_problem": 1, "partial_variant": 2, "subproblem": 3}, "records": [{"answer_unit_count": 10, "answer_unit_ids": ["a.1", "b.1", "c.1", "c.2", "c.3", "c.4", "c.5", "c.6", "c.7", "c.8"], "canonical_answer_chars": 1877, "canonical_excerpt": "Problem 2.  Rules:\n\n     • Word order:\n\n       – inalienable possession: possessor + noun\n       – alienable possession: noun + possessor\n\n     • Possessive pronouns:\n\n                            inalienable possession     alienable possession\n                         kinship term  body part   singular     plural\n           singular              V́*               V̄         V́V̀n       V́nV̀g\n             plural                    V̄g               V̄yV̀n      V̄yV̀g\n                                                  V: 1/2/3 →ə/u/i       person marking      V: 1/2/3 →a/ɔ/ɛ                                                 3 →add a suffix -ī\n        * High tone is only applicable for first and ", "item_type": "full_problem", "prompt_chars": 2794, "record_id": "iol-2025-individual-p2", "target_subparts": [], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["a.1"], "canonical_answer_chars": 166, "canonical_excerpt": "(a)\n        1    J    2  A   3  D   4  T    5  E    6  S    7   F    8  B   9  P   10  L\n       11  H  12  N  13  G  14 M  15  Q  16  K  17  O  18    I   19  C  20  R", "item_type": "subproblem", "prompt_chars": 1589, "record_id": "iol-2025-individual-p2-sub-a", "target_subparts": ["a"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 86, "canonical_excerpt": "(b) his grandmother; his grandfather — ɛ does not have high tone before kinship terms.", "item_type": "subproblem", "prompt_chars": 1905, "record_id": "iol-2025-individual-p2-sub-b", "target_subparts": ["b"], "visible_solved_subparts": []}, {"answer_unit_count": 8, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5", "c.6", "c.7", "c.8"], "canonical_answer_chars": 521, "canonical_excerpt": "(c)   21. āg bɔ̀ɔ̀rāāg — our shoulders\n        22.  ɖɔ̄ɔ̄rɛ́ɛ̄g ə̄yə̀g — our hammers\n        23.  ɛ̄ bɔ̄ɔ̄rààg — his shoulders\n        24.  ɔ̄ túndúlììg — yoursg elbows\n        25.  ɔ́ máàm — yoursg aunt\n  (d)   26. my grindstone — gùùr ə́ə̀n\n        27. their cheeks — ɛ̄g fə̀ndə̄g\n        28. yourpl anchor — ʈɛ́ɛ̀l ūyùn\n        29. our uncle — āg ābɛ́ɛ́\n        30. yoursg dogs — áðág únùg\n\nTwenty-second International Linguistics Olympiad (2025)                                          3\nIndividual Contest Solutions", "item_type": "subproblem", "prompt_chars": 3667, "record_id": "iol-2025-individual-p2-sub-c", "target_subparts": ["c"], "visible_solved_subparts": []}, {"answer_unit_count": 1, "answer_unit_ids": ["b.1"], "canonical_answer_chars": 86, "canonical_excerpt": "(b) his grandmother; his grandfather — ɛ does not have high tone before kinship terms.", "item_type": "partial_variant", "prompt_chars": 2124, "record_id": "iol-2025-individual-p2-partial-through-a-target-b", "target_subparts": ["b"], "visible_solved_subparts": ["a"]}, {"answer_unit_count": 8, "answer_unit_ids": ["c.1", "c.2", "c.3", "c.4", "c.5", "c.6", "c.7", "c.8"], "canonical_answer_chars": 521, "canonical_excerpt": "(c)   21. āg bɔ̀ɔ̀rāāg — our shoulders\n        22.  ɖɔ̄ɔ̄rɛ́ɛ̄g ə̄yə̀g — our hammers\n        23.  ɛ̄ bɔ̄ɔ̄rààg — his shoulders\n        24.  ɔ̄ túndúlììg — yoursg elbows\n        25.  ɔ́ máàm — yoursg aunt\n  (d)   26. my grindstone — gùùr ə́ə̀n\n        27. their cheeks — ɛ̄g fə̀ndə̄g\n        28. yourpl anchor — ʈɛ́ɛ̀l ūyùn\n        29. our uncle — āg ābɛ́ɛ́\n        30. yoursg dogs — áðág únùg\n\nTwenty-second International Linguistics Olympiad (2025)                                          3\nIndividual Contest Solutions", "item_type": "partial_variant", "prompt_chars": 4000, "record_id": "iol-2025-individual-p2-partial-through-ab-target-c", "target_subparts": ["c"], "visible_solved_subparts": ["a", "b"]}], "requires_images": false, "requires_tables": true, "review_instructions": ["Compare model-facing prompt/problem text with the cited problem PDF pages.", "Compare solution_text/canonical answers/answer_units with cited solution PDF pages.", "Check subproblem labels, partial-variant shown solutions, and target question do not leak target answers.", "Mark pass only if all derived records for this source are faithful and answerable."], "round": "individual", "solution_extraction_confidence": "heading", "solution_page_end": 3, "solution_page_start": 2, "solution_pdf_local_path": "data/raw/pdfs/iol-2025-indiv-sol.en.pdf", "solution_pdf_url": "https://ioling.org/booklets/iol-2025-indiv-sol.en.pdf", "source_problem_id": "2025-individual-2", "target_subparts": ["a", "b", "c", "full"], "title": "Gaahmg", "year": 2025}

benchmark/IOL/ioling_hf/reports/manual_verification/verification_summary.md

@@ -0,0 +1,299 @@
+# Manual Verification Summary
+
+Expected source problems: 130
+Covered source problems: 130
+Finding rows: 130
+Missing source problems: 0
+Duplicate source problems: 0
+
+## Verdicts
+
+| verdict | count |
+| --- | ---: |
+| fix_needed | 130 |
+
+## Issue Severity
+
+| severity | count |
+| --- | ---: |
+| critical | 61 |
+| high | 101 |
+| low | 27 |
+| medium | 83 |
+
+## Issues
+
+| source | verdict | severity | field | record | description | suggested fix |
+| --- | --- | --- | --- | --- | --- | --- |
+| 2003-individual-1 | fix_needed | critical | prompt |  | The rendered problem PDF uses special mathematical/fraction glyphs for the artificial language, but the dataset OCR replaces them with ordinary letters and punctuation, so the examples and assignments are not faithful or reliably answerable from text. | Re-extract this problem from rendered pages or encode the glyph inventory explicitly so all examples and assignment items match the source. |
+| 2003-individual-1 | fix_needed | critical | answer_units |  | Official answers for assignment items 9, 13, and 15 are missing from answer_units; several remaining formula answers are OCR-corrupted. | Regenerate answer_units from the rendered official solution with one required unit for each target item, including 9, 13, and 15. |
+| 2003-individual-1 | fix_needed | high | canonical |  | The canonical solution text corrupts the source glyphs throughout the noun/verb rules and Assignment 2 formulas. | Replace canonical solution_text with a faithful transcription or rendered-safe notation for the source symbols. |
+| 2003-individual-2 | fix_needed | high | prompt |  | The image-PDF prompt OCR replaces source transliteration symbols and fraction layout with incorrect characters such as yen/2/c-like placeholders, changing the Arabic forms and notes from the rendered source page. | Re-transcribe the problem from the rendered PDF, preserving the source transliteration letters and displayed fractions. |
+| 2003-individual-2 | fix_needed | critical | canonical |  | The solution/canonical text and shown solution for assignment 1 contain garbled equations such as "$+3=$" instead of the official fraction equations; assignment 2 also has corrupted fractions and transliteration. | Rebuild canonical answers and shown solutions from the rendered solution page, e.g. assignment 1 should contain the five explicit fraction equalities and assignment 2 the square-root or square sign alternatives. |
+| 2003-individual-2 | fix_needed | medium | answer_units |  | Assignment 1 is one bundled answer unit for five equalities, which is too coarse for partial-credit RL scoring even after transcription is fixed. | Split assignment 1 into separate answer units for each numbered equality and keep assignment 2 as its own unit(s). |
+| 2003-individual-2 | fix_needed | low | provenance |  | The solution page range includes page 3, which is the next problem; Problem 2 solution is on page 2. | Set the solution page range for this source to page 2 only. |
+| 2003-individual-3 | fix_needed | medium | prompt |  | OCR text drops the visible underline blanks in the Basque/English lists and carries in an unrelated Egyptian Arabic footnote from the same PDF page; the prompt is mostly readable but not a clean rendering of the cited problem. | Manually transcribe Problem 3 from page 2, preserving blank positions and only the Basque footnote. |
+| 2003-individual-3 | fix_needed | medium | answer_units |  | Each assignment is a single answer unit even where Assignment 1 has ten matches/fill-ins and Assignment 2 has four translations, which is too coarse for RL scoring. | Split answer_units into one unit per requested date/match/translation and keep the literal-name explanation as separate units. |
+| 2003-individual-4 | fix_needed | critical | prompt |  | The image-only problem page OCR does not match the dataset on target Adyghe forms; for example sentence 8 reads as a throw-form ending like "deza" in the rendered/OCR source, while the dataset prompt has "dega", and other special consonants drift in sentences 4 and 6. | Manually transcribe the Adyghe forms from the rendered PDF and correct the prompt text before using these records. |
+| 2003-individual-4 | fix_needed | high | answer_units |  | Answer units omit required numbered answers: assignment 2 lacks item 8, assignment 3 lacks items 10 and 11, and assignment 4 collapses three possible answers into one unit. | Create separate answer_units for every numbered answer and every assignment 4 alternative, including all units in full, subproblem, and partial records. |
+| 2003-individual-4 | fix_needed | medium | canonical |  | The canonical solution has OCR artifacts such as "ll." for item 11 and questionable Adyghe special-consonant transcription. | Recheck the rendered solution page and repair the canonical solution text. |
+| 2003-individual-5 | fix_needed | high | prompt | iol-2003-individual-p5 | The French table's shaded/no-counterpart cells are OCRed as long nonsense strings instead of being represented as shaded or empty cells, so the model-facing table is not faithful to the PDF. | Re-extract or manually transcribe the table, preserving blank prefixes and shaded no-counterpart cells in a consistent textual notation. |
+| 2003-individual-5 | fix_needed | high | canonical | iol-2003-individual-p5 | The solution answer lines contain OCR garbage in the shaded-cell positions, including corrupted strings after entries such as reformer/refuter/remunerate/distribute. | Replace the corrupted solution table with a clean transcription of the intended completed rows and explanation. |
+| 2003-individual-5 | fix_needed | medium | answer_units | iol-2003-individual-p5 | The single full.1 answer unit covers all gap fills plus the explanation, which is too coarse for RL scoring of the individual table entries. | Split the expected answer into separate units for each filled gap and a separate explanation/rule unit. |
+| 2003-team-1 | fix_needed | high | prompt | iol-2003-team-p1 | The scanned source page is OCR-corrupted in the model prompt: multiple Proto-Tocharian forms and the note symbols are misread or replaced by placeholder characters, affecting core linguistic data. | Re-extract from rendered pages or manually transcribe the problem, preserving the special vowel/length/ng symbols and the reconstructed word forms. |
+| 2003-team-1 | fix_needed | medium | answer_units | iol-2003-team-p1 | The only answer unit is a short fragment of Assignment 1 and does not cover the language allocation, proto-form reconstructions, or stress explanation required by the full problem. | Split answer_units by the required outputs for Assignments 1, 2(a-c), and 3, with normalized acceptable values for each. |
+| 2003-team-2 | fix_needed | critical | prompt |  | The source relies on subscripted variables and asterisk notation, but the dataset flattens these into semicolons and stray punctuation, changing the linguistic data. | Rebuild the prompt from a rendered transcription that preserves subscripts, asterisks, and sentence pair structure. |
+| 2003-team-2 | fix_needed | critical | answer_units |  | Assignment 1 explanation is omitted from answer_units, while the a/b/c solved forms contain corrupted subscript alternatives; the c unit also includes a footer. | Create faithful units for the Assignment 1 rule explanation and each requested translation, without page headers or footers. |
+| 2003-team-2 | fix_needed | high | canonical |  | The canonical solution has the same subscript/OCR corruption as the prompt, so partial variants show unreliable solved forms. | Replace solution_text and shown partial solutions with a notation-preserving transcription. |
+| 2003-team-3 | fix_needed | medium | answer_units |  | Several answer units bundle multiple independently scorable items: assignment 2 has seven verbs in one unit and assignment 3 asks for two examples in one unit. | Split assignment 2 into one unit per accepted verb and assignment 3 into separate example units or an explicit two-example set rubric. |
+| 2003-team-3 | fix_needed | low | prompt |  | The prompt has an OCR typo: "Tt is known" should be "It is known". | Correct the prompt text against the rendered source page. |
+| 2003-team-3 | fix_needed | low | canonical |  | The solution text has minor OCR errors in prose, including a missing opening quote before "I denounce" and "of" where the source says "or" in the parenthetical example. | Correct the solution prose from the rendered solution page. |
+| 2004-individual-1 | fix_needed | high | prompt |  | Kayapo special vowels are corrupted in the prompt: PDF forms such as mé/mẽ and aje/ije are represented with ! and quotation-mark artifacts (for example "Ba m! m! anhê", "Ij" m! akuru kêt", "M! aj" inhêrê kêt"). | Re-extract or manually transcribe the Kayapo forms using the PDF glyphs, e.g. mé, aje, ije, and the note vowels. |
+| 2004-individual-1 | fix_needed | high | canonical |  | The solution answer for Assignment 1 contains OCR damage: "Tréré két Tam not swimming" instead of the PDF answer "Irêrê két — I am not swimming". | Correct the Assignment 1 canonical answer and derived answer_units from the rendered solution page. |
+| 2004-individual-1 | fix_needed | medium | answer_units |  | Assignment-level answer units group several translations together, making partial scoring too coarse. | Create separate units for each Assignment 1 sentence and each Assignment 2 translation. |
+| 2004-individual-2 | fix_needed | low | prompt |  | The prompt includes a stray Kayapo footnote from Problem 1 and has "SwiftNews" without the source-space in "Swift News". | Remove the unrelated Kayapo footnote from Problem 2 and restore the agency name spacing. |
+| 2004-individual-2 | fix_needed | high | answer_units | iol-2004-individual-p2 | The five answer_units are truncated fragments: the corrected titles for 6, 7, and 9 omit essential continuation text, and the explanatory group units stop mid-sentence. | Rebuild units as complete explanation components and complete corrected titles, e.g. including "Fraught with Tax Reform" and the full 7 and 9 titles. |
+| 2004-individual-3 | fix_needed | high | prompt |  | Latin long-vowel marks in the problem word lists are corrupted into placeholder punctuation such as d$vidit, mord%x, f&mus, frac's, fov're, rubr$ca, combr'tum, and d'beo. | Manually restore the Latin forms with their proper macrons/long-vowel notation throughout all derived prompts. |
+| 2004-individual-3 | fix_needed | high | canonical |  | The Assignment 2 solution includes OCR/footer debris after the actual answer and several character confusions in the explanatory solution text. | Clean the official solution text so Assignment 2 ends after the debeo explanation and remove unrelated footer/OCR fragments. |
+| 2004-individual-3 | fix_needed | medium | answer_units | iol-2004-individual-p3-sub-assignment_2 | Assignment 2 is represented as one unit even though it asks which of six words are excluded and requires multiple answer items. | Split Assignment 2 into separate answer units for brevis, gurdus, debeo, and the shared rationale or acceptance criteria. |
+| 2004-individual-4 | fix_needed | high | prompt |  | The Lakhota problem text has corrupted special characters in key data and target words, e.g. placeholder characters appear in forms that the rendered source shows as distinct Lakhota symbols. | Manually transcribe the Lakhota forms and notes from the rendered PDF, preserving the source transliteration consistently. |
+| 2004-individual-4 | fix_needed | high | canonical |  | The solution text and canonical answers contain OCR garbage in the explanatory table and answer strings, including corrupted fricative symbols and several damaged word forms. | Replace the solution_text and canonical subpart answers with a clean transcription of the official solution for Problem 4. |
+| 2004-individual-4 | fix_needed | medium | answer_units |  | Assignments 1 and 2 are each represented as one free-text unit even though they require multiple separate matches/translations, which is too coarse for RL scoring. | Create separate answer_units for each matched item in Assignment 1 and each translation requested in Assignment 2. |
+| 2004-individual-5 | fix_needed | critical | prompt | iol-2004-individual-p5 | Chuvash letters and diacritics from the source are replaced by font-encoding artifacts such as commas, ampersands, apostrophes, and parentheses, which changes the forms needed to solve the problem. | Re-extract or manually transcribe the Chuvash data with the correct Unicode letters and diacritics. |
+| 2004-individual-5 | fix_needed | critical | answer_units | iol-2004-individual-p5 | The answer_units contain only two suffix-rule statements and omit the completed table/gap answers required by the official solution. | Add granular answer_units for each requested filled form and keep rule statements separate from final answers. |
+| 2004-individual-5 | fix_needed | high | canonical | iol-2004-individual-p5 | The canonical solution loses much of the completed answer table and ends with garbled symbol strings. | Replace the canonical answer with a clean transcription of the official solution pages. |
+| 2005-individual-1 | fix_needed | medium | answer_units |  | Assignment 1's four translations and assignment 2's five translations are each bundled into a single answer unit, which is too coarse for RL scoring. | Create separate answer units for problem items 16-24. |
+| 2005-individual-1 | fix_needed | medium | canonical |  | The canonical solution includes OCR artifacts, including a slash in the final answer ("Ch/abal" instead of the source form) and page-footer text from the next page. | Clean the canonical/subpart answers against the OCR-rendered solution page and remove page footer noise. |
+| 2005-individual-1 | fix_needed | low | provenance |  | The cited solution range includes page 2, which starts Problem 2; Problem 1 solution is on page 1. | Set the solution page range to page 1 only for this source. |
+| 2005-individual-2 | fix_needed | high | prompt |  | The Lango prompt loses distinctive source glyphs and tones: rendered PDF has forms such as dyè ɔ̀t, nìg wàŋ, ɔ̀t cèm, but the dataset has plain/incorrect forms such as dye ot, nig way, at cem. | Manually transcribe the Lango forms and note from the PDF, preserving ɔ/ɛ, ŋ, and tone marks. |
+| 2005-individual-2 | fix_needed | high | canonical |  | The canonical solution contains OCR substitutions in target forms, including "dye dt", "St cém", and "way ot" where the PDF shows dyè ɔ̀t, ɔ̀t cèm, and wàŋ ɔ̀t. | Correct the solution_text, canonical answers, and answer_units from the rendered solution page. |
+| 2005-individual-2 | fix_needed | medium | answer_units |  | Assignment 1 is one large unit with eight correspondences, too coarse for RL scoring. | Split into one unit per correspondence and one unit per short answer in Assignments 2 and 3. |
+| 2005-individual-3 | fix_needed | high | prompt |  | The Mansi prompt has OCR/provenance contamination: the 918 example is rendered as "... nollowxujplow" in the source OCR but appears as "... hollowxujplow" in the dataset, and an unrelated Lango footnote plus page footer are included. | Manually re-extract Problem 3 from the rendered PDF, correct the 918 numeral, and remove unrelated footnotes/footers. |
+| 2005-individual-3 | fix_needed | medium | answer_units |  | Assignments 1 and 2 each contain three requested answers but are stored as one coarse unit per assignment; the assignment 2 unit also includes the solution page footer. | Split into six numbered answer_units and remove page footer text from canonical/unit values. |
+| 2005-individual-4 | fix_needed | high | prompt |  | Yoruba special consonants/vowels and tone marks are corrupted in the prompt, e.g. bracket loss around a city-witch phrase, s/y substitutions, and note symbols rendered as 3, c, or 0 rather than the source characters. | Manually transcribe the Yoruba forms and note from the rendered PDF, preserving brackets, special letters, and tone marks. |
+| 2005-individual-4 | fix_needed | high | canonical |  | The solution answers carry the same OCR damage, e.g. [oko sya], laze elu], [ila a3e], missing spacing in 'house of love', and a page footer included in the answer. | Replace the solution answer blocks with clean Yoruba forms from the PDF and remove page footer text from canonical answers. |
+| 2005-individual-4 | fix_needed | medium | answer_units |  | Each assignment answer unit bundles four separate translations, which is too coarse for item-level scoring. | Create one answer unit per numbered item 14-21. |
+| 2005-individual-5 | fix_needed | high | prompt |  | The Lithuanian accent marks and paradigm table are corrupted in the prompt, replacing source accent/intonation notation with ambiguous or wrong characters. This damages the core data needed to solve the problem. | Rebuild the prompt from a manual transcription that preserves the Lithuanian accent marks, nasal notation, and table entries. |
+| 2005-individual-5 | fix_needed | high | canonical |  | The official solution/canonical text loses the crucial accent-pattern notation, reducing it to ambiguous strings such as aa, dd, ad, and da instead of the source-marked patterns. | Transcribe the solution table and Saussure-law explanation with faithful accent and stress notation. |
+| 2005-individual-5 | fix_needed | medium | subparts | iol-2005-individual-p5-partial-through-assignment_1-target-assignment_2 | The partial variant for Assignment 2 shows extra reasoning under the solved Assignment 1 block that belongs to deriving Assignment 2, giving target-relevant information before the target question. | Limit the visible solved Assignment 1 block to the actual answer to Assignment 1, or relabel/split the official solution spans more carefully. |
+| 2006-individual-1 | fix_needed | medium | answer_units |  | Some answer units are not granular or complete enough for RL scoring: assignment_1.3 is truncated after 'we can't', and assignment_2.1 combines two possible translations in one required unit. | Split alternatives into separate acceptable values and restore the missing continuation of the assignment_1.3 explanation. |
+| 2006-individual-1 | fix_needed | low | prompt |  | The prompt includes a stray page/footer line from the next page, though the problem text itself is otherwise faithful. | Trim page headers and footers from the extracted prompt. |
+| 2006-individual-2 | fix_needed | medium | answer_units |  | Assignment 1 is one large answer unit covering all missing Catalan forms, which is too coarse for partial-credit scoring. | Split assignment 1 into separate units for each missing singular/plural/article form or row. |
+| 2006-individual-2 | fix_needed | low | canonical |  | The canonical and subpart answers include page-footer noise and a wrong footer page number. | Remove contest footer text from canonical/subpart answers. |
+| 2006-individual-2 | fix_needed | low | provenance |  | Problem and solution page ranges include page 3, which is the next problem; the Catalan problem and solution are on page 2. | Narrow both problem and solution ranges to page 2. |
+| 2006-individual-3 | fix_needed | medium | answer_units |  | Assignment 1 and Assignment 2 are each single answer units even though they contain many correspondences/romanizations; this is too coarse for RL scoring. | Split Assignment 1 into individual word correspondence units and Assignment 2 into one unit per requested form. |
+| 2006-individual-4 | fix_needed | medium | answer_units |  | Each assignment is a single coarse unit even when it contains multiple translations or yes/no judgments; assignment 4 units also include source footer text. | Split assignment 1-4 answers into individual phrase/judgment units and remove attribution/page footer text from answer values. |
+| 2006-individual-5 | fix_needed | medium | answer_units | iol-2006-individual-p5 | The prompt and solution text match the cited pages, but the single full.1 unit covers translations, error diagnoses, and four corrected Ngoni sentences for each case. | Split the expected answer into units for each erroneous sentence's English translation, error explanation, and accepted corrected Ngoni alternatives. |
+| 2007-individual-1 | fix_needed | critical | prompt | iol-2007-individual-p1 | The Braille dot diagrams from the source problem are absent from the model-facing prompt, leaving only blank space after the example sentences; the task is not answerable. | Include rendered Braille diagrams as images or encode each Braille cell in an unambiguous textual representation. |
+| 2007-individual-1 | fix_needed | critical | canonical | iol-2007-individual-p1 | The canonical answer and answer_unit are for 2007 Problem 2 (Movima), while this source problem is 2007 Problem 1 (Braille). | Replace the solution provenance and canonical answer with the official Braille solution, or remove the record until a correct source solution is available. |
+| 2007-individual-1 | fix_needed | critical | provenance | iol-2007-individual-p1 | The cited solution page range points to the Movima Problem 2 solution, not the Braille Problem 1 solution. | Correct the solution page range/source or mark the Braille solution as unavailable if it is not present in the local PDF. |
+| 2007-individual-2 | fix_needed | high | prompt | iol-2007-individual-p2 | Movima consonant symbols from the source are encoded as ASCII punctuation in the prompt and notes, making examples such as the special c/l/n letters visually non-faithful. | Use correct Unicode characters or an explicit normalized notation for the Movima symbols in the prompt. |
+| 2007-individual-2 | fix_needed | medium | canonical | iol-2007-individual-p2 | The canonical solution uses the same degraded symbol encoding, so the official forms are not faithful to the PDF. | Correct the solution transcription to use the same notation as the repaired prompt. |
+| 2007-individual-2 | fix_needed | medium | answer_units | iol-2007-individual-p2 | The single full.1 answer unit contains the whole solution and a footer, which is too coarse for scoring individual targets. | Split the final answers and rule explanations into separate units and remove page footer text. |
+| 2007-individual-3 | fix_needed | critical | canonical | iol-2007-individual-p3 | The canonical solution is corrupted by interleaving prose with the layout table, making the explanation hard to read and not faithful to the solution PDF order. | Re-extract or manually transcribe the solution so prose, final correspondence table, and tense table are separated in source order. |
+| 2007-individual-3 | fix_needed | critical | answer_units | iol-2007-individual-p3 | The only answer unit has key "keni" and value "tthe suffix -eb by contrasting visadileb and vsadilob. We can't", which is a stray fragment rather than the requested Georgian-English correspondences. | Replace with structured answer units for the 15 final correspondences and the ambiguity/plausibility note. |
+| 2007-individual-3 | fix_needed | low | prompt | iol-2007-individual-p3 | The prompt has minor OCR artifacts such as "Problem !3" and a garbled section marker before the task. | Correct headings/section markers from the source PDF. |
+| 2007-individual-4 | fix_needed | high | prompt | iol-2007-individual-p4 | Important mathematical symbols are corrupted: the prompt has "mer abo sas $ meregh" where the PDF has a multiplication sign, and problem number/sign glyphs are also mapped incorrectly. | Manually correct multiplication signs, problem markers, and symbols from the rendered PDF. |
+| 2007-individual-4 | fix_needed | high | canonical | iol-2007-individual-p4 | The solution_text/canonical maps multiplication and exponent notation to punctuation such as %, &, and apostrophes (for example 6& and 7'x) instead of 6² and 7×x. | Replace the corrupted notation with the actual PDF math symbols throughout the canonical solution. |
+| 2007-individual-4 | fix_needed | medium | answer_units | iol-2007-individual-p4 | The full problem has a single full-answer unit, too coarse for the four requested tasks. | Create units for §1 square identifications, §2 equality, §3 numerals, and §4 Ndom forms. |
+| 2007-individual-5 | fix_needed | medium | answer_units | iol-2007-individual-p5 | The four units pair answers 19-22 with 23-26 in the same values and do not provide separate keys/units for items 23-26. | Create eight answer_units, one for each missing Turkish/Tatar word pair from 19 through 26. |
+| 2007-team-1 | fix_needed | critical | prompt | iol-2007-team-p1 | The genealogical tree diagram on problem page 1 is omitted from the dataset prompt. The prompt only says a tree is presented, but does not include the empty-box tree structure to fill. | Include a rendered image reference or a faithful textual encoding of the family-tree graph and empty boxes in the prompt. |
+| 2007-team-1 | fix_needed | medium | answer_units | iol-2007-team-p1 | The single full.1 unit is a flattened graphical tree answer for both requested tasks, too coarse for scoring tree reconstruction and the eleven numbered members separately. | Add structured answer units for each tree position/name and for each of the eleven numbered family members. |
+| 2008-individual-1 | fix_needed | medium | answer_units |  | Subparts (a) and (b) each have a single free-text answer unit despite requiring five transcriptions and four orthographic forms respectively. | Split answer_units into one unit per numbered item 12-20. |
+| 2008-individual-2 | fix_needed | high | prompt |  | The two-column verse and word-list layout is interleaved in plain text, and the boxed answer slots are lost, making part b hard to reconstruct from the dataset prompt. | Recreate the stanza and word list with stable line/table structure and explicit blank labels. |
+| 2008-individual-2 | fix_needed | critical | answer_units |  | The part b answer unit does not contain the a-m gap assignments from the official solution; it mostly records leftover words and includes a footer. Full-problem units also omit useful part a scoring units. | Add one unit per filled blank or translation target and separate part a rules/answers from part b final assignments. |
+| 2008-individual-2 | fix_needed | medium | subparts |  | The source has multiple parts, but the dataset only exposes a sub-b variant and no usable sub-a record. | Generate subproblem records for each source subpart with matching target labels. |
+| 2008-individual-3 | fix_needed | medium | answer_units |  | Answer units are too coarse: subpart (a) bundles all correspondences, subpart (b) bundles four lexical meanings, and subpart (c) bundles two translations. | Split (a) by correspondence row, (b) by requested word, and (c) by requested Drehu translation. |
+| 2008-individual-4 | fix_needed | high | prompt | iol-2008-individual-p4 | Copainalá Zoque transcription glyphs are corrupted: PDF characters such as ʌ, ŋ, ʔ, and š appear as 2, N, P, and ˇ/letter sequences in the prompt. | Manually transcribe the problem with the actual IPA-like characters shown in the PDF. |
+| 2008-individual-4 | fix_needed | high | canonical | iol-2008-individual-p4 | The same transcription corruption appears in canonical answers, changing forms such as -kʌsmʌ, ʔas, ŋgumguy, and šeh. | Correct solution_text/canonical answer forms from the rendered solution page. |
+| 2008-individual-4 | fix_needed | critical | answer_units | iol-2008-individual-p4 | The answer_units capture numbered suffix-rule bullets rather than the requested final answers for parts (a) and (b), so scoring would not check the actual task outputs. | Replace units with the six English translations in part (a) and six Copainalá Zoque translations in part (b). |
+| 2008-individual-5 | fix_needed | high | answer_units |  | The answer_units omit the first answer in each subpart: item 13 in part (a) and item 19 in part (b) are present in the canonical solution but absent from units. | Add units for 13 "The wolf saw your shaman" and 19 "Angatkuup aanniqtaatit" across the full, subproblem, and partial records. |
+| 2008-team-1 | fix_needed | medium | answer_units |  | Several answer units bundle multiple numbered outputs, especially h (six missing tones), i (four Cantonese readings), and j (eleven Mandarin readings), so scoring is too coarse. | Split h, i, and j into one unit per numbered transcription/tone answer while keeping explanatory subparts as separate rubric units. |
+| 2009-individual-1 | fix_needed | medium | answer_units | iol-2009-individual-p1 | The full-problem record uses one free-text answer unit for all eight requested translations plus explanation, which is too coarse for RL scoring. | Create separate units for each item in parts (a) and (b), with explanation/rules either separate or optional. |
+| 2009-individual-1 | fix_needed | low | prompt | iol-2009-individual-p1 | The language note/footer text is garbled, although the core Sulka examples and questions are readable. | Clean the source note text when rebuilding the prompt. |
+| 2009-individual-2 | fix_needed | critical | prompt |  | The N'Ko-script column present in the source problem is omitted from the dataset prompt, and the script note is collapsed into garbled run-on text. | Rebuild the prompt from rendered pages so both the N'Ko and Roman-script columns and the explanatory note are preserved. |
+| 2009-individual-2 | fix_needed | high | canonical |  | The official solution text does not faithfully represent the script-side answers from the PDF, so solved forms are incomplete in canonical and partial records. | Transcribe the official solution with the N'Ko forms or an explicit source-faithful notation. |
+| 2009-individual-2 | fix_needed | medium | answer_units |  | Answer units are coarse and one part b unit includes footer text. | Split the official answers into one unit per requested item and remove headers/footers. |
+| 2009-individual-3 | fix_needed | medium | prompt | iol-2009-individual-p3 | The phonetic note in the prompt is line-smashed into an unreadable fragment around the definitions of c, E, h, and N. | Reformat the note as normal prose from the source PDF. |
+| 2009-individual-3 | fix_needed | medium | answer_units | iol-2009-individual-p3 | All six Burmese child name/date answers are in a single free-text unit with reasoning, which is too coarse for RL scoring. | Create one answer unit per requested child/date assignment. |
+| 2009-individual-4 | fix_needed | high | prompt |  | Old Indic diacritics and the explanatory note are badly garbled: the model-facing prompt uses lossy forms like v´r.k-a- and the note about ṇ, ṣ, ṭ, and ṛ is scrambled, even though stress/diacritics are central to the problem. | Manually transcribe the Sanskrit stems and note using proper Unicode diacritics from the PDF. |
+| 2009-individual-4 | fix_needed | medium | canonical |  | Canonical answer forms use decomposed/placeholder stress and length notation rather than the PDF glyphs. | Normalize canonical solution forms to the PDF’s Unicode diacritics and stress marks. |
+| 2009-individual-4 | fix_needed | medium | answer_units |  | Part (b) is a single unit containing all stressed stems, too coarse for scoring individual answers. | Split part (b) into one unit per stem and keep part (a) as a separate explanatory unit. |
+| 2009-individual-5 | fix_needed | high | answer_units |  | The units omit item 18 from part (a) and item 24 from part (b), even though both are in the canonical solution. | Add separate units for 18 and 24 in the full, subproblem, and partial records. |
+| 2009-individual-5 | fix_needed | low | prompt |  | The Nahuatl note is garbled into a fused line beginning "TheClassicalNahuatl...". | Repair the note text from the PDF so it reads as the Classical Nahuatl background and orthography note. |
+| 2009-team-1 | fix_needed | medium | prompt | iol-2009-team-p1 | The PDF highlights frequent Vietnamese words in bold inside the reading units, and the prompt explicitly refers to highlighted words; the dataset prompt flattens away that highlighting. | Preserve highlighted target words with markup, e.g. bold Markdown or another explicit token-level marker. |
+| 2009-team-1 | fix_needed | medium | answer_units | iol-2009-team-p1 | The full answer is one giant unit covering translations for ten reading units, which is far too coarse for scoring. | Split answer units by reading unit and preferably by sentence or highlighted vocabulary item. |
+| 2010-individual-1 | fix_needed | medium | answer_units | iol-2010-individual-p1 | A single full-problem answer unit contains all required Budukh table fills and rules, making scoring too coarse. | Split answer_units by each required vacant cell/form and keep rules separate from final filled forms. |
+| 2010-individual-1 | fix_needed | low | provenance |  | The manifest problem page range includes page 2, which contains later problems rather than Problem 1. | Set the problem page range to page 1 only for this source problem. |
+| 2010-individual-2 | fix_needed | medium | answer_units |  | Answer units are too coarse for RL scoring: a.1 bundles many correspondences, b.1 bundles multiple equations, and c.1 bundles all Drehu forms plus footer text. | Split answer_units into one required unit per correspondence, equation, or translation target, and remove footer text. |
+| 2010-individual-3 | fix_needed | critical | prompt |  | The Blissymbolics symbols are omitted from the model-facing prompt; the prompt lists only English words and blank tasks, so subparts (a) and (b) are not answerable. | Include rendered images or an exact textual/vector representation of every Blissymbolic symbol in the examples and target questions. |
+| 2010-individual-3 | fix_needed | critical | canonical |  | The canonical solution and shown solutions describe symbol compositions in words but do not preserve the actual Blissymbolic glyphs needed for source fidelity. | Store the official symbol glyphs or linked rendered crops alongside the textual composition descriptions. |
+| 2010-individual-3 | fix_needed | medium | answer_units |  | Answer units are large composition tables rather than individual symbol correspondences/outputs. | Split by each requested symbol correspondence and each requested produced Blissymbolic word. |
+| 2010-individual-3 | fix_needed | low | provenance |  | The cited page ranges include adjacent Problem 4 pages although Problem 3 content is on page 2. | Narrow problem and solution ranges to the page containing Problem 3, while retaining any image assets from that page. |
+| 2010-individual-4 | fix_needed | high | answer_units | iol-2010-individual-p4 | The full-problem answer_units cover only part (c); answers for parts (a) and (b) are present in canonical text but unscored. | Add units for the synthesized polypeptide answer in (a) and the valid mRNA-sequence choices/count in (b). |
+| 2010-individual-4 | fix_needed | medium | canonical |  | The solution table and part (b) combinatorics are linearized with glyph artifacts such as private-use braces and !/" markers, making the canonical solution hard to verify mechanically. | Manually clean the codon table and part (b) answer formatting from the solution PDF. |
+| 2010-individual-5 | fix_needed | medium | answer_units |  | Part (a) stores all gap fills as one unit and part (d) stores both requested plural translations as one unit, which is too coarse for RL scoring. | Split part (a) into individual gap/table-row units and part (d) into separate units for "elms" and "angles". |
+| 2010-team-1 | fix_needed | medium | answer_units | iol-2010-team-p1 | The single full.1 unit covers all phrase translations and the large Mongolian lexicon table, which is too coarse for RL scoring. | Split part (a) phrase translations into separate units and represent part (b) lexicon entries as separate optional/graded units. |
+| 2011-individual-1 | fix_needed | medium | answer_units |  | The Menominee answer units are grouped by subpart, but subpart (a) has three translations and subpart (b) has four translations, so the units are too coarse for RL scoring. | Create one answer_unit for each requested form/translation in parts (a) and (b). |
+| 2011-individual-2 | fix_needed | critical | answer_units |  | The answer_units capture only two rule lines and omit the seven actual gap transcriptions for part a; there is no record or unit coverage for part b. | Regenerate units from the official solution with one answer per gap for part a and add the missing part b target coverage. |
+| 2011-individual-2 | fix_needed | medium | canonical |  | The side-by-side official solution layout is interleaved in extraction, mixing part a answers and part b rules. | Clean the canonical solution into separate part a and part b sections. |
+| 2011-individual-2 | fix_needed | medium | subparts |  | The source has more than one part, but the dataset only includes the full problem and sub-a record. | Generate subproblem and partial records that match all source subpart labels. |
+| 2011-individual-3 | fix_needed | medium | answer_units |  | Subpart (a) has three translations in one unit and subpart (c) has four Vai translations in one unit, which is too coarse for partial-credit RL scoring. | Split answer units by each requested phrase/translation in (a) and (c). |
+| 2011-individual-4 | fix_needed | medium | answer_units |  | Each subpart is a single large answer unit even though part (a) has many correspondences and parts (b)/(c) each have multiple translations; too coarse for RL scoring. | Split part (a) by word correspondence and parts (b)/(c) by requested word or phrase. |
+| 2011-individual-5 | fix_needed | critical | prompt |  | The barcode source requires diagrams, barcode images, and a partially filled grid, but the dataset prompt contains only arrows/labels and omits the visual information needed to solve the problem. | Attach rendered page images or encode the barcode/grid data structurally; otherwise exclude this source from text-only records. |
+| 2011-individual-5 | fix_needed | critical | subparts |  | Official part (c) is not represented as its own target; the sub-b and partial-b records ask only to draw barcode (b), but their canonical answer is the textual part (c) Norway/code answer. | Create separate records for (b) and (c), or mark drawing-only (b) unsupported; align each target with its own canonical answer. |
+| 2011-individual-5 | fix_needed | high | answer_units |  | Part (a) units start at item 2 and omit item 1, and the current b unit combines empty (b) text with the part (c) answer. | Add item 1 to part (a) units and rebuild units after separating (b) and (c). |
+| 2011-team-1 | fix_needed | high | answer_units | iol-2011-team-p1-sub-b | Part (b) answer units are incomplete: they omit the deleted words 'jarau, lagau' and several added/deleted macron corrections, even though the canonical text includes them. | Create answer units for the two deleted words, three changed letters, four added-macron removals, and five deleted-macron restorations. |
+| 2011-team-1 | fix_needed | high | answer_units | iol-2011-team-p1 | The full-problem answer units inherit the incomplete part (b) coverage, so the full record cannot score all requested corrections. | Use the same complete granular part (b) units in the full and partial-through-a records. |
+| 2012-individual-1 | fix_needed | high | answer_units |  | Answer units omit the first required item in multiple subparts: part (b) starts at 18 instead of 17 and part (c) starts at 21 instead of 20. The full record has the same omissions. | Add answer_units for 17 and 20 and verify all Dyirbal numbered answers are represented individually. |
+| 2012-individual-2 | fix_needed | medium | prompt |  | A boxed comparison in the source is collapsed into the surrounding numeral list, weakening the table/layout fidelity needed to read the data. | Represent the tables and boxed comparison with explicit table markup or clear separated lines. |
+| 2012-individual-2 | fix_needed | medium | answer_units |  | The part a and part b answer units each bundle multiple final answers, which is too coarse for item-level scoring. | Split units so each requested numeral or expression has its own required answer unit. |
+| 2012-individual-3 | fix_needed | medium | answer_units |  | Answer units are too coarse and partly layout-scrambled: subpart (a) is one correspondence table unit, while (b) and (c) each bundle two answers into one unit. | Split (a) by Basque-English correspondence and split (b)/(c) by requested sentence. |
+| 2012-individual-3 | fix_needed | medium | canonical |  | The canonical subpart (a) correspondence table is flattened into a hard-to-read layout and subpart (d) includes solution footer text. | Reformat the solution table and remove page footers from canonical/subpart answers. |
+| 2012-individual-3 | fix_needed | low | provenance |  | The page ranges include adjacent Problems 2 and 4 pages around Problem 3. | Use page spans or text spans that isolate Problem 3 where possible. |
+| 2012-individual-4 | fix_needed | high | answer_units |  | The first answer on each labeled solution line is omitted from answer_units: part (a) misses 13, part (b) misses 16, and part (c) misses the first target answer. | Regenerate units so answers 13, 16, and the first part (c) translation are included, and split every numbered target separately. |
+| 2012-individual-4 | fix_needed | medium | canonical |  | Part (b) canonical includes explanatory rule text after the target translations, which is acceptable for solution_text but noisy as the canonical answer for the subproblem/partial variant. | Keep rationale in solution_text but set canonical for subproblem records to the requested translations only. |
+| 2012-individual-5 | fix_needed | high | answer_units |  | Subpart (b) has a single unit "poga: hole" and omits the other four requested translations; parts (a), (c), and (d) are also grouped as coarse multi-answer units. | Create separate units for every requested body-part correspondence, every part (b) translation, every part (c) translation, and each theoretical form in part (d). |
+| 2012-team-1 | fix_needed | critical | prompt | iol-2012-team-p1 | The prompt omits the entire list of 57 Lao country names visible on the source PDF page, leaving only the sentence 'Here are the names of 57 countries in Lao'. | Add a rendered image or faithful Unicode/manual transcription of all 57 Lao names to the prompt. |
+| 2012-team-1 | fix_needed | high | answer_units | iol-2012-team-p1 | The answer units are only four large, mis-grouped chunks of the 57-country solution rather than item-level country/pronunciation answers. | Split into one unit per country identification and pronunciation, or at least one unit per numbered country. |
+| 2013-individual-1 | fix_needed | high | answer_units |  | Part (c) is represented by a single unit containing only the last two translations, omitting most requested Yidiny outputs. Part (b) is also a single unit for four outputs. | Create separate answer_units for every requested item in parts (a), (b), and (c), especially all eight translations in part (c). |
+| 2013-individual-2 | fix_needed | medium | answer_units |  | The answer units are too coarse for RL scoring: part a, b, and c each bundle several final answers into one unit, and c includes footer text. | Split the official answers into one unit per requested item and strip page footer text. |
+| 2013-individual-3 | fix_needed | critical | answer_units | iol-2013-individual-p3 | The five answer units are rule fragments only; none of the ten requested Piraha pronunciations are represented as answer units. | Create answer units for each requested item: xaaibi, xaapisi, xitiixisi, bigi, kagahoaogii, kaibai, kapiigaiitoii, poogaihiai toio, xabagi kapioxio, and xabagi xogiai. |
+| 2013-individual-3 | fix_needed | medium | prompt | iol-2013-individual-p3 | The prompt table has layout artifacts, including a misplaced "toucan" label near the kaoaibogi row and page-footer noise. | Reformat the training and target tables from the source PDF and remove footers. |
+| 2013-individual-3 | fix_needed | low | provenance | iol-2013-individual-p3 | The problem page range includes page 4, which is the next problem. | Narrow the problem page range to page 3 for this source. |
+| 2013-individual-4 | fix_needed | critical | subparts | iol-2013-individual-p4-sub-a | Subpart solution splitting is wrong: sub-a asks for items 12 and 13, but the canonical omits item 12 and includes answers for items 14 and 15, which belong to part (b). | Manually split the solution so part (a) contains answers 12 and 13 only. |
+| 2013-individual-4 | fix_needed | critical | canonical | iol-2013-individual-p4-sub-b | Sub-b asks for items 14-17, but its canonical contains only items 16 and 17 because answers 14 and 15 were attached to part (a). | Move answers 14 and 15 into the part (b) canonical and units. |
+| 2013-individual-4 | fix_needed | high | subparts | iol-2013-individual-p4-partial-through-a-target-b | The partial-through-a variant leaks part (b) answers 14 and 15 in the shown official solution for part (a). | After correcting subpart splits, regenerate partial variants so shown_solutions contain only prior target answers. |
+| 2013-individual-5 | fix_needed | medium | answer_units | iol-2013-individual-p5 | The full record has one large unit containing the whole explanation/table instead of separate scoreable correspondences for the six numbered words. | Split the answer into units for refrigerator, cow, bed, butterfly, spoon, and cat, with their corresponding numbers/locations. |
+| 2014-individual-1 | fix_needed | high | prompt |  | The source uses the letter sequence 'fi' in many Benabena forms, but the PDF text extraction lost the ligature/control character, producing blanks such as 'noli 'ibe', 'no ne', 'noni be', and 'li latagihe'. | Manually restore all lost 'fi' sequences in examples and targets, e.g. nolifi'ibe, nofine, nifila'ibe, nonifibe, lifilatagihe. |
+| 2014-individual-1 | fix_needed | high | canonical |  | The same 'fi' loss appears in the solution rules and answers, including the pierce root and answers such as kifilune and filanagihe. | Clean the canonical solution and answer units against the rendered solution PDF. |
+| 2014-individual-1 | fix_needed | medium | answer_units |  | Subpart (a) bundles five translations in one unit and subpart (b) bundles four forms in one unit. | Split each requested item into its own answer unit. |
+| 2014-individual-2 | fix_needed | critical | answer_units | iol-2014-individual-p2 | The only answer unit is a fragment of the rule table (fruit suffixes) rather than the filled Kiowa noun table, so the expected answer is effectively unscoreable. | Replace the full.1 unit with separate answer_units for each question-mark cell in the Kiowa table. |
+| 2014-individual-2 | fix_needed | low | prompt | iol-2014-individual-p2 | Minor text extraction damage appears in the prompt, such as the word fish rendered as sh. | Clean ligature/font extraction artifacts in the prompt text. |
+| 2014-individual-3 | fix_needed | critical | prompt |  | The source problem depends on a family-tree diagram, but the dataset prompt reduces the diagram to scattered punctuation and letters, losing the relationships needed to solve the task. | Include the rendered diagram as an image asset or transcribe the family tree into explicit relationship edges. |
+| 2014-individual-3 | fix_needed | critical | canonical |  | The official solution family tree is also collapsed; the part a answer unit is effectively just '(a) 1' rather than the reconstructed tree. | Rebuild the canonical and part a answer_units from the rendered solution diagram. |
+| 2014-individual-3 | fix_needed | high | subparts | iol-2014-individual-p3-partial-through-a-target-b | The partial variant shows solved part a as an unusable collapsed tree, so the visible solved context is not faithful. | Use a diagram image or explicit relationship transcription for shown solved part a. |
+| 2014-individual-4 | fix_needed | critical | answer_units |  | Answer units omit major required answers: subpart (a) lacks item 8 and truncates items 9/10; subpart (b) lacks item 11 and the second line of item 12; subpart (c) includes footer text. | Rebuild answer units from the canonical answer so every numbered English/Engenni sentence and base-form explanation is separately represented. |
+| 2014-individual-4 | fix_needed | medium | prompt |  | The note explaining tone marks loses the mark symbols, and several pronoun glosses are mashed together as strings like "hethe". | Re-transcribe the note and affected English glosses with correct spacing and tone mark names/symbols. |
+| 2014-individual-4 | fix_needed | low | canonical |  | Canonical prose contains OCR artifacts such as missing letters in "affected"/"affirmative" and includes solution footnote/footer noise in subpart answers. | Clean OCR artifacts and remove footers from canonical and subpart_solutions. |
+| 2014-individual-5 | fix_needed | medium | prompt |  | The fi ligature is dropped in several English words, producing "eld edge" and "good eld" instead of "field edge" and "good field" in prompt/canonical text. | Normalize PDF ligatures so field/simplified are restored. |
+| 2014-individual-5 | fix_needed | medium | answer_units |  | Subpart units are large blocks: part (a) has all correspondences as one unit, and parts (b)/(c) each group several answers. | Split answer_units by each correspondence/translation requested. |
+| 2014-team-1 | fix_needed | high | answer_units | iol-2014-team-p1 | Although there are 30 units, multi-line Armenian articles are truncated to their first line in answer_units; for example article 2 and articles 29-30 omit most of their official solution text. | Store the complete Armenian text for each numbered article, or split long articles into complete sub-units without dropping continuation lines. |
+| 2015-individual-1 | fix_needed | high | answer_units | iol-2015-individual-p1 | The full-problem prompt asks parts (a), (b), and (c), but the answer_units cover only b and c; part (a)'s numerical equalities are missing from scoring. | Add granular answer units for all sixteen items in part (a), and add a sub-a record if subproblem coverage is intended. |
+| 2015-individual-1 | fix_needed | medium | answer_units |  | Parts (b) and (c) each bundle two requested translations in a single unit. | Split b into separate units for 42 and 494, and c into separate units for 43 and 569. |
+| 2015-individual-2 | fix_needed | critical | prompt | iol-2015-individual-p2-sub-b | The standalone part (b) prompt omits the second block of Beslenei examples that the source provides before question (b), so the target is not answerable from the shown data. | Include the additional Beslenei forms block before part (b) in the sub-b prompt. |
+| 2015-individual-2 | fix_needed | critical | prompt | iol-2015-individual-p2-partial-through-a-target-b | The partial variant targeting part (b) also omits the second Beslenei examples block, so the partial record remains not answerable despite showing part (a). | Include the omitted examples in the partial-through-a target-b prompt. |
+| 2015-individual-2 | fix_needed | medium | answer_units |  | Each subpart is scored as a single free-text unit even though parts (a), (b), and (c) each require multiple separate translations/forms. | Split answer_units by each requested Kabardian/Beslenei item. |
+| 2015-individual-3 | fix_needed | high | answer_units |  | The Soundex rule units are incomplete: step 1 is missing, step 2 is truncated mid-sentence, and step 5 is truncated. Part b and c units bundle many codes into single units. | Restore all Soundex algorithm steps and split surname/code outputs into item-level answer_units. |
+| 2015-individual-3 | fix_needed | low | prompt |  | The source includes a sample census-card image, but the dataset prompt keeps only surrounding text/source URL context. | Include the image or a concise transcription of the relevant card fields if they are needed as context. |
+| 2015-individual-4 | fix_needed | high | answer_units |  | Answer units omit the first answer in both subparts: item 18 in (a) and item 24 in (b). The unit keys also use solution-local numbering 2-6 and 8-11 instead of prompt item numbers 18-28. | Add the missing units for 18 and 24 and renumber/split units to match prompt items 18-28. |
+| 2015-individual-4 | fix_needed | low | provenance |  | Problem and solution ranges include adjacent Problem 5 pages, and subpart prompts include page-footer text. | Narrow page ranges or text spans to Problem 4 and strip footers. |
+| 2015-individual-5 | fix_needed | high | answer_units | iol-2015-individual-p5 | Full-problem answer_units only cover part (b) and omit the descriptive answer for part (a). Within part (b), item 36 is also omitted because it is on the same line as the (b) marker. | Add a part (a) meter-structure unit and include item 36 plus one unit for each 36-45 genuine/rearranged judgment. |
+| 2015-team-1 | fix_needed | high | subparts |  | The official problem includes part (d) asking for a bilingual dictionary and grammar, but derived records and target_subparts only cover (a)-(c); the full record units also omit the dictionary/grammar answer material. | Add a part (d) record/partial sequence and answer_units for the dictionary/grammar solution, or explicitly mark that open-ended part unsupported while keeping provenance clear. |
+| 2016-individual-1 | fix_needed | critical | prompt | iol-2016-individual-p1 | The map of Salu Leang hamlets, roads, river, bridges, stream direction, and relative heights is missing from the dataset prompt; only the legend text survives. | Include a rendered map image or a faithful textual graph with hamlet positions, road/river connections, bridges, and elevation numbers. |
+| 2016-individual-2 | fix_needed | high | answer_units |  | Part (c) answer_units omit item 1 (the acute-sign function) in the full, sub-c, and partial-through-ab records. Part (b) is also one unit for two answers. | Add a unit for c.1 and split part (b) into two units, then verify all 13 matches in part (a) remain covered. |
+| 2016-individual-3 | fix_needed | critical | answer_units |  | Answer units are on the wrong side of the translation task: part a units contain source Kunuz Nubian sentences instead of English translations and miss item 11; part b units contain English prompts instead of Kunuz Nubian translations and miss item 16. | Regenerate answer_units from the official solution so each target item stores the requested translated answer, including items 11 and 16. |
+| 2016-individual-4 | fix_needed | high | answer_units |  | Answer units omit the first required answer in every subpart: (a) item 1, (b) item 1 plus the synonym explanation, and (c) item 1. | Add units for gu-na vaala = canoe, vaala = canoe with the synonym explanation, and pig = bak; align keys with source item numbering. |
+| 2016-individual-4 | fix_needed | low | canonical |  | Canonical/subpart answers include page-footer text from the solution PDF. | Remove footer text from canonical answers. |
+| 2016-individual-5 | fix_needed | high | answer_units |  | The first numbered answer in each subpart is missing from answer_units: item 1 in part (a) and item 4 in part (b) are present in canonical text but not unitized. | Regenerate units to include items 1-3 and 4-9 separately. |
+| 2016-team-1 | fix_needed | critical | other | iol-2016-team-p1 | The source is an audio matching problem requiring 114 audio files, but the dataset record has no audio assets and metadata has has_audio=false, making the prompt unanswerable. | Include the audio files/references in the dataset task or exclude this source from non-audio records. |
+| 2016-team-1 | fix_needed | high | answer_units | iol-2016-team-p1 | All 114 audio-word matches are stored as a single full answer unit. | Create one unit per audio-word match and a separate unit for the observations requested in part (b). |
+| 2017-individual-1 | fix_needed | high | prompt |  | Superscript exponents in the Birom equalities are flattened or line-broken, e.g. powers are shown as adjacent digits or misplaced words, and the a^k explanatory formula is garbled. | Represent powers explicitly with caret notation or Unicode superscripts in every equality and in the formula note. |
+| 2017-individual-1 | fix_needed | high | answer_units |  | Part (a) answer units omit equality 1 (52 + 3 + 4 = 32); parts (b) and (c) also bundle multiple requested outputs into single units. | Add the missing item 1 unit and split parts (b) and (c) into one unit per number/equality/output. |
+| 2017-individual-2 | fix_needed | high | answer_units |  | Answer units omit the first required answer in part (a) (1. a-bang) and part (c) (1. pig), and part (b) is a single unit for two translations. | Add the missing first units and split all multi-answer units so every requested Abui item is scoreable. |
+| 2017-individual-3 | fix_needed | critical | answer_units |  | Answer units are on the wrong side of the translation task: part a units contain Kimbundu source sentences instead of English translations and miss item 12; part b units contain English prompts instead of Kimbundu translations and miss item 16. | Regenerate answer_units from the official solution with one requested translated answer per numbered target. |
+| 2017-individual-4 | fix_needed | critical | prompt | iol-2017-individual-p4 | The Khom script column and source illustrations are omitted from the prompt; the task asks to fill question-mark cells including script cells, so the record is not answerable from text alone. | Include rendered table/image assets or exact Khom glyph representations for all source and target cells. |
+| 2017-individual-4 | fix_needed | critical | canonical | iol-2017-individual-p4 | The canonical answer represents Khom-script answers with arrows and blank positions rather than actual glyphs, losing the official written answers. | Store the rendered Khom glyph answers or a faithful transcription/image crop for each target cell. |
+| 2017-individual-4 | fix_needed | medium | answer_units | iol-2017-individual-p4 | All target cells are in one large free-text unit, too coarse for scoring even if glyphs are restored. | Create one answer unit per filled cell or per row/cell target. |
+| 2017-individual-5 | fix_needed | high | answer_units |  | Part (b)/(c) answer_units omit item 1 (laradi) and item 8 (heart) because those answers share lines with subpart markers. | Include all items 1-14 as separate units, with part (a) as its own literal-translation unit. |
+| 2017-team-1 | fix_needed | critical | prompt | iol-2017-team-p1 | The emoji list needed for solving is corrupted in the prompt: many emoji render as punctuation, letters, blanks, or private-use placeholders instead of the actual glyphs. | Use rendered problem pages or verified Unicode emoji data for the numbered emoji list. |
+| 2017-team-1 | fix_needed | high | canonical | iol-2017-team-p1 | The canonical solution also contains corrupted/missing emoji glyphs in places, such as black circle appearing as a stray letter/blank and some symbol answers lacking the glyph. | Repair canonical emoji mappings from the rendered solution PDF. |
+| 2017-team-1 | fix_needed | medium | answer_units | iol-2017-team-p1 | The entire 87-item solution plus extra vocabulary is one answer_unit. | Split into per-emoji mapping units and separate units for the additional requested descriptions/dictionary/grammar outputs. |
+| 2018-individual-1 | fix_needed | medium | canonical | iol-2018-individual-p1 | The official solution's stress-rule formula is flattened into a largely unreadable text block, losing the visual structure of the PDF. | Manually rewrite the rule formula in clear textual notation or preserve it as a rendered image plus text. |
+| 2018-individual-1 | fix_needed | medium | answer_units | iol-2018-individual-p1 | The single answer unit bundles all twelve stress-marking targets, which is too coarse for RL scoring. | Create one answer unit for each target word's correctly stressed form. |
+| 2018-individual-2 | fix_needed | high | subparts |  | The source problem has parts (a) and (b), but the derived records include only a full record and a sub-b record; there is no subproblem record for part (a). | Add a subproblem record for part (a), or explain intentionally omitted subparts in metadata. |
+| 2018-individual-2 | fix_needed | high | answer_units |  | The full record answer_units cover only part (b) items 8-10, omitting all part (a) answers and part (b) item 7; the sub-b record also omits item 7. | Create answer_units for part (a) items 1-6 and part (b) items 7-10. |
+| 2018-individual-3 | fix_needed | high | answer_units |  | Part a answer_units bundle all fourteen gaps into one unit, and the part b unit omits explanatory content from the official solution. | Split part a into one unit per gap and include the complete part b answer/rationale expected by the official solution. |
+| 2018-individual-3 | fix_needed | medium | prompt |  | The extracted table and language note are linearized and partially mangled, reducing fidelity even though the main data are mostly present. | Clean table structure and note text against the PDF. |
+| 2018-individual-4 | fix_needed | high | answer_units |  | Answer units omit required first answers in translation subparts: item 12 in (b) and item 17 in (c). | Add answer units for 12. biabys atub bobys = 3 betel nuts and 17. 4 betel nuts = biabys atubys atub bobys, preserving source diacritics. |
+| 2018-individual-4 | fix_needed | medium | prompt |  | The subpart (c) note is line-scrambled into text like "15,000Mountainpeople inArapesh...", reducing prompt fidelity. | Reformat the note from the source PDF with normal spacing. |
+| 2018-individual-4 | fix_needed | medium | answer_units |  | Subpart (a)'s multi-part explanation is a single unit; this is coarse for partial credit. | Split subpart (a) into units for word X, examples 6/9/11, and how the word is implied. |
+| 2018-individual-5 | fix_needed | critical | prompt |  | The family-tree diagram is not represented coherently in the text prompt; the PDF geometry/legend needed for part (a) becomes a jumbled sequence of names and symbols. | Provide the family tree as an image attachment or manually encode the graph structure in text. |
+| 2018-individual-5 | fix_needed | high | answer_units |  | Part (b) answer_units only include statement 2 and omit statement 1; part (a) has only a shape sequence rather than a scoreable named family tree. | Create scoreable units for the names in the family tree and both filled statements. |
+| 2018-team-1 | fix_needed | high | answer_units |  | Part (a) is one very large unit for the whole matching table, and part (c) has incorrect units: item 5 is truncated and item 6 combines answers 6 and 7. | Split part (a) into individual match units; repair part (c) so items 5, 6, and 7 are complete and separately keyed. |
+| 2019-individual-1 | fix_needed | high | answer_units |  | Answer units omit the first requested item in each subpart: item 12 for part (a) and item 18 for part (b), although both appear in the canonical answer. | Add answer units for item 12 ('Om benu aneen' translation) and item 18 ('We will go away and they will come' in Yonggom), and propagate them to full/sub/partial records. |
+| 2019-individual-2 | fix_needed | high | subparts |  | Source parts (c) and (d) are merged into records labeled only target c; the target question says solve (c) only but includes the (d) translation requests, and no separate d record exists. | Split source parts (c) and (d) into distinct target_subparts/records or relabel the merged target as c+d consistently. |
+| 2019-individual-2 | fix_needed | high | answer_units |  | Answer units repeatedly omit the first item of each section: a.1 for source item 1, b.1 for item 19, c.1 for item 26, and the first part (d) item 30 are missing from scoring. | Add units for source items 1, 19, 26, and 30 and verify all Yurok outputs are represented individually. |
+| 2019-individual-3 | fix_needed | critical | prompt |  | The source uses Book Pahlavi glyphs, but the dataset exposes internal font-encoding strings instead of the visual forms and ligatures needed to solve the problem. | Use rendered image crops or a faithful custom transcription for the Book Pahlavi forms. |
+| 2019-individual-3 | fix_needed | critical | canonical |  | The canonical solution and shown partial solutions use the same font-encoding artifacts, so solved script answers are not source-faithful. | Rebuild solution_text and partial solved sections from rendered pages. |
+| 2019-individual-3 | fix_needed | high | answer_units |  | Answer units bundle table rows, include stray artifacts and footers, and do not provide clean item-level Pahlavi answers. | Create one clean unit per requested subpart item using the repaired script representation. |
+| 2019-individual-4 | fix_needed | medium | answer_units | iol-2019-individual-p4 | All nine North-dialect and nine Coast-dialect gap fills are bundled into one free-text answer unit, which is too coarse for RL scoring. | Create separate answer units for each requested form or each row/dialect cell. |
+| 2019-individual-4 | fix_needed | low | provenance | iol-2019-individual-p4 | Problem and solution page ranges include adjacent problems; the prompt also contains a page footer between the example table and target table. | Use ranges/text spans that isolate Problem 4 and strip footers. |
+| 2019-individual-5 | fix_needed | high | answer_units |  | Answer_units omit the first item in parts (a) and (c): item 16 and item 20 are in canonical text but not scored. Part (b) is also a single list unit. | Create separate units for items 16-19, the week-order list, and items 20-24. |
+| 2019-team-1 | fix_needed | critical | other | iol-2019-team-p1 | The source states that some entries require videos, but the dataset provides only text/PDF extraction and no video assets, so the full team problem is not answerable as given. | Include the designated videos or exclude/mark the video-dependent source unsupported. |
+| 2019-team-1 | fix_needed | high | answer_units | iol-2019-team-p1 | All entries 49-77 and their scores/descriptions are stored as one huge full answer unit. | Create one scoreable unit per numbered answer, with score and description kept together. |
+| 2021-individual-1 | fix_needed | medium | answer_units |  | Part (a) is one unit for three digit answers and part (b) is one unit for five Ekari outputs, so scoring is too coarse. | Split part (a) into units for each numeral and part (b) into units for 1, 19, 26, 104, and 292. |
+| 2021-individual-2 | fix_needed | high | subparts |  | Source part (d) is merged into records labeled target c, while the target question for sub-c asks only part (c); no separate d record exists. | Split part (d) into its own subproblem/partial target or label the merged record as covering both c and d. |
+| 2021-individual-2 | fix_needed | high | answer_units |  | Answer units omit the first answer in parts (a), (b), (c), and (d), and the c units combine c and d answers into single paired values. | Add units for source items 1, 11, 20, and 26, and separate part (c) English translations from part (d) Zuni translations. |
+| 2021-individual-3 | fix_needed | critical | answer_units |  | Answer units are on the wrong side of the translation task: part a units contain Kilivila prompts instead of English translations and miss item 14; part b units contain English prompts instead of Kilivila translations and miss item 19. | Regenerate answer_units from the official solution with one target-side answer per numbered item. |
+| 2021-individual-4 | fix_needed | high | answer_units |  | Subpart (b) answer units omit item 1; units start at key 2 even though the canonical has the form for rutu/chewing-stick. | Add an answer unit for item 1 and align keys with source numbering. |
+| 2021-individual-4 | fix_needed | low | canonical |  | The rule table in the canonical contains glyph artifacts shown as question marks and includes page-footer text. | Clean or replace the rule-table extraction and remove footer text. |
+| 2021-individual-5 | fix_needed | critical | prompt | iol-2021-individual-p5 | The two-column table with bracketed alternative translations is scrambled in the prompt; brackets and alternatives are interleaved with unrelated entries, making the text record unreliable and not fully answerable. | Manually encode the Rikbaktsa table row-by-row, preserving bracketed alternative translations. |
+| 2021-individual-5 | fix_needed | high | canonical | iol-2021-individual-p5 | The solution’s paradigm table and several answers are similarly scrambled by linearization, including bracket artifacts in part (b). | Clean the solution_text/canonical from the rendered solution page. |
+| 2021-individual-5 | fix_needed | medium | answer_units | iol-2021-individual-p5 | The full record has a single full-answer unit, too coarse for multiple requested translations. | Split into one unit per part (a) and part (b) requested item. |
+| 2021-team-1 | fix_needed | high | answer_units |  | Several parts are too coarse as single-table/list units, and part (f) units omit item 1, item 14, and the N-Q translation answers while truncating items 12 and 13. | Split parts (a)-(g) into granular units and rebuild part (f) to include every numbered/lettered answer completely. |
+| 2022-individual-1 | fix_needed | high | subparts |  | The source PDF has separate parts (b) and (c), but the dataset collapses both into target subpart b and has no separate c record; sub-b prompts ask both 'Translate into English' and 'Translate into Ubykh'. | Create a distinct c subproblem/partial target for items 13-17 and keep b limited to items 10-12. |
+| 2022-individual-1 | fix_needed | high | answer_units |  | The combined b answer units omit the first item of each source subpart: item 10 'aš'əntʷən — he gives him to us' and item 13 'they give youpl to me — ŝʷəsənatʷan'. | Add missing units for items 10 and 13 and relabel all b/c units to match the source subparts. |
+| 2022-individual-2 | fix_needed | high | subparts |  | Source parts (b) and (c) are merged into records labeled only target b; the prompt says solve (b) only while including the (c) Alabama translations. | Split part (c) into its own target_subpart/record or relabel the merged record as b+c. |
+| 2022-individual-2 | fix_needed | high | answer_units |  | Answer units omit first answers in multiple sections: source item 1 in part (a), item 27 in part (b), and item 30 in part (c). | Add the missing units and align b/c unit ids with the source subpart labels. |
+| 2022-individual-3 | fix_needed | high | prompt | iol-2022-individual-p3 | The two-column task list is interleaved in extraction, with later items appearing inside or after earlier item text, which makes the prompt hard to follow. | Recreate the task list with explicit columns or separate numbered lines. |
+| 2022-individual-3 | fix_needed | critical | answer_units | iol-2022-individual-p3 | Answer units omit source targets 17 and 23, truncate the answer for 26, and degrade click-symbol forms to ASCII punctuation in some entries. | Regenerate all answer_units from the official solution, preserving click characters and complete text for every numbered item. |
+| 2022-individual-3 | fix_needed | medium | canonical | iol-2022-individual-p3 | The word-order chart/rule section is mangled by brace and line-layout extraction, although many final answers are present. | Clean the canonical rule chart into a readable structured form. |
+| 2022-individual-4 | fix_needed | critical | prompt | iol-2022-individual-p4 | The Arabana family-tree diagrams are linearized into symbols and numbers, losing marriage, sibling, and parent-child structure. The record is not answerable without the rendered PDF. | Include rendered family-tree images or a structured tree representation with all relationships. |
+| 2022-individual-4 | fix_needed | critical | canonical | iol-2022-individual-p4 | The canonical solution is a lossy one-unit diagram dump rather than structured A-X position mappings and blank (i)-(ix) answers. | Represent the solution as explicit mappings for A-X to 1-24 and separate values for blanks i-ix, with any diagrams preserved as images if needed. |
+| 2022-individual-4 | fix_needed | medium | answer_units | iol-2022-individual-p4 | The entire solution is one answer unit, too coarse for scoring the 24 mappings and 9 blanks. | Create one unit per family-member mapping and one unit per blank. |
+| 2022-individual-5 | fix_needed | medium | answer_units |  | Part (b) has 18 blanks but is represented as one answer unit, too coarse for RL scoring. | Split part (b) into one unit per numbered blank and keep part (a) as a separate unit. |
+| 2022-individual-5 | fix_needed | low | prompt |  | The source is table-heavy and the linearized prompt moves some column labels/entries out of visual order; values remain present but require careful manual cleanup for fidelity. | Prefer a manually transcribed Markdown table or image-backed prompt for this problem. |
+| 2022-team-1 | fix_needed | critical | prompt |  | The Manchu script material in the prompt is garbled by font extraction, especially the initial manuscript excerpt; this is core data for the problem. | Provide rendered images or manually verified Unicode/structured Manchu script text for all Manchu-script passages. |
+| 2022-team-1 | fix_needed | critical | canonical |  | The canonical answer for part (c) is also garbled Manchu script and includes extraction artifacts rather than reliably readable target script. | Repair part (c) from the rendered solution PDF or include the answer as an image/validated Unicode text. |
+| 2022-team-1 | fix_needed | high | answer_units |  | Units omit answers b54, f90, and h112; part (f) units capture only source Manchu prompts rather than full translated dialogue answers, and some units include headers/chart spillover. | Rebuild answer_units from the canonical solution with one complete unit per requested item and no page headers or explanatory charts inside item values. |
+| 2023-individual-1 | fix_needed | critical | prompt | iol-2023-individual-p1-sub-b | The sub-b prompt asks to fill rows 39-48 but omits the entire second table containing rows 19-48 from the source PDF, so the target is not answerable. | Include the second verb-form table (rows 19-48) in the b prompt before the target question. |
+| 2023-individual-1 | fix_needed | critical | prompt | iol-2023-individual-p1-partial-through-a-target-b | The partial-through-a target-b prompt has the same omission of rows 19-48, even though those rows are required for solving b. | Regenerate the partial b prompt with rows 19-48 included after the shown solution for part a. |
+| 2023-individual-2 | fix_needed | high | answer_units |  | Answer units omit source item 17 and item 21, and the part (c) unit values contain only the English prompts after the dash, not the required Apurina translations. | Rebuild answer_units for all items 17-26, preserving both acceptable Apurina alternatives for part (c). |
+| 2023-individual-2 | fix_needed | medium | canonical | iol-2023-individual-p2-sub-b | The part (b) canonical/answer unit is contaminated by page-layout spillover from the following part (c), including an unrelated answer fragment. | Clean the part (b) canonical to contain only the grammatical correction, both valid equivalents, and its English translation. |
+| 2023-individual-3 | fix_needed | critical | answer_units | iol-2023-individual-p3 | Several answer_units are rule fragments rather than requested final answers, and actual official answers for items 18 and 23 are missing. | Replace rule-fragment units with item-level final-answer units and add the missing items 18 and 23. |
+| 2023-individual-4 | fix_needed | high | answer_units | iol-2023-individual-p4 | Answer units omit requested items 22 and 26; units start at 23 for subpart (a) and 27 for subpart (b). | Add units for 22. e-wapamikoyek = as he sees you(pl) and 26. if we ask them = kakwecimayahkwawi, preserving source diacritics. |
+| 2023-individual-4 | fix_needed | low | prompt | iol-2023-individual-p4 | The note for the long-vowel mark is awkwardly formatted as "mark" joined with the macron symbol. | Reformat the note so it clearly says the macron indicates a long vowel. |
+| 2023-individual-5 | fix_needed | medium | answer_units |  | Part (a) is one unit for two digit answers and part (b) is one unit for five Supyire numerals, too coarse for RL scoring. | Create one answer unit per requested number/numeral. |
+| 2023-team-1 | fix_needed | critical | answer_units | iol-2023-team-p1 | The canonical answer contains the full page 57-61 matching key plus the 60-T-54 restoration, but answer_units contain only the 60-T-54 restoration; almost all required mappings are unscored. | Create answer_units for every dictionary-entry match on pages 57-61 and a separate unit for the restored 60-T-54 class. |
+| 2023-team-1 | fix_needed | medium | prompt | iol-2023-team-p1 | The long dictionary tables are flattened with many adjacent entries concatenated, making parts of the prompt hard to parse. | Preserve table layout or encode dictionary entries as structured rows. |
+| 2024-individual-1 | fix_needed | high | answer_units | iol-2024-individual-p1 | The answer_units include explanatory rule-structure bullets as required units and omit actual answer items 19 and 24 from the requested translations. | Remove rule bullets from required answer_units or mark them as rationale, and add answer units for item 19 (kullenan/kulleŋən translation) and item 24 (you(sg) will see them(du)). |
+| 2024-individual-1 | fix_needed | medium | answer_units | iol-2024-individual-p1 | Only one full-problem record exists even though the source has parts (a) and (b), limiting subpart-level evaluation. | Add subproblem records or at least label answer units under source subparts a and b. |
+| 2024-individual-2 | fix_needed | high | subparts |  | Source part (c) is merged into records labeled only target b; target-b prompts say solve (b) only but include the Hadza-to-English and English-to-Hadza tasks together. | Split source part (c) into its own target_subpart/record or relabel merged records as b+c. |
+| 2024-individual-2 | fix_needed | high | answer_units |  | Answer units omit the first item in part (a) (19), part (b) (23), and merged part (c) (29). | Add missing units for 19, 23, and 29 and align unit ids with the correct source subparts. |
+| 2024-individual-3 | fix_needed | critical | prompt |  | The source problem includes a family-tree diagram above the statements, but the dataset prompt omits the actual tree structure and keeps only a name plus legend words. | Include the rendered family tree or transcribe it into explicit relationship edges and node labels. |
+| 2024-individual-3 | fix_needed | high | canonical |  | The part a official answer is reduced to a loose list of names, losing the solved family-tree structure from the PDF. | Represent the solved tree with an image or explicit relationship structure in canonical and answer_units. |
+| 2024-individual-3 | fix_needed | critical | answer_units |  | Part b answer_units miss gap 1 and part a has only a coarse text list rather than a scoreable tree answer; part c includes footer text. | Add the missing part b gap 1 answer, split units by target, remove footers, and encode the tree answer structurally. |
+| 2024-individual-3 | fix_needed | high | subparts | iol-2024-individual-p3-partial-through-a-target-b | The partial variant shows solved part a as the same unstructured name list, so prior solved context is not faithful. | Use the repaired tree representation in shown solved subparts. |
+| 2024-individual-4 | fix_needed | high | answer_units |  | Answer units omit the first required answer in each subpart: (a) item 1, (b) item 21, and (c) item 24. | Add units for 1. cum aa = C flip-flops, 21. daw sob piis = finger, and 24. brook/stream = nax piis, preserving source characters. |
+| 2024-individual-4 | fix_needed | medium | prompt |  | The source problem refers to images on the next page, but the prompt only contains captions for capybara, tapir, and anteater without the images. | Include the rendered animal images or mark them as omitted only if the dataset intentionally relies on the textual note instead. |
+| 2024-individual-5 | fix_needed | high | answer_units |  | The first answer in every subpart is omitted from answer_units (15 in parts a/b and 19 in part c). Part (c) units also capture only the English prompts for 20-22, not the required Yanyuwa translations. | Regenerate units for items 15-22, including the actual women/men Yanyuwa forms for each part (c) answer. |
+| 2024-individual-5 | fix_needed | low | metadata |  | Records are marked requires_images even though the cited Yanyuwa problem itself is textual; the image count appears to come from unrelated material on the page range. | Set requires_images based on the extracted source problem span rather than whole cited pages. |
+| 2024-team-1 | fix_needed | critical | answer_units | iol-2024-team-p1 | Only three garbled units were extracted for a multi-assignment team problem, and they do not correspond to the official assignment answers. | Rebuild answer_units by assignment from the solution material, with complete values for each final answer/tree/distance/procedure. |
+| 2024-team-1 | fix_needed | high | canonical | iol-2024-team-p1 | The cited solution is a slide deck with graphical trees; the dataset canonical is a flattened slide text dump rather than a clean, source-faithful answer text. | Use rendered tree images or a structured representation of each tree and extract a clean answer key from the slides. |
+| 2025-individual-1 | fix_needed | high | canonical |  | Several solution equalities involving fractions are flattened incorrectly in canonical text, e.g. part (b)(5) appears as '2 x 43 + 21 = 2' instead of using 3/4 and 1/2, and (b)(6) appears with '1100 x 12' instead of '1100 x 1/2'. | Manually transcribe the fraction expressions from the rendered solution PDF using explicit slash notation such as 3/4 and 1/2. |
+| 2025-individual-1 | fix_needed | high | answer_units |  | Part (a) answer_units omit [X] = 60 = ɖukcu; part (b) is a single unit for ten equalities and part (c) a single unit for two outputs. | Add the missing X unit and split b into ten units and c into separate units for 75 and 570. |
+| 2025-individual-2 | fix_needed | high | subparts |  | Source part (d) is merged into records labeled only target c; the target-c prompt says solve (c) only but includes both c and d requests, and no separate d record exists. | Split part (d) into its own target_subpart/record or relabel the merged target as c+d. |
+| 2025-individual-2 | fix_needed | medium | answer_units |  | Part (a) is a single unit for 20 correspondences, which is too coarse; merged c/d units also omit source items 21 and 26. | Create one unit per correspondence for part (a), and add separate units for all c and d translations including 21 and 26. |
+| 2025-individual-3 | fix_needed | critical | answer_units |  | Answer units for the combined b/c target omit official answers 21 and 25, and c answers are mislabeled under b. | Add units for 21 and 25, preserve source labels b and c, and keep one required unit per numbered target. |
+| 2025-individual-3 | fix_needed | high | subparts |  | The source has subparts a, b, and c, but the dataset has only sub-a and sub-b records; the sub-b and partial target-b records include both b and c tasks. | Generate separate sub-b and sub-c records, plus partial variants whose target labels match the source. |
+| 2025-individual-3 | fix_needed | low | prompt |  | The full and partial prompts include an irrelevant header from page 4 after the Problem 3 text. | Trim extraction at the end of Problem 3 rather than carrying over the next page header. |
+| 2025-individual-4 | fix_needed | critical | subparts |  | Subpart (b) records and the partial variant incorrectly merge the two-column (b)/(c) target into one target question, so a record labeled target (b) also asks for (c). | Separate subpart (b) and subpart (c) into distinct records/targets, or label the current record as a combined b+c task if that is intended. |
+| 2025-individual-4 | fix_needed | high | answer_units |  | The subpart (b) answer units mix (b) and (c) answers under label b, omit item 40 from (b), and omit item 47 from (c). | Create separate b units for 40-46 and c units for 47-53, including the missing repena = tree/fire and white man = bali answers. |
+| 2025-individual-4 | fix_needed | medium | answer_units | iol-2025-individual-p4-sub-a | Subpart (a)'s 25 matching answers are bundled into one unit, too coarse for RL scoring. | Split subpart (a) into one answer unit for each item 15-39. |
+| 2025-individual-4 | fix_needed | low | provenance |  | The problem page range includes page 5, which starts Problem 5, and prompts include page-footer text. | Narrow the page/text span to Problem 4 and remove footer noise. |
+| 2025-individual-5 | fix_needed | critical | prompt |  | The Kaqchikel task depends on colored stick-figure pictures, but the dataset prompt contains only sentence text/gaps and no image descriptions or attachments. Parts (a), (b), and (c) are not answerable from the text record alone. | Attach/render the relevant PDF images or manually encode each picture scene in text. |
+| 2025-individual-5 | fix_needed | high | subparts | iol-2025-individual-p5-sub-c | Part (d) is merged into the sub-c target_question and canonical answer, so a record labeled target_subparts=["c"] actually asks for and answers both (c) and (d). | Split part (d) into its own subproblem/partial target or exclude it from sub-c. |
+| 2025-individual-5 | fix_needed | high | answer_units |  | Answer_units are incomplete/coarse: part (a) omits item 1, part (b) groups two picture-description answers as one unit, and part (c)/(d) is one large merged unit. | Create per-item units for A-I, 9-10, the four possible sentences in (c), and the three cortex predictions in (d). |
+| 2025-team-1 | fix_needed | critical | subparts |  | Official part (f) is merged into the part (e) canonical/unit, and there is no sub-f record or partial variant; the sequence jumps from e to g. | Split parts (e) and (f), add sub-f and appropriate partial records, and align visible solved parts accordingly. |
+| 2025-team-1 | fix_needed | high | answer_units |  | Large matching parts (a)-(d) and (h) are single coarse units, part (g) omits item 1, and the part (e) unit includes part (f) plus page footer text. | Create granular units for every numbered/lettered answer, add g item 1, remove footers, and separate e/f units. |

benchmark/IOL/ioling_hf/reports/microsteps/qwen3_4b_microsteps_balanced_v2_e5_train_greedy.json

@@ -0,0 +1,2163 @@
+{
+  "summary": {
+    "schema_version": "ioling_microstep_eval_v1",
+    "model": "checkpoints/qwen3-4b-ioling-microsteps-balanced-v2-e5-merged",
+    "adapter": null,
+    "tasks": 100,
+    "samples_per_task": 1,
+    "samples": 100,
+    "correct_samples": 28,
+    "correct_sample_rate": 0.28,
+    "strict_correct_samples": 28,
+    "strict_correct_sample_rate": 0.28,
+    "pass_at_n_tasks": 28,
+    "pass_at_n_task_rate": 0.28,
+    "pass_at_n_by_type": {
+      "allomorph_rule": {
+        "passed": 3,
+        "tasks": 8
+      },
+      "arithmetic_rule": {
+        "passed": 2,
+        "tasks": 4
+      },
+      "class_mapping": {
+        "passed": 1,
+        "tasks": 6
+      },
+      "lexical_mapping": {
+        "passed": 4,
+        "tasks": 12
+      },
+      "morpheme_mapping": {
+        "passed": 9,
+        "tasks": 38
+      },
+      "morpheme_order": {
+        "passed": 1,
+        "tasks": 4
+      },
+      "prosody_rule": {
+        "passed": 2,
+        "tasks": 2
+      },
+      "role_rule": {
+        "passed": 0,
+        "tasks": 1
+      },
+      "transformation_rule": {
+        "passed": 3,
+        "tasks": 13
+      },
+      "word_order": {
+        "passed": 3,
+        "tasks": 12
+      }
+    },
+    "temperature": 0.0,
+    "max_tokens": 256,
+    "seed": 42
+  },
+  "results": [
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-definite-order",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-definite-order",
+      "step_type": "word_order",
+      "question": "What is the order of subject, definite object, and verb in a definite transitive clause?",
+      "accepted_answers": [
+        "subject-up, object-(q), verb"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-definite-order\",\"answer\":\"subject, definite object, verb\"}",
+          "parsed_answer": "subject, definite object, verb",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-indefinite-object",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-indefinite-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which noun suffix marks an indefinite object?",
+      "accepted_answers": [
+        "-(r)mik"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-indefinite-object\",\"answer\":\"-sijiit\"}",
+          "parsed_answer": "-sijiit",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-indefinite-verb",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-indefinite-verb",
+      "step_type": "morpheme_mapping",
+      "question": "Which verb element accompanies an indefinite object?",
+      "accepted_answers": [
+        "-si"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-indefinite-verb\",\"answer\":\"-q\"}",
+          "parsed_answer": "-q",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-linker-vowel",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-linker-vowel",
+      "step_type": "allomorph_rule",
+      "question": "Which linker precedes a person ending after a vowel-final verb?",
+      "accepted_answers": [
+        "-j"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-linker-vowel\",\"answer\":\"-u-\"}",
+          "parsed_answer": "-u-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-linker-consonant",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-linker-consonant",
+      "step_type": "allomorph_rule",
+      "question": "Which linker precedes a person ending after a consonant-final verb?",
+      "accepted_answers": [
+        "-t"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-linker-consonant\",\"answer\":\"-q\"}",
+          "parsed_answer": "-q",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-third-to-second",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-third-to-second",
+      "step_type": "morpheme_mapping",
+      "question": "Which ending marks a third-person subject acting on a second-person object?",
+      "accepted_answers": [
+        "-aatit"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-third-to-second\",\"answer\":\"-q\"}",
+          "parsed_answer": "-q",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-your-noun",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-your-noun",
+      "step_type": "transformation_rule",
+      "question": "How is 'your' marked on a subject or object noun ending in -(q)?",
+      "accepted_answers": [
+        "replace -(q) with -it"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-your-noun\",\"answer\":\"-q\"}",
+          "parsed_answer": "-q",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-order",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-order",
+      "step_type": "morpheme_order",
+      "question": "What is the order of components in a Menominee verb?",
+      "accepted_answers": [
+        "person prefix, lexical element, verb element, person ending"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-order\",\"answer\":\"subject, object, verb\"}",
+          "parsed_answer": "subject, object, verb",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-first-prefix",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-first-prefix",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks first-person singular?",
+      "accepted_answers": [
+        "ne-"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-first-prefix\",\"answer\":\"ne-\"}",
+          "parsed_answer": "ne-",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-inclusive-prefix",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-inclusive-prefix",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks inclusive we (we1+2)?",
+      "accepted_answers": [
+        "ke-"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-inclusive-prefix\",\"answer\":\"ke-\"}",
+          "parsed_answer": "ke-",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-by-mouth",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-by-mouth",
+      "step_type": "morpheme_mapping",
+      "question": "Which transitive element means 'by mouth'?",
+      "accepted_answers": [
+        "-aht"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-by-mouth\",\"answer\":\"-tæq\"}",
+          "parsed_answer": "-tæq",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-by-hand",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-by-hand",
+      "step_type": "morpheme_mapping",
+      "question": "Which transitive element means 'by hand'?",
+      "accepted_answers": [
+        "-en"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-by-hand\",\"answer\":\"kēskenam\"}",
+          "parsed_answer": "kēskenam",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-transitive-first",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-transitive-first",
+      "step_type": "morpheme_mapping",
+      "question": "Which transitive ending marks first-person singular?",
+      "accepted_answers": [
+        "-an"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-transitive-first\",\"answer\":\"-næm\"}",
+          "parsed_answer": "-næm",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-length-rule",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-length-rule",
+      "step_type": "transformation_rule",
+      "question": "What happens when the first two vowels of an assembled word are both short?",
+      "accepted_answers": [
+        "the second vowel becomes long"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-length-rule\",\"answer\":\"The first two vowels of the assembled word are both short, so the first vowel is not long.\"}",
+          "parsed_answer": "The first two vowels of the assembled word are both short, so the first vowel is not long.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 34
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-compound-order",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-compound-order",
+      "step_type": "word_order",
+      "question": "Which part of a Nahuatl compound modifies which other part?",
+      "accepted_answers": [
+        "the left-hand part modifies the right-hand part"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-compound-order\",\"answer\":\"The first element modifies the second element.\"}",
+          "parsed_answer": "The first element modifies the second element.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 23
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-noun-after-l",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-noun-after-l",
+      "step_type": "allomorph_rule",
+      "question": "Which ordinary noun ending occurs after l?",
+      "accepted_answers": [
+        "-li"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-noun-after-l\",\"answer\":\"-tl\"}",
+          "parsed_answer": "-tl",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-noun-elsewhere",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-noun-elsewhere",
+      "step_type": "allomorph_rule",
+      "question": "Which ordinary noun ending occurs when the stem does not end in l?",
+      "accepted_answers": [
+        "-tl"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-noun-elsewhere\",\"answer\":\"-tl\"}",
+          "parsed_answer": "-tl",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-many-after-l",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-many-after-l",
+      "step_type": "allomorph_rule",
+      "question": "Which 'place of many' suffix occurs after l?",
+      "accepted_answers": [
+        "-lah"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-many-after-l\",\"answer\":\"-huah\"}",
+          "parsed_answer": "-huah",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-diminutive",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-diminutive",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix is the diminutive?",
+      "accepted_answers": [
+        "-capil"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-diminutive\",\"answer\":\"-capil\"}",
+          "parsed_answer": "-capil",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-order",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-order",
+      "step_type": "word_order",
+      "question": "What is the basic constituent order in the Dyirbal data?",
+      "accepted_answers": [
+        "object, subject, verb"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-order\",\"answer\":\"Subject-Object-Verb\"}",
+          "parsed_answer": "Subject-Object-Verb",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-adjective-order",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-adjective-order",
+      "step_type": "word_order",
+      "question": "Where does an adjective occur relative to its noun?",
+      "accepted_answers": [
+        "after the noun"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-adjective-order\",\"answer\":\"before the noun\"}",
+          "parsed_answer": "before the noun",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-causative",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-causative",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix forms a causative verb from an adjective?",
+      "accepted_answers": [
+        "-man"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-causative\",\"answer\":\"-man\"}",
+          "parsed_answer": "-man",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-object-dangerous",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-object-dangerous",
+      "step_type": "class_mapping",
+      "question": "Which object article is used for women and dangerous animals or objects?",
+      "accepted_answers": [
+        "balan"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-object-dangerous\",\"answer\":\"bala\"}",
+          "parsed_answer": "bala",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-object-men",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-object-men",
+      "step_type": "class_mapping",
+      "question": "Which object article is used for men and other animals?",
+      "accepted_answers": [
+        "bayi"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-object-men\",\"answer\":\"bala\"}",
+          "parsed_answer": "bala",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-object-other",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-object-other",
+      "step_type": "class_mapping",
+      "question": "Which object article is used for the remaining noun class?",
+      "accepted_answers": [
+        "bala"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-object-other\",\"answer\":\"baNgul\"}",
+          "parsed_answer": "baNgul",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-changed-position",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-changed-position",
+      "step_type": "word_order",
+      "question": "Which word in a Rotuman two-word phrase undergoes the phrase-form change?",
+      "accepted_answers": [
+        "the first word"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-changed-position\",\"answer\":\"the first word\"}",
+          "parsed_answer": "the first word",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-vcv",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-vcv",
+      "step_type": "transformation_rule",
+      "question": "What phrase form results from a citation form ending in -VCV?",
+      "accepted_answers": [
+        "-VC"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-vcv\",\"answer\":\"-VCV\"}",
+          "parsed_answer": "-VCV",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-vcu",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-vcu",
+      "step_type": "transformation_rule",
+      "question": "What phrase form results from a citation form ending in -VCu?",
+      "accepted_answers": [
+        "-VC"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-vcu\",\"answer\":\"-VCu is a citation form ending in -VCu\"}",
+          "parsed_answer": "-VCu is a citation form ending in -VCu",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 26
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-vca",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-vca",
+      "step_type": "transformation_rule",
+      "question": "What happens to final -VCa in phrase form?",
+      "accepted_answers": [
+        "it becomes -VaC"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-vca\",\"answer\":\"It becomes -VCa in phrase form.\"}",
+          "parsed_answer": "It becomes -VCa in phrase form.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-noun-adjective",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-noun-adjective",
+      "step_type": "word_order",
+      "question": "What is the order of a noun and its adjective?",
+      "accepted_answers": [
+        "noun, adjective"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-noun-adjective\",\"answer\":\"noun before adjective\"}",
+          "parsed_answer": "noun before adjective",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-his",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-his",
+      "step_type": "morpheme_mapping",
+      "question": "Which Jaqaru possessor suffix means 'his'?",
+      "accepted_answers": [
+        "-p”a"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-his\",\"answer\":\"-shun-\"}",
+          "parsed_answer": "-shun-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-our",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-our",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor suffix means inclusive 'our1+2'?",
+      "accepted_answers": [
+        "-sa"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-our\",\"answer\":\"-npsa\"}",
+          "parsed_answer": "-npsa",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-my",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-my",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor suffix means 'my'?",
+      "accepted_answers": [
+        "-nha"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-my\",\"answer\":\"-nha\"}",
+          "parsed_answer": "-nha",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-your",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-your",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor suffix means 'your'?",
+      "accepted_answers": [
+        "-ma"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-your\",\"answer\":\"-ma\"}",
+          "parsed_answer": "-ma",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-from",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-from",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix means 'from'?",
+      "accepted_answers": [
+        "-t”a"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-from\",\"answer\":\"-t”a\"}",
+          "parsed_answer": "-t”a",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-interrogative",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-interrogative",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix is interrogative?",
+      "accepted_answers": [
+        "-txi"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-interrogative\",\"answer\":\"-t”xi\"}",
+          "parsed_answer": "-t”xi",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-case-vowel-rule",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-case-vowel-rule",
+      "step_type": "transformation_rule",
+      "question": "When an oral case suffix follows a root-final vowel, is that root vowel consumed or preserved?",
+      "accepted_answers": [
+        "preserved"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-case-vowel-rule\",\"answer\":\"preserved\"}",
+          "parsed_answer": "preserved",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-possessor-order",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-possessor-order",
+      "step_type": "word_order",
+      "question": "Where does a possessor occur relative to the possessed noun?",
+      "accepted_answers": [
+        "before the possessed noun"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-possessor-order\",\"answer\":\"before\"}",
+          "parsed_answer": "before",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-first-possessor",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-first-possessor",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor prefix marks first-person singular?",
+      "accepted_answers": [
+        "n-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-first-possessor\",\"answer\":\"ma-\"}",
+          "parsed_answer": "ma-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-second-possessor",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-second-possessor",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor prefix marks second-person singular?",
+      "accepted_answers": [
+        "zero"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-second-possessor\",\"answer\":\"-t-\"}",
+          "parsed_answer": "-t-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-third-possessor",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-third-possessor",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor prefix marks third-person singular?",
+      "accepted_answers": [
+        "h-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-third-possessor\",\"answer\":\"-\"}",
+          "parsed_answer": "-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-body-part",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-body-part",
+      "step_type": "class_mapping",
+      "question": "Except after ri-, which prefix appears on a possessed body-part noun?",
+      "accepted_answers": [
+        "a-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-body-part\",\"answer\":\"eku-\"}",
+          "parsed_answer": "eku-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-other-noun",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-other-noun",
+      "step_type": "class_mapping",
+      "question": "Except after ri-, which prefix appears on another possessed noun?",
+      "accepted_answers": [
+        "e-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-other-noun\",\"answer\":\"eku-\"}",
+          "parsed_answer": "eku-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-basic-order",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-basic-order",
+      "step_type": "word_order",
+      "question": "What is the basic order of subject, verb, optional object, and optional place?",
+      "accepted_answers": [
+        "subject, verb, object, place"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-basic-order\",\"answer\":\"subject, verb, object, place\"}",
+          "parsed_answer": "subject, verb, object, place",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-first-subject",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-first-subject",
+      "step_type": "morpheme_mapping",
+      "question": "Which verb prefix expresses a first-person-singular subject?",
+      "accepted_answers": [
+        "ŋg-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-first-subject\",\"answer\":\"ŋ-\"}",
+          "parsed_answer": "ŋ-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-possessed-order",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-possessed-order",
+      "step_type": "word_order",
+      "question": "Where does a possessed noun occur relative to its possessive pronoun?",
+      "accepted_answers": [
+        "before the possessive pronoun"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-possessed-order\",\"answer\":\"after the possessive pronoun\"}",
+          "parsed_answer": "after the possessive pronoun",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 23
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-place-concord",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-place-concord",
+      "step_type": "class_mapping",
+      "question": "Which concord prefix marks place expressions?",
+      "accepted_answers": [
+        "mù-"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-place-concord\",\"answer\":\"mù-\"}",
+          "parsed_answer": "mù-",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-statement-tone",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-statement-tone",
+      "step_type": "prosody_rule",
+      "question": "What is the sentence-final tone of a statement?",
+      "accepted_answers": [
+        "low"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-statement-tone\",\"answer\":\"low\"}",
+          "parsed_answer": "low",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-question-tone",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-question-tone",
+      "step_type": "prosody_rule",
+      "question": "What is the sentence-final tone of a question?",
+      "accepted_answers": [
+        "high"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-question-tone\",\"answer\":\"high\"}",
+          "parsed_answer": "high",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-order",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-order",
+      "step_type": "word_order",
+      "question": "What is the order of subject, optional object, predicate, agreement complex, and final ne?",
+      "accepted_answers": [
+        "subject, object, predicate, agreement complex, ne"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-order\",\"answer\":\"subject, optional object, predicate, agreement complex, final ne\"}",
+          "parsed_answer": "subject, optional object, predicate, agreement complex, final ne",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 25
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-singular-transitive",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-singular-transitive",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker follows a first- or second-person-singular transitive subject?",
+      "accepted_answers": [
+        "-bə"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-singular-transitive\",\"answer\":\"ŋa\"}",
+          "parsed_answer": "ŋa",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-other-transitive",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-other-transitive",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker follows a third-person or plural transitive subject?",
+      "accepted_answers": [
+        "kəmə"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-other-transitive\",\"answer\":\"ŋa\"}",
+          "parsed_answer": "ŋa",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-hierarchy",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-hierarchy",
+      "step_type": "role_rule",
+      "question": "What is the person hierarchy from highest to lowest?",
+      "accepted_answers": [
+        "first > second > third"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-hierarchy\",\"answer\":\"subject-object-verb\"}",
+          "parsed_answer": "subject-object-verb",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-present-higher-subject",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-present-higher-subject",
+      "step_type": "morpheme_mapping",
+      "question": "Which agreement prefix marks present tense when the subject outranks the object?",
+      "accepted_answers": [
+        "k-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-present-higher-subject\",\"answer\":\"n-\"}",
+          "parsed_answer": "n-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-present-higher-object",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-present-higher-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which agreement prefix marks present tense when the object outranks the subject?",
+      "accepted_answers": [
+        "r-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-present-higher-object\",\"answer\":\"n-\"}",
+          "parsed_answer": "n-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-sleep",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-sleep",
+      "step_type": "lexical_mapping",
+      "question": "Which predicate means 'sleep'?",
+      "accepted_answers": [
+        "ʒip"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-sleep\",\"answer\":\"ʒip\"}",
+          "parsed_answer": "ʒip",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-first-p",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-first-p",
+      "step_type": "transformation_rule",
+      "question": "For first-person singular, what does the first voiceless p become?",
+      "accepted_answers": [
+        "mb"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-first-p\",\"answer\":\"b\"}",
+          "parsed_answer": "b",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-first-t",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-first-t",
+      "step_type": "transformation_rule",
+      "question": "For first-person singular, what does the first voiceless t become?",
+      "accepted_answers": [
+        "nd"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-first-t\",\"answer\":\"n\"}",
+          "parsed_answer": "n",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-first-k",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-first-k",
+      "step_type": "transformation_rule",
+      "question": "For first-person singular, what does the first voiceless k become?",
+      "accepted_answers": [
+        "ng"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-first-k\",\"answer\":\"g\"}",
+          "parsed_answer": "g",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-second-a",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-second-a",
+      "step_type": "transformation_rule",
+      "question": "For second-person singular in a native word, what does the first eligible a become?",
+      "accepted_answers": [
+        "e"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-second-a\",\"answer\":\"e\"}",
+          "parsed_answer": "e",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-second-o",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-second-o",
+      "step_type": "transformation_rule",
+      "question": "For second-person singular, what does the first eligible o become?",
+      "accepted_answers": [
+        "e"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-second-o\",\"answer\":\"e\"}",
+          "parsed_answer": "e",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-portuguese-accent",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-portuguese-accent",
+      "step_type": "transformation_rule",
+      "question": "In a Portuguese loan, what does á become for second-person singular?",
+      "accepted_answers": [
+        "eá"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-portuguese-accent\",\"answer\":\"e\"}",
+          "parsed_answer": "e",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-copular-order",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-copular-order",
+      "step_type": "word_order",
+      "question": "What is the order of the two terms in the Nooni copular construction?",
+      "accepted_answers": [
+        "A nú B"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-copular-order\",\"answer\":\"subject-then-day\"}",
+          "parsed_answer": "subject-then-day",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-clause-order",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-clause-order",
+      "step_type": "word_order",
+      "question": "What is the ordinary order of subject, tense, verb, object, place, and day?",
+      "accepted_answers": [
+        "subject, tense, verb, object, place, day"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-clause-order\",\"answer\":\"subject, tense, verb, object, place, day\"}",
+          "parsed_answer": "subject, tense, verb, object, place, day",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 25
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-see",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-see",
+      "step_type": "morpheme_order",
+      "question": "What is the discontinuous construction meaning 'see'?",
+      "accepted_answers": [
+        "yɛn (object) lé"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-see\",\"answer\":\"yɛn\"}",
+          "parsed_answer": "yɛn",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-past-today",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-past-today",
+      "step_type": "morpheme_mapping",
+      "question": "Which tense marker means past today?",
+      "accepted_answers": [
+        "bɛ̂ː"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-past-today\",\"answer\":\"cǐ\"}",
+          "parsed_answer": "cǐ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-near-future",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-near-future",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker means future fewer than three days ahead?",
+      "accepted_answers": [
+        "nú bóy"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-near-future\",\"answer\":\"ntfǔː\"}",
+          "parsed_answer": "ntfǔː",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-far-future",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-far-future",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker means future more than three days ahead?",
+      "accepted_answers": [
+        "nú yúː"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-far-future\",\"answer\":\"ntfǔː\"}",
+          "parsed_answer": "ntfǔː",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-first-person",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-first-person",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks first-person singular on the tense marker or verb?",
+      "accepted_answers": [
+        "n-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-first-person\",\"answer\":\"me-\"}",
+          "parsed_answer": "me-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-le-after-consonant",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-le-after-consonant",
+      "step_type": "allomorph_rule",
+      "question": "What does lé become after a consonant?",
+      "accepted_answers": [
+        "é"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-le-after-consonant\",\"answer\":\"é\"}",
+          "parsed_answer": "é",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-one",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-one",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 1?",
+      "accepted_answers": [
+        "ena"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-one\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-six",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-six",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 6?",
+      "accepted_answers": [
+        "benomi"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-six\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-nine",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-nine",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 9?",
+      "accepted_answers": [
+        "ije"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-nine\",\"answer\":\"ije mepija\"}",
+          "parsed_answer": "ije mepija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-ten",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-ten",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 10?",
+      "accepted_answers": [
+        "rati"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-ten\",\"answer\":\"wido\"}",
+          "parsed_answer": "wido",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-sixty",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-sixty",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 60?",
+      "accepted_answers": [
+        "moeto"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-sixty\",\"answer\":\"wido\"}",
+          "parsed_answer": "wido",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-addition",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-addition",
+      "step_type": "arithmetic_rule",
+      "question": "What operation is expressed when a unit with -ma precedes a multiple of ten?",
+      "accepted_answers": [
+        "addition"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-addition\",\"answer\":\"addition\"}",
+          "parsed_answer": "addition",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-subtractive-nine",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-subtractive-nine",
+      "step_type": "arithmetic_rule",
+      "question": "Which unit may precede the next decade to express subtraction by one?",
+      "accepted_answers": [
+        "ije"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-subtractive-nine\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-daemita",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-daemita",
+      "step_type": "arithmetic_rule",
+      "question": "What value does X daemita moeto represent?",
+      "accepted_answers": [
+        "60 + X"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-daemita\",\"answer\":\"80\"}",
+          "parsed_answer": "80",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-role-order",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-role-order",
+      "step_type": "morpheme_order",
+      "question": "In what role order do person prefixes occur around the fixed verb stem?",
+      "accepted_answers": [
+        "direct object, indirect object, subject"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-role-order\",\"answer\":\"person prefixes occur in the order: subject, object, recipient\"}",
+          "parsed_answer": "person prefixes occur in the order: subject, object, recipient",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 25
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-first-singular",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-first-singular",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks first-person singular?",
+      "accepted_answers": [
+        "s-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-first-singular\",\"answer\":\"a-\"}",
+          "parsed_answer": "a-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-first-plural",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-first-plural",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks first-person plural?",
+      "accepted_answers": [
+        "šʼ-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-first-plural\",\"answer\":\"šʼ\"}",
+          "parsed_answer": "šʼ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-second-singular",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-second-singular",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks second-person singular?",
+      "accepted_answers": [
+        "w-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-second-singular\",\"answer\":\"s-\"}",
+          "parsed_answer": "s-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-second-plural",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-second-plural",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks second-person plural?",
+      "accepted_answers": [
+        "ŝʷ-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-second-plural\",\"answer\":\"šʼ\"}",
+          "parsed_answer": "šʼ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-a-vowel",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-a-vowel",
+      "step_type": "allomorph_rule",
+      "question": "When is the verb vowel a rather than ə?",
+      "accepted_answers": [
+        "when the direct object is plural or a second-person-plural participant is present"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-a-vowel\",\"answer\":\"when the object is plural and the subject is first-person singular\"}",
+          "parsed_answer": "when the object is plural and the subject is first-person singular",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 26
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-one",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-one",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 1?",
+      "accepted_answers": [
+        "niŋkin"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-one\",\"answer\":\"kɛ\"}",
+          "parsed_answer": "kɛ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-five",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-five",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 5?",
+      "accepted_answers": [
+        "kaŋkuro"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-five\",\"answer\":\"kampwoo\"}",
+          "parsed_answer": "kampwoo",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-ten",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-ten",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 10?",
+      "accepted_answers": [
+        "kɛ"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-ten\",\"answer\":\"kɛ\"}",
+          "parsed_answer": "kɛ",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-twenty",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-twenty",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 20?",
+      "accepted_answers": [
+        "beɲjaaga"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-twenty\",\"answer\":\"beɲjaaga\"}",
+          "parsed_answer": "beɲjaaga",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-eighty",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-eighty",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 80?",
+      "accepted_answers": [
+        "ŋkuu"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-eighty\",\"answer\":\"kampwoo\"}",
+          "parsed_answer": "kampwoo",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-four-hundred",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-four-hundred",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 400?",
+      "accepted_answers": [
+        "kampwoo"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-four-hundred\",\"answer\":\"kampwoo\"}",
+          "parsed_answer": "kampwoo",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-addition",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-addition",
+      "step_type": "arithmetic_rule",
+      "question": "What arithmetic operation does na express?",
+      "accepted_answers": [
+        "addition"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-addition\",\"answer\":\"addition\"}",
+          "parsed_answer": "addition",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-400-multiplier",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-400-multiplier",
+      "step_type": "allomorph_rule",
+      "question": "Which form of 400 occurs before a multiplier?",
+      "accepted_answers": [
+        "kampwɔhii"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-400-multiplier\",\"answer\":\"kampwɔhii\"}",
+          "parsed_answer": "kampwɔhii",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 24
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-verb-order",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-verb-order",
+      "step_type": "morpheme_order",
+      "question": "Which comes first in the verb: subject prefix or tense prefix?",
+      "accepted_answers": [
+        "subject prefix"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-verb-order\",\"answer\":\"subject prefix\"}",
+          "parsed_answer": "subject prefix",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-first-subject",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-first-subject",
+      "step_type": "morpheme_mapping",
+      "question": "Which subject prefix marks first-person singular?",
+      "accepted_answers": [
+        "t-"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-first-subject\",\"answer\":\"t-\"}",
+          "parsed_answer": "t-",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-third-nonsingular",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-third-nonsingular",
+      "step_type": "morpheme_mapping",
+      "question": "Which subject prefix marks third-person dual or plural?",
+      "accepted_answers": [
+        "ne-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-third-nonsingular\",\"answer\":\"-k-\"}",
+          "parsed_answer": "-k-",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-present",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-present",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks present tense?",
+      "accepted_answers": [
+        "ku-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-present\",\"answer\":\"-n\"}",
+          "parsed_answer": "-n",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-future",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-future",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks future tense?",
+      "accepted_answers": [
+        "je-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-future\",\"answer\":\"-tək\"}",
+          "parsed_answer": "-tək",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-third-singular-object",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-third-singular-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which ending marks a third-person-singular object?",
+      "accepted_answers": [
+        "-n"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-third-singular-object\",\"answer\":\"-tək\"}",
+          "parsed_answer": "-tək",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-third-plural-object",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-third-plural-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which ending marks a third-person-plural object?",
+      "accepted_answers": [
+        "-new"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-third-plural-object\",\"answer\":\"-new\"}",
+          "parsed_answer": "-new",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-tj",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-tj",
+      "step_type": "transformation_rule",
+      "question": "What does the sequence t+j become?",
+      "accepted_answers": [
+        "cc"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-tj\",\"answer\":\"t+j becomes t+j\"}",
+          "parsed_answer": "t+j becomes t+j",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    }
+  ]
+}

benchmark/IOL/ioling_hf/reports/microsteps/qwen3_4b_microsteps_mixed_v1_e2_train_greedy.json

@@ -0,0 +1,2163 @@
+{
+  "summary": {
+    "schema_version": "ioling_microstep_eval_v1",
+    "model": "checkpoints/qwen3-4b-ioling-microsteps-mixed-v1-e2-merged",
+    "adapter": null,
+    "tasks": 100,
+    "samples_per_task": 1,
+    "samples": 100,
+    "correct_samples": 8,
+    "correct_sample_rate": 0.08,
+    "strict_correct_samples": 8,
+    "strict_correct_sample_rate": 0.08,
+    "pass_at_n_tasks": 8,
+    "pass_at_n_task_rate": 0.08,
+    "pass_at_n_by_type": {
+      "allomorph_rule": {
+        "passed": 1,
+        "tasks": 8
+      },
+      "arithmetic_rule": {
+        "passed": 2,
+        "tasks": 4
+      },
+      "class_mapping": {
+        "passed": 0,
+        "tasks": 6
+      },
+      "lexical_mapping": {
+        "passed": 0,
+        "tasks": 12
+      },
+      "morpheme_mapping": {
+        "passed": 2,
+        "tasks": 38
+      },
+      "morpheme_order": {
+        "passed": 1,
+        "tasks": 4
+      },
+      "prosody_rule": {
+        "passed": 1,
+        "tasks": 2
+      },
+      "role_rule": {
+        "passed": 0,
+        "tasks": 1
+      },
+      "transformation_rule": {
+        "passed": 1,
+        "tasks": 13
+      },
+      "word_order": {
+        "passed": 0,
+        "tasks": 12
+      }
+    },
+    "temperature": 0.0,
+    "max_tokens": 256,
+    "seed": 42
+  },
+  "results": [
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-definite-order",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-definite-order",
+      "step_type": "word_order",
+      "question": "What is the order of subject, definite object, and verb in a definite transitive clause?",
+      "accepted_answers": [
+        "subject-up, object-(q), verb"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-definite-order\",\"answer\":\"subject, definite object, verb\"}",
+          "parsed_answer": "subject, definite object, verb",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-indefinite-object",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-indefinite-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which noun suffix marks an indefinite object?",
+      "accepted_answers": [
+        "-(r)mik"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-indefinite-object\",\"answer\":\"-it\"}",
+          "parsed_answer": "-it",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-indefinite-verb",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-indefinite-verb",
+      "step_type": "morpheme_mapping",
+      "question": "Which verb element accompanies an indefinite object?",
+      "accepted_answers": [
+        "-si"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-indefinite-verb\",\"answer\":\"iputujait\"}",
+          "parsed_answer": "iputujait",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-linker-vowel",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-linker-vowel",
+      "step_type": "allomorph_rule",
+      "question": "Which linker precedes a person ending after a vowel-final verb?",
+      "accepted_answers": [
+        "-j"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-linker-vowel\",\"answer\":\"iluaqhaiji\"}",
+          "parsed_answer": "iluaqhaiji",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-linker-consonant",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-linker-consonant",
+      "step_type": "allomorph_rule",
+      "question": "Which linker precedes a person ending after a consonant-final verb?",
+      "accepted_answers": [
+        "-t"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-linker-consonant\",\"answer\":\"iputuj\"}",
+          "parsed_answer": "iputuj",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-third-to-second",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-third-to-second",
+      "step_type": "morpheme_mapping",
+      "question": "Which ending marks a third-person subject acting on a second-person object?",
+      "accepted_answers": [
+        "-aatit"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-third-to-second\",\"answer\":\"-uit\"}",
+          "parsed_answer": "-uit",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2008-individual-5-microstep-inuktitut-your-noun",
+      "source_problem_id": "2008-individual-5",
+      "step_id": "inuktitut-your-noun",
+      "step_type": "transformation_rule",
+      "question": "How is 'your' marked on a subject or object noun ending in -(q)?",
+      "accepted_answers": [
+        "replace -(q) with -it"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"inuktitut-your-noun\",\"answer\":\"The suffix -(q) is used to mark 'your' on a subject noun\"}",
+          "parsed_answer": "The suffix -(q) is used to mark 'your' on a subject noun",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 32
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-order",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-order",
+      "step_type": "morpheme_order",
+      "question": "What is the order of components in a Menominee verb?",
+      "accepted_answers": [
+        "person prefix, lexical element, verb element, person ending"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-order\",\"answer\":\"The order of components in a Menominee verb is subject + verb + object, with the subject marked by prefixes like 'we1+2' or 'I' and the verb root followed by directional or manner markers.\"}",
+          "parsed_answer": "The order of components in a Menominee verb is subject + verb + object, with the subject marked by prefixes like 'we1+2' or 'I' and the verb root followed by directional or manner markers.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 58
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-first-prefix",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-first-prefix",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks first-person singular?",
+      "accepted_answers": [
+        "ne-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-first-prefix\",\"answer\":\"nepī\"}",
+          "parsed_answer": "nepī",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-inclusive-prefix",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-inclusive-prefix",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks inclusive we (we1+2)?",
+      "accepted_answers": [
+        "ke-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-inclusive-prefix\",\"answer\":\"kewǣ\"}",
+          "parsed_answer": "kewǣ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-by-mouth",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-by-mouth",
+      "step_type": "morpheme_mapping",
+      "question": "Which transitive element means 'by mouth'?",
+      "accepted_answers": [
+        "-aht"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-by-mouth\",\"answer\":\"kekēskahtæq\"}",
+          "parsed_answer": "kekēskahtæq",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-by-hand",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-by-hand",
+      "step_type": "morpheme_mapping",
+      "question": "Which transitive element means 'by hand'?",
+      "accepted_answers": [
+        "-en"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-by-hand\",\"answer\":\"kēskenam\"}",
+          "parsed_answer": "kēskenam",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-transitive-first",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-transitive-first",
+      "step_type": "morpheme_mapping",
+      "question": "Which transitive ending marks first-person singular?",
+      "accepted_answers": [
+        "-an"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-transitive-first\",\"answer\":\"-næm\"}",
+          "parsed_answer": "-næm",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-1-microstep-menominee-length-rule",
+      "source_problem_id": "2011-individual-1",
+      "step_id": "menominee-length-rule",
+      "step_type": "transformation_rule",
+      "question": "What happens when the first two vowels of an assembled word are both short?",
+      "accepted_answers": [
+        "the second vowel becomes long"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"menominee-length-rule\",\"answer\":\"The first two vowels of an assembled word being both short may indicate a specific verb form or action related to hand or tool use, such as breaking or cutting.\"}",
+          "parsed_answer": "The first two vowels of an assembled word being both short may indicate a specific verb form or action related to hand or tool use, such as breaking or cutting.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 46
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-compound-order",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-compound-order",
+      "step_type": "word_order",
+      "question": "Which part of a Nahuatl compound modifies which other part?",
+      "accepted_answers": [
+        "the left-hand part modifies the right-hand part"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-compound-order\",\"answer\":\"The modifier in a Nahuatl compound typically precedes the modified element, such as 'chili' modifying 'water' to form 'chilatl'.\"}",
+          "parsed_answer": "The modifier in a Nahuatl compound typically precedes the modified element, such as 'chili' modifying 'water' to form 'chilatl'.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 48
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-noun-after-l",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-noun-after-l",
+      "step_type": "allomorph_rule",
+      "question": "Which ordinary noun ending occurs after l?",
+      "accepted_answers": [
+        "-li"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-noun-after-l\",\"answer\":\"-tl\"}",
+          "parsed_answer": "-tl",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-noun-elsewhere",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-noun-elsewhere",
+      "step_type": "allomorph_rule",
+      "question": "Which ordinary noun ending occurs when the stem does not end in l?",
+      "accepted_answers": [
+        "-tl"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-noun-elsewhere\",\"answer\":\"-tl\"}",
+          "parsed_answer": "-tl",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-many-after-l",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-many-after-l",
+      "step_type": "allomorph_rule",
+      "question": "Which 'place of many' suffix occurs after l?",
+      "accepted_answers": [
+        "-lah"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-many-after-l\",\"answer\":\"-huah\"}",
+          "parsed_answer": "-huah",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2011-individual-4-microstep-nahuatl-diminutive",
+      "source_problem_id": "2011-individual-4",
+      "step_id": "nahuatl-diminutive",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix is the diminutive?",
+      "accepted_answers": [
+        "-capil"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nahuatl-diminutive\",\"answer\":\"-tli\"}",
+          "parsed_answer": "-tli",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-order",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-order",
+      "step_type": "word_order",
+      "question": "What is the basic constituent order in the Dyirbal data?",
+      "accepted_answers": [
+        "object, subject, verb"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-order\",\"answer\":\"Subject-Verb-Object\"}",
+          "parsed_answer": "Subject-Verb-Object",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-adjective-order",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-adjective-order",
+      "step_type": "word_order",
+      "question": "Where does an adjective occur relative to its noun?",
+      "accepted_answers": [
+        "after the noun"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-adjective-order\",\"answer\":\"An adjective occurs before its noun in Dyirbal.\"}",
+          "parsed_answer": "An adjective occurs before its noun in Dyirbal.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 26
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-causative",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-causative",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix forms a causative verb from an adjective?",
+      "accepted_answers": [
+        "-man"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-causative\",\"answer\":\"-man\"}",
+          "parsed_answer": "-man",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-object-dangerous",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-object-dangerous",
+      "step_type": "class_mapping",
+      "question": "Which object article is used for women and dangerous animals or objects?",
+      "accepted_answers": [
+        "balan"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-object-dangerous\",\"answer\":\"The object article 'bala' is used for women and dangerous animals or objects.\"}",
+          "parsed_answer": "The object article 'bala' is used for women and dangerous animals or objects.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 32
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-object-men",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-object-men",
+      "step_type": "class_mapping",
+      "question": "Which object article is used for men and other animals?",
+      "accepted_answers": [
+        "bayi"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-object-men\",\"answer\":\"balan\"}",
+          "parsed_answer": "balan",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-1-microstep-dyirbal-object-other",
+      "source_problem_id": "2012-individual-1",
+      "step_id": "dyirbal-object-other",
+      "step_type": "class_mapping",
+      "question": "Which object article is used for the remaining noun class?",
+      "accepted_answers": [
+        "bala"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"dyirbal-object-other\",\"answer\":\"baNgul\"}",
+          "parsed_answer": "baNgul",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-changed-position",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-changed-position",
+      "step_type": "word_order",
+      "question": "Which word in a Rotuman two-word phrase undergoes the phrase-form change?",
+      "accepted_answers": [
+        "the first word"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-changed-position\",\"answer\":\"ele\"}",
+          "parsed_answer": "ele",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-vcv",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-vcv",
+      "step_type": "transformation_rule",
+      "question": "What phrase form results from a citation form ending in -VCV?",
+      "accepted_answers": [
+        "-VC"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-vcv\",\"answer\":\"a citation form ending in -VCV would result in a phrase form with a root containing a consonant-vowel-consonant structure\"}",
+          "parsed_answer": "a citation form ending in -VCV would result in a phrase form with a root containing a consonant-vowel-consonant structure",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 41
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-vcu",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-vcu",
+      "step_type": "transformation_rule",
+      "question": "What phrase form results from a citation form ending in -VCu?",
+      "accepted_answers": [
+        "-VC"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-vcu\",\"answer\":\"-VCu is a citation form ending that typically results in a phrase form involving a verb root with a specific suffix indicating a state or action, such as 'to be near' or 'to cut'\"}",
+          "parsed_answer": "-VCu is a citation form ending that typically results in a phrase form involving a verb root with a specific suffix indicating a state or action, such as 'to be near' or 'to cut'",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 55
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-vca",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-vca",
+      "step_type": "transformation_rule",
+      "question": "What happens to final -VCa in phrase form?",
+      "accepted_answers": [
+        "it becomes -VaC"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-vca\",\"answer\":\"Final -VCa in phrase form undergoes vowel harmony and assimilation to align with the surrounding phonological environment, particularly in verb stems and compound constructions.\"}",
+          "parsed_answer": "Final -VCa in phrase form undergoes vowel harmony and assimilation to align with the surrounding phonological environment, particularly in verb stems and compound constructions.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 44
+        }
+      ]
+    },
+    {
+      "record_id": "2012-individual-5-microstep-rotuman-noun-adjective",
+      "source_problem_id": "2012-individual-5",
+      "step_id": "rotuman-noun-adjective",
+      "step_type": "word_order",
+      "question": "What is the order of a noun and its adjective?",
+      "accepted_answers": [
+        "noun, adjective"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"rotuman-noun-adjective\",\"answer\":\"noun comes before adjective\"}",
+          "parsed_answer": "noun comes before adjective",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-his",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-his",
+      "step_type": "morpheme_mapping",
+      "question": "Which Jaqaru possessor suffix means 'his'?",
+      "accepted_answers": [
+        "-p”a"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-his\",\"answer\":\"chushp”a\"}",
+          "parsed_answer": "chushp”a",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 22
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-our",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-our",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor suffix means inclusive 'our1+2'?",
+      "accepted_answers": [
+        "-sa"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-our\",\"answer\":\"-p”sa\"}",
+          "parsed_answer": "-p”sa",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-my",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-my",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor suffix means 'my'?",
+      "accepted_answers": [
+        "-nha"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-my\",\"answer\":\"unh\"}",
+          "parsed_answer": "unh",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-possessor-your",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-possessor-your",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor suffix means 'your'?",
+      "accepted_answers": [
+        "-ma"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-possessor-your\",\"answer\":\"isha\"}",
+          "parsed_answer": "isha",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-from",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-from",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix means 'from'?",
+      "accepted_answers": [
+        "-t”a"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-from\",\"answer\":\"stxi\"}",
+          "parsed_answer": "stxi",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-interrogative",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-interrogative",
+      "step_type": "morpheme_mapping",
+      "question": "Which suffix is interrogative?",
+      "accepted_answers": [
+        "-txi"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-interrogative\",\"answer\":\"-txi\"}",
+          "parsed_answer": "-txi",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2016-individual-5-microstep-jaqaru-case-vowel-rule",
+      "source_problem_id": "2016-individual-5",
+      "step_id": "jaqaru-case-vowel-rule",
+      "step_type": "transformation_rule",
+      "question": "When an oral case suffix follows a root-final vowel, is that root vowel consumed or preserved?",
+      "accepted_answers": [
+        "preserved"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"jaqaru-case-vowel-rule\",\"answer\":\"preserved\"}",
+          "parsed_answer": "preserved",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-possessor-order",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-possessor-order",
+      "step_type": "word_order",
+      "question": "Where does a possessor occur relative to the possessed noun?",
+      "accepted_answers": [
+        "before the possessed noun"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-possessor-order\",\"answer\":\"The possessor occurs before the possessed noun in Abui.\"}",
+          "parsed_answer": "The possessor occurs before the possessed noun in Abui.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 27
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-first-possessor",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-first-possessor",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor prefix marks first-person singular?",
+      "accepted_answers": [
+        "n-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-first-possessor\",\"answer\":\"a\"}",
+          "parsed_answer": "a",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-second-possessor",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-second-possessor",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor prefix marks second-person singular?",
+      "accepted_answers": [
+        "zero"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-second-possessor\",\"answer\":\"atáng\"}",
+          "parsed_answer": "atáng",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-third-possessor",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-third-possessor",
+      "step_type": "morpheme_mapping",
+      "question": "Which possessor prefix marks third-person singular?",
+      "accepted_answers": [
+        "h-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-third-possessor\",\"answer\":\"hatáng\"}",
+          "parsed_answer": "hatáng",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-body-part",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-body-part",
+      "step_type": "class_mapping",
+      "question": "Except after ri-, which prefix appears on a possessed body-part noun?",
+      "accepted_answers": [
+        "a-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-body-part\",\"answer\":\"riya\"}",
+          "parsed_answer": "riya",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-2-microstep-abui-other-noun",
+      "source_problem_id": "2017-individual-2",
+      "step_id": "abui-other-noun",
+      "step_type": "class_mapping",
+      "question": "Except after ri-, which prefix appears on another possessed noun?",
+      "accepted_answers": [
+        "e-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"abui-other-noun\",\"answer\":\"riya\"}",
+          "parsed_answer": "riya",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-basic-order",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-basic-order",
+      "step_type": "word_order",
+      "question": "What is the basic order of subject, verb, optional object, and optional place?",
+      "accepted_answers": [
+        "subject, verb, object, place"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-basic-order\",\"answer\":\"subject-verb-object-place\"}",
+          "parsed_answer": "subject-verb-object-place",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-first-subject",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-first-subject",
+      "step_type": "morpheme_mapping",
+      "question": "Which verb prefix expresses a first-person-singular subject?",
+      "accepted_answers": [
+        "ŋg-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-first-person-singular\",\"answer\":\"ŋga-\"}",
+          "parsed_answer": null,
+          "strict_json": false,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-possessed-order",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-possessed-order",
+      "step_type": "word_order",
+      "question": "Where does a possessed noun occur relative to its possessive pronoun?",
+      "accepted_answers": [
+        "before the possessive pronoun"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-possessed-order\",\"answer\":\"A possessed noun occurs after its possessive pronoun in Kimbundu.\"}",
+          "parsed_answer": "A possessed noun occurs after its possessive pronoun in Kimbundu.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 33
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-place-concord",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-place-concord",
+      "step_type": "class_mapping",
+      "question": "Which concord prefix marks place expressions?",
+      "accepted_answers": [
+        "mù-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-place-concord\",\"answer\":\"mù\"}",
+          "parsed_answer": "mù",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 18
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-statement-tone",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-statement-tone",
+      "step_type": "prosody_rule",
+      "question": "What is the sentence-final tone of a statement?",
+      "accepted_answers": [
+        "low"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-statement-tone\",\"answer\":\"high\"}",
+          "parsed_answer": "high",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2017-individual-3-microstep-kimbundu-question-tone",
+      "source_problem_id": "2017-individual-3",
+      "step_id": "kimbundu-question-tone",
+      "step_type": "prosody_rule",
+      "question": "What is the sentence-final tone of a question?",
+      "accepted_answers": [
+        "high"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"kimbundu-question-tone\",\"answer\":\"high\"}",
+          "parsed_answer": "high",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-order",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-order",
+      "step_type": "word_order",
+      "question": "What is the order of subject, optional object, predicate, agreement complex, and final ne?",
+      "accepted_answers": [
+        "subject, object, predicate, agreement complex, ne"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-order\",\"answer\":\"subject-optional-object-predicate-agreement-complex-final-ne\"}",
+          "parsed_answer": "subject-optional-object-predicate-agreement-complex-final-ne",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 25
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-singular-transitive",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-singular-transitive",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker follows a first- or second-person-singular transitive subject?",
+      "accepted_answers": [
+        "-bə"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-singular-transitive\",\"answer\":\"lapkʰi\"}",
+          "parsed_answer": "lapkʰi",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-other-transitive",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-other-transitive",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker follows a third-person or plural transitive subject?",
+      "accepted_answers": [
+        "kəmə"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-other-transitive\",\"answer\":\"lapkʰi\"}",
+          "parsed_answer": "lapkʰi",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-hierarchy",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-hierarchy",
+      "step_type": "role_rule",
+      "question": "What is the person hierarchy from highest to lowest?",
+      "accepted_answers": [
+        "first > second > third"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-hierarchy\",\"answer\":\"The person hierarchy from highest to lowest is: pl, sg, him, me, us, they, he\"}",
+          "parsed_answer": "The person hierarchy from highest to lowest is: pl, sg, him, me, us, they, he",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 36
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-present-higher-subject",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-present-higher-subject",
+      "step_type": "morpheme_mapping",
+      "question": "Which agreement prefix marks present tense when the subject outranks the object?",
+      "accepted_answers": [
+        "k-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-present-higher-subject\",\"answer\":\"The prefix 'nirum' marks present tense when the subject outranks the object.\"}",
+          "parsed_answer": "The prefix 'nirum' marks present tense when the subject outranks the object.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 35
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-present-higher-object",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-present-higher-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which agreement prefix marks present tense when the object outranks the subject?",
+      "accepted_answers": [
+        "r-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-present-higher-object\",\"answer\":\"The prefix 'ŋa-' marks present tense when the object outranks the subject.\"}",
+          "parsed_answer": "The prefix 'ŋa-' marks present tense when the object outranks the subject.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 33
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-2-microstep-hakhun-sleep",
+      "source_problem_id": "2018-individual-2",
+      "step_id": "hakhun-sleep",
+      "step_type": "lexical_mapping",
+      "question": "Which predicate means 'sleep'?",
+      "accepted_answers": [
+        "ʒip"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"hakhun-sleep\",\"answer\":\"tuʔ\"}",
+          "parsed_answer": "tuʔ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-first-p",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-first-p",
+      "step_type": "transformation_rule",
+      "question": "For first-person singular, what does the first voiceless p become?",
+      "accepted_answers": [
+        "mb"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-first-p\",\"answer\":\"p becomes b\"}",
+          "parsed_answer": "p becomes b",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-first-t",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-first-t",
+      "step_type": "transformation_rule",
+      "question": "For first-person singular, what does the first voiceless t become?",
+      "accepted_answers": [
+        "nd"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-first-t\",\"answer\":\"t becomes d\"}",
+          "parsed_answer": "t becomes d",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-first-k",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-first-k",
+      "step_type": "transformation_rule",
+      "question": "For first-person singular, what does the first voiceless k become?",
+      "accepted_answers": [
+        "ng"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-first-k\",\"answer\":\"k becomes g in first-person singular\"}",
+          "parsed_answer": "k becomes g in first-person singular",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-second-a",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-second-a",
+      "step_type": "transformation_rule",
+      "question": "For second-person singular in a native word, what does the first eligible a become?",
+      "accepted_answers": [
+        "e"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-second-a\",\"answer\":\"y\"}",
+          "parsed_answer": "y",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-second-o",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-second-o",
+      "step_type": "transformation_rule",
+      "question": "For second-person singular, what does the first eligible o become?",
+      "accepted_answers": [
+        "e"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-second-o\",\"answer\":\"o becomes e\"}",
+          "parsed_answer": "o becomes e",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2018-individual-3-microstep-terena-portuguese-accent",
+      "source_problem_id": "2018-individual-3",
+      "step_id": "terena-portuguese-accent",
+      "step_type": "transformation_rule",
+      "question": "In a Portuguese loan, what does á become for second-person singular?",
+      "accepted_answers": [
+        "eá"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"terena-portuguese-accent\",\"answer\":\"á becomes â in second-person singular\"}",
+          "parsed_answer": "á becomes â in second-person singular",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 23
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-copular-order",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-copular-order",
+      "step_type": "word_order",
+      "question": "What is the order of the two terms in the Nooni copular construction?",
+      "accepted_answers": [
+        "A nú B"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-copular-order\",\"answer\":\"The order of the two terms in the Nooni copular construction is date first, then the subject.\"}",
+          "parsed_answer": "The order of the two terms in the Nooni copular construction is date first, then the subject.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 36
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-clause-order",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-clause-order",
+      "step_type": "word_order",
+      "question": "What is the ordinary order of subject, tense, verb, object, place, and day?",
+      "accepted_answers": [
+        "subject, tense, verb, object, place, day"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-clause-order\",\"answer\":\"Subject, tense, verb, object, place, and day\"}",
+          "parsed_answer": "Subject, tense, verb, object, place, and day",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 26
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-see",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-see",
+      "step_type": "morpheme_order",
+      "question": "What is the discontinuous construction meaning 'see'?",
+      "accepted_answers": [
+        "yɛn (object) lé"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-see\",\"answer\":\"see\"}",
+          "parsed_answer": "see",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-past-today",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-past-today",
+      "step_type": "morpheme_mapping",
+      "question": "Which tense marker means past today?",
+      "accepted_answers": [
+        "bɛ̂ː"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-past-today\",\"answer\":\"yo᷅w\"}",
+          "parsed_answer": "yo᷅w",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-near-future",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-near-future",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker means future fewer than three days ahead?",
+      "accepted_answers": [
+        "nú bóy"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-near-future\",\"answer\":\"fɔ̀ shě ntfǔː\"}",
+          "parsed_answer": "fɔ̀ shě ntfǔː",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 24
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-far-future",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-far-future",
+      "step_type": "morpheme_mapping",
+      "question": "Which marker means future more than three days ahead?",
+      "accepted_answers": [
+        "nú yúː"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-far-future\",\"answer\":\"fɔ̀ shě ntfǔː\"}",
+          "parsed_answer": "fɔ̀ shě ntfǔː",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 24
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-first-person",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-first-person",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks first-person singular on the tense marker or verb?",
+      "accepted_answers": [
+        "n-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-first-person\",\"answer\":\"me\"}",
+          "parsed_answer": "me",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2019-individual-5-microstep-nooni-le-after-consonant",
+      "source_problem_id": "2019-individual-5",
+      "step_id": "nooni-le-after-consonant",
+      "step_type": "allomorph_rule",
+      "question": "What does lé become after a consonant?",
+      "accepted_answers": [
+        "é"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"nooni-le-after-consonant\",\"answer\":\"lé becomes le after a consonant\"}",
+          "parsed_answer": "lé becomes le after a consonant",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 23
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-one",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-one",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 1?",
+      "accepted_answers": [
+        "ena"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-one\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-six",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-six",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 6?",
+      "accepted_answers": [
+        "benomi"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-six\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-nine",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-nine",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 9?",
+      "accepted_answers": [
+        "ije"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-nine\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-ten",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-ten",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 10?",
+      "accepted_answers": [
+        "rati"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-ten\",\"answer\":\"wido\"}",
+          "parsed_answer": "wido",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-sixty",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-sixty",
+      "step_type": "lexical_mapping",
+      "question": "Which Ekari numeral means 60?",
+      "accepted_answers": [
+        "moeto"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-sixty\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-addition",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-addition",
+      "step_type": "arithmetic_rule",
+      "question": "What operation is expressed when a unit with -ma precedes a multiple of ten?",
+      "accepted_answers": [
+        "addition"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-addition\",\"answer\":\"addition\"}",
+          "parsed_answer": "addition",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-subtractive-nine",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-subtractive-nine",
+      "step_type": "arithmetic_rule",
+      "question": "Which unit may precede the next decade to express subtraction by one?",
+      "accepted_answers": [
+        "ije"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-subtractive-nine\",\"answer\":\"wija\"}",
+          "parsed_answer": "wija",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2021-individual-1-microstep-ekari-daemita",
+      "source_problem_id": "2021-individual-1",
+      "step_id": "ekari-daemita",
+      "step_type": "arithmetic_rule",
+      "question": "What value does X daemita moeto represent?",
+      "accepted_answers": [
+        "60 + X"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ekari-daemita\",\"answer\":\"80\"}",
+          "parsed_answer": "80",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-role-order",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-role-order",
+      "step_type": "morpheme_order",
+      "question": "In what role order do person prefixes occur around the fixed verb stem?",
+      "accepted_answers": [
+        "direct object, indirect object, subject"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-role-order\",\"answer\":\"person prefixes occur in the order: subject, object, recipient\"}",
+          "parsed_answer": "person prefixes occur in the order: subject, object, recipient",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 25
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-first-singular",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-first-singular",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks first-person singular?",
+      "accepted_answers": [
+        "s-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-first-singular\",\"answer\":\"a\"}",
+          "parsed_answer": "a",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 15
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-first-plural",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-first-plural",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks first-person plural?",
+      "accepted_answers": [
+        "šʼ-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-first-plural\",\"answer\":\"šʼ\"}",
+          "parsed_answer": "šʼ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-second-singular",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-second-singular",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks second-person singular?",
+      "accepted_answers": [
+        "w-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-second-singular\",\"answer\":\"sawtʷən\"}",
+          "parsed_answer": "sawtʷən",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 20
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-second-plural",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-second-plural",
+      "step_type": "morpheme_mapping",
+      "question": "Which person prefix marks second-person plural?",
+      "accepted_answers": [
+        "ŝʷ-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-second-plural\",\"answer\":\"šʼ\"}",
+          "parsed_answer": "šʼ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2022-individual-1-microstep-ubykh-a-vowel",
+      "source_problem_id": "2022-individual-1",
+      "step_id": "ubykh-a-vowel",
+      "step_type": "allomorph_rule",
+      "question": "When is the verb vowel a rather than ə?",
+      "accepted_answers": [
+        "when the direct object is plural or a second-person-plural participant is present"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"ubykh-a-vowel\",\"answer\":\"The verb vowel is 'a' when the object is a person or when the verb is in the first person singular or plural with a specific object direction.\"}",
+          "parsed_answer": "The verb vowel is 'a' when the object is a person or when the verb is in the first person singular or plural with a specific object direction.",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 45
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-one",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-one",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 1?",
+      "accepted_answers": [
+        "niŋkin"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-one\",\"answer\":\"kampwɔhii\"}",
+          "parsed_answer": "kampwɔhii",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-five",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-five",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 5?",
+      "accepted_answers": [
+        "kaŋkuro"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-five\",\"answer\":\"kampwɔhii shuunni na kɛ\"}",
+          "parsed_answer": "kampwɔhii shuunni na kɛ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 26
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-ten",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-ten",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 10?",
+      "accepted_answers": [
+        "kɛ"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-ten\",\"answer\":\"kampwɔhii shuunni na kɛ\"}",
+          "parsed_answer": "kampwɔhii shuunni na kɛ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 27
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-twenty",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-twenty",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 20?",
+      "accepted_answers": [
+        "beɲjaaga"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-twenty\",\"answer\":\"baashuunni\"}",
+          "parsed_answer": "baashuunni",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-eighty",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-eighty",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 80?",
+      "accepted_answers": [
+        "ŋkuu"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-eighty\",\"answer\":\"kampwɔhii shuunni na kɛ\"}",
+          "parsed_answer": "kampwɔhii shuunni na kɛ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 27
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-four-hundred",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-four-hundred",
+      "step_type": "lexical_mapping",
+      "question": "Which Supyire numeral means 400?",
+      "accepted_answers": [
+        "kampwoo"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-four-hundred\",\"answer\":\"kampwɔhii shuunni na kɛ\"}",
+          "parsed_answer": "kampwɔhii shuunni na kɛ",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 28
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-addition",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-addition",
+      "step_type": "arithmetic_rule",
+      "question": "What arithmetic operation does na express?",
+      "accepted_answers": [
+        "addition"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-addition\",\"answer\":\"addition\"}",
+          "parsed_answer": "addition",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2023-individual-5-microstep-supyire-400-multiplier",
+      "source_problem_id": "2023-individual-5",
+      "step_id": "supyire-400-multiplier",
+      "step_type": "allomorph_rule",
+      "question": "Which form of 400 occurs before a multiplier?",
+      "accepted_answers": [
+        "kampwɔhii"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"supyire-400-multiplier\",\"answer\":\"403\"}",
+          "parsed_answer": "403",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 21
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-verb-order",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-verb-order",
+      "step_type": "morpheme_order",
+      "question": "Which comes first in the verb: subject prefix or tense prefix?",
+      "accepted_answers": [
+        "subject prefix"
+      ],
+      "task_passed": true,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-verb-order\",\"answer\":\"subject prefix\"}",
+          "parsed_answer": "subject prefix",
+          "strict_json": true,
+          "correct": true,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-first-subject",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-first-subject",
+      "step_type": "morpheme_mapping",
+      "question": "Which subject prefix marks first-person singular?",
+      "accepted_answers": [
+        "t-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-first-subject\",\"answer\":\"tə\"}",
+          "parsed_answer": "tə",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-third-nonsingular",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-third-nonsingular",
+      "step_type": "morpheme_mapping",
+      "question": "Which subject prefix marks third-person dual or plural?",
+      "accepted_answers": [
+        "ne-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-third-nonsingular\",\"answer\":\"du\"}",
+          "parsed_answer": "du",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 16
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-present",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-present",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks present tense?",
+      "accepted_answers": [
+        "ku-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-present\",\"answer\":\"-net\"}",
+          "parsed_answer": "-net",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 14
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-future",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-future",
+      "step_type": "morpheme_mapping",
+      "question": "Which prefix marks future tense?",
+      "accepted_answers": [
+        "je-"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-future\",\"answer\":\"-tək\"}",
+          "parsed_answer": "-tək",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 17
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-third-singular-object",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-third-singular-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which ending marks a third-person-singular object?",
+      "accepted_answers": [
+        "-n"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-third-singular-object\",\"answer\":\"-tək\"}",
+          "parsed_answer": "-tək",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-third-plural-object",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-third-plural-object",
+      "step_type": "morpheme_mapping",
+      "question": "Which ending marks a third-person-plural object?",
+      "accepted_answers": [
+        "-new"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-third-plural-object\",\"answer\":\"-tək\"}",
+          "parsed_answer": "-tək",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    },
+    {
+      "record_id": "2024-individual-1-microstep-arapesh-tj",
+      "source_problem_id": "2024-individual-1",
+      "step_id": "arapesh-tj",
+      "step_type": "transformation_rule",
+      "question": "What does the sequence t+j become?",
+      "accepted_answers": [
+        "cc"
+      ],
+      "task_passed": false,
+      "samples": [
+        {
+          "text": "{\"step_id\":\"arapesh-tj\",\"answer\":\"t+j becomes təj\"}",
+          "parsed_answer": "t+j becomes təj",
+          "strict_json": true,
+          "correct": false,
+          "finish_reason": "stop",
+          "token_count": 19
+        }
+      ]
+    }
+  ]
+}

benchmark/IOL/ioling_hf/reports/qa/manual_checks.json

@@ -0,0 +1,67 @@
+[
+  {
+    "record_id": "iol-2025-individual-p1",
+    "problem_pdf": "data/raw/pdfs/iol-2025-indiv-prob.en.pdf",
+    "solution_pdf": "data/raw/pdfs/iol-2025-indiv-sol.en.pdf",
+    "problem_images": [
+      "reports/qa/iol-2025-individual-p1-problem-p1.png",
+      "reports/qa/iol-2025-individual-p1-problem-p2.png"
+    ],
+    "solution_images": [
+      "reports/qa/iol-2025-individual-p1-solution-p1.png",
+      "reports/qa/iol-2025-individual-p1-solution-p2.png"
+    ],
+    "problem_extraction_confidence": "heading",
+    "solution_extraction_confidence": "heading",
+    "manual_check_status": "sample_rendered_for_review"
+  },
+  {
+    "record_id": "iol-2003-individual-p1",
+    "problem_pdf": "data/raw/pdfs/iol-2003-indiv-prob.en.pdf",
+    "solution_pdf": "data/raw/pdfs/iol-2003-indiv-sol.en.pdf",
+    "problem_images": [
+      "reports/qa/iol-2003-individual-p1-problem-p1.png",
+      "reports/qa/iol-2003-individual-p1-problem-p2.png"
+    ],
+    "solution_images": [
+      "reports/qa/iol-2003-individual-p1-solution-p1.png",
+      "reports/qa/iol-2003-individual-p1-solution-p2.png"
+    ],
+    "problem_extraction_confidence": "ocr_heading",
+    "solution_extraction_confidence": "ocr_heading",
+    "manual_check_status": "sample_rendered_for_review"
+  },
+  {
+    "record_id": "iol-2005-individual-p5",
+    "problem_pdf": "data/raw/pdfs/iol-2005-indiv-prob.en.pdf",
+    "solution_pdf": "data/raw/pdfs/iol-2005-indiv-sol.en.pdf",
+    "problem_images": [
+      "reports/qa/iol-2005-individual-p5-problem-p4.png"
+    ],
+    "solution_images": [
+      "reports/qa/iol-2005-individual-p5-solution-p4.png"
+    ],
+    "problem_extraction_confidence": "ocr_heading",
+    "solution_extraction_confidence": "ocr_heading",
+    "manual_check_status": "sample_rendered_for_review"
+  },
+  {
+    "record_id": "iol-2014-team-p1",
+    "problem_pdf": "data/raw/pdfs/iol-2014-team-prob.en-gb.pdf",
+    "solution_pdf": "data/raw/pdfs/iol-2014-team-sol.m.pdf",
+    "problem_images": [
+      "reports/qa/iol-2014-team-p1-problem-p1.png",
+      "reports/qa/iol-2014-team-p1-problem-p2.png",
+      "reports/qa/iol-2014-team-p1-problem-p3.png",
+      "reports/qa/iol-2014-team-p1-problem-p4.png"
+    ],
+    "solution_images": [
+      "reports/qa/iol-2014-team-p1-solution-p1.png",
+      "reports/qa/iol-2014-team-p1-solution-p2.png",
+      "reports/qa/iol-2014-team-p1-solution-p3.png"
+    ],
+    "problem_extraction_confidence": "team_full_document",
+    "solution_extraction_confidence": "team_full_document",
+    "manual_check_status": "sample_rendered_for_review"
+  }
+]

benchmark/IOL/ioling_hf/reports/qa/qa_samples.html

@@ -0,0 +1,473 @@
+<!doctype html>
+<html><head><meta charset="utf-8"><title>IOLing QA Samples</title>
+<style>
+body { font:14px/1.45 system-ui, sans-serif; margin:24px; color:#202124; }
+section { border-top:1px solid #d8dee4; padding-top:20px; margin-top:20px; }
+.images { display:flex; gap:14px; overflow:auto; align-items:flex-start; }
+img { max-width:520px; border:1px solid #d8dee4; }
+pre { white-space:pre-wrap; border:1px solid #d8dee4; padding:12px; background:#fafafa; }
+</style></head><body><h1>IOLing QA Samples</h1>
+<section>
+  <h2>iol-2025-individual-p1</h2>
+  <p>2025 individual P1 · heading / heading</p>
+  <h3>Problem PDF Pages</h3>
+  <div class="images"><img src="iol-2025-individual-p1-problem-p1.png" alt="iol-2025-individual-p1-problem-p1.png">
+<img src="iol-2025-individual-p1-problem-p2.png" alt="iol-2025-individual-p1-problem-p2.png"></div>
+  <h3>Parsed Prompt</h3>
+  <pre>Solve this International Linguistics Olympiad problem. Reason step by step, then give the final answer inside \boxed{}.
+
+Problem 1 (20 points).  Here are some numbers in Dzongkha and their numerical values:
+
+      1 — ci      3 — sum      8 — ge      12 — cuɲi      17 — cupdyn      19 — cygu
+
+   For higher numbers, Dzongkha uses two different systems (referred to here as A and B).
+   Below are some numbers written in both systems, as well as their numerical values:
+
+          System A     System B  Value       System A        System B  Value
+          ke ci da ɲi     tsaɲi         22       ke ko-da sum    ŋaŋa         55
+          ke ci da ŋa    tsaŋa         25       ke sum da cuɖu  dønɖu        76
+          ke pɟe-da ɲi  sumcu        30       ke ʑi            gepcu        80
+          ke ci da cyʑi   soʑi          34       ke ʑi da gu       ɟagu          89
+          ke ɲi da dyn  ʑedyn        47       ke ceŋa         sumɟa       300
+
+    Finally, some equalities are given with left-hand side written in system A and right-hand side writ-
+ten in system B. Some numbers are missing.
+                                         System A      System B
+              (1)                 cusum + ke pɟe-da ʑi =  ɟasum
+              (2)                                 ɲiɕu ɲi =  ɲiɕu × ʑipcu
+              (3)                  ɲiɕu ci da ke sum da gu =  (ŋapcu × gu) + cygu
+              (4)            ɲiɕu pɟe-da ɲi + ke pɟe-da ɖu =  ŋapɟa + ɲiɟa cutãm
+              (5)                              (ɲi × ko) + pɟe =   ɲi
+              (6)   (ɲiɕu ko-da sum × pɟe) + ke pɟe-da sum =  ɖukɟa
+              (7)               ɲiɕu ci da ke cuɖu da cuɖu =   (ɟaʑi × ʑi) + ʑipɟa
+              (8)            ɲi × ɲiɕu ci da ke cutãm da gu =   (    X × ɲiɕu) + copge
+              (9)                             Y + ke ci da ʑi =  ɟaɖu
+            (10)                         Z + ke ko-da ɖu =  dynɟa + sumɟa
+
+  (a)  Fill in the blanks X–Z with Dzongkha num-  !△  The Dzongkha language belongs to the Sino-
+      bers.                                     Tibetan family.  It is spoken by approx. 171,000
+                                               people in Bhutan.
+  (b) Write with digits the equalities (1–10).
+                                              The words are given in a simplified transcrip-
+   (c) Write  in  Dzongkha  in  both  systems:  tion.  ɖ,  ɟ, ɲ, ŋ, ɕ and ʑ are consonants.   ã, ø
+      75; 570.                               and y are vowels.            —Vlad A. Neacșu
+
+Twenty-second International Linguistics Olympiad (2025)                                          2
+Individual Contest Problems</pre>
+  <h3>Solution PDF Pages</h3>
+  <div class="images"><img src="iol-2025-individual-p1-solution-p1.png" alt="iol-2025-individual-p1-solution-p1.png">
+<img src="iol-2025-individual-p1-solution-p2.png" alt="iol-2025-individual-p1-solution-p2.png"></div>
+  <h3>Parsed Canonical Answer</h3>
+  <pre>Problem 1.
+
+                         1       2     3      4     5      6      7      8     9
+     k   k                  ci       ɲi   sum    ʑi    ŋa    ɖu    dyn   ge    gu
+          10( + k)        cu-     cu-   cu-    cy-   ce-    cu-    cup-   cop-   cy-
+   k k × 10        cutãm   ɲi-   sum-   ʑip-  ŋap-  ɖuk-  dyn-  gep-   (gup-)     k
+   k     k   k × 10( + . . .)            tsa-   so-     ʑe-   ŋa-     (re-)   døn-   ɟa-     (go-)
+     • System A:
+
+       – β × 20 ( + α) = ke β (da α )
+
+        ∗β × 20 + 10 →ke pɟe-da β + 1
+
+        ∗β × 20 + 15 →ke ko-da β + 1
+
+       – γ × 400 ( + β × 20 + α) = ɲiɕu γ (da ke β da α )
+
+        ∗γ × 400 + 200 →ɲiɕu pɟe-da γ + 1
+        ∗γ × 400 + 300 →ɲiɕu ko-da γ + 1
+
+     • System B:
+
+       – β × 10 = βk-cu, β &gt; 2;
+        β × 10 = βk-ɕu, β = 2
+       – β × 10 + α = βk- α
+       – γ × 100 ( + ω) = γk-ɟa ω
+
+  (a)      [X] = 60 = ɖukcu
+           [Y] = 62 = ke sum da ɲi
+           [Z] = 885 = ɲiɕu ɲi da ke ʑi da ŋa
+
+  (b)   (1) 13 + 70 = 83
+         (2) 800 = 20 × 40
+         (3) 469 = 50 × 9 + 19
+         (4) 600 + 110 = 500 + 210
+         (5) 2 × 43 + 21 = 2
+         (6) 1100 × 12 + 50 = 600
+         (7) 736 = 84 × 4 + 400
+         (8) 2 × 609 = 60X × 20 + 18
+         (9) 62Y + 24 = 86
+       (10) 885Z + 115 = 700 + 300
+
+   (c)     • 75 = ke ko-da ʑi = dønŋa
+           • 570 = ɲiɕu ci da ke pɟe-da gu = ŋapɟa dyncu
+
+Twenty-second International Linguistics Olympiad (2025)                                          2
+Individual Contest Solutions</pre>
+</section>
+
+<section>
+  <h2>iol-2003-individual-p1</h2>
+  <p>2003 individual P1 · ocr_heading / ocr_heading</p>
+  <h3>Problem PDF Pages</h3>
+  <div class="images"><img src="iol-2003-individual-p1-problem-p1.png" alt="iol-2003-individual-p1-problem-p1.png">
+<img src="iol-2003-individual-p1-problem-p2.png" alt="iol-2003-individual-p1-problem-p2.png"></div>
+  <h3>Parsed Prompt</h3>
+  <pre>Solve this International Linguistics Olympiad problem. Reason step by step, then give the final answer inside \boxed{}.
+
+Problem 1 (20 marks)
+In 1916 the Russian scholar Jacob Linzbach invented a universal writing system, which he thought
+should be understandable to all people, regardless of their native tongue. Linzbach called his new
+language “Transcendental Algebra’.
+Several sentences have been written in Linzbach’s language and translated into English:
+1. (Adis +4)s The father and the brother are talking.
+2. n(&gt;h-t The giants are working without haste.
+3. Eo =WN The orphans are writing a letter.
+4. (—nh)*% -t=b It wasn’t us who wrote about you (sg.).
+5. Sv —-t= —A3 It was not by her that the letter was written.
+6. (Adis) -S =[F- The father doesn’t like the work.
+7% (SD-@)4-t= Adis The wicked giant ate the parents.
+8. A;t She is not in a hurry.
+Assignment 1. Translate into English:
+9 IP-v
+AAiA =&lt;)4 — AAia , AAiA
+10. ( KAT -S) tts “ge + is
+i. APTS +
+12 RVA8-1=4-A
+Assignment 2. Write in “Transcedental Algebra’:
+13. It wasn’t about them that my husband and I (say: I and the husband) talked.
+14. The people are working reluctantly.
+15. The good widow loves the unemployed dwarf.
+16. You (pl.) will be talked about.
+Explain your solution. (Ksenia Guiliarova)
+
+1st IOL: Borovetz 03. Individual Contest</pre>
+  <h3>Solution PDF Pages</h3>
+  <div class="images"><img src="iol-2003-individual-p1-solution-p1.png" alt="iol-2003-individual-p1-solution-p1.png">
+<img src="iol-2003-individual-p1-solution-p2.png" alt="iol-2003-individual-p1-solution-p2.png"></div>
+  <h3>Parsed Canonical Answer</h3>
+  <pre>Solution of Problem 1
+1. Nouns:
+e A ‘man’, A ‘woman’, i ‘boy’, A ‘girl’, etter’, [-- ‘work’.
+
+— Combinations: AA ‘man + woman = husband + wife’, iA ‘boy + girl = brother
++ sister’, AAiA ‘man + woman + boy + girl = family’.
+
+— Family members are singled out by division and cancellation: 4¢% ‘family /(woman
++ kids) = father’, is ‘kids/girl = brother’, AAA ‘family /kids = parents’.
+
+— Missing (deceased) family members are preceded by a minus sign: “ ‘kids
+(—parents)/(—parents) = orphans’ (apparently orphaned children of one and the
+same family).
+
+e I ‘person’, (&gt; 1) ‘giant’.
+
+2. Pronouns are composed of the character { or A (for feminine gender) and the subscripts 1
+to 3, which indicate the person.
+
+3. The plural of nouns and pronouns is expressed by the coefficient n. The plus sign plays the
+part of the conjunction ‘and’.
+
+4. Verbs: &lt; ‘talk’, [-- ‘work’, t ‘hurry’, 7 ‘write’, [&gt; ‘like, love’, (Q) ‘eat’. If what the
+verb denotes is absent or uncharacteristic, a minus sign expresses that: —[&gt; ‘not inclined
+to affection = wicked’. (We can assume that a characteristic property is expressed by a plus
+sign, hence +&lt;&gt; ‘good’, a concept we need.)
+
+5. Sentence structure:
+
+e the subject is the base of the power;
+e the predicate is the exponent, whereby negation is expressed by a minus sign (—C&gt;
+‘not like’) and passive voice by a radical sign (V7 ‘be written’); additional activities
+can be added or subtracted (i ‘he is working and doesn’t hurry = he is working
+without haste’);
+© past tense is marked by —t d- —t ‘he worked’), future tense by +t;
+e the direct object, if there is one, follows an equals sign.
+Assignment 1. 9. He loves with an unrequited love (i. e. loves without being loved).
+10. The taciturn (or mute) daughter will write about the father and the mother.
+11. You (sg. fem.) worked quickly (or hastily) and silently.
+12. The letter was eaten by the hungry sister.
+Assignment 2. 13. (A; + AAs —t=—ni3
+14. (njjF-2
+i. (AGP ESI =(&lt;D-1-
+16. (nix)¥S+t
+
+1st IOL: Borovetz ’03. Solutions to the Problems of the Individual Contest</pre>
+</section>
+
+<section>
+  <h2>iol-2005-individual-p5</h2>
+  <p>2005 individual P5 · ocr_heading / ocr_heading</p>
+  <h3>Problem PDF Pages</h3>
+  <div class="images"><img src="iol-2005-individual-p5-problem-p4.png" alt="iol-2005-individual-p5-problem-p4.png"></div>
+  <h3>Parsed Prompt</h3>
+  <pre>Solve this International Linguistics Olympiad problem. Reason step by step, then give the final answer inside \boxed{}.
+
+Problem 5 (20 marks)
+
+In Lithuanian® nouns the accent may move according to the number and the case of the
+nouns, i.e., different syllables may be accented in different forms of the same word. The
+pattern of accent movement is called the accent paradigm of the noun.
+
+There are two types of syllables in Lithuanian. If a syllable of the first type is accented, that
+syllable has falling intonation marked «’», e.g.: ie, 6, al. If a syllable of the second type is
+accented, that syllable has rising intonation marked «™», e.g.: afi, 6, ié.
+
+Within the same root or the same ending, the syllable type always remains the same. For
+example, the root /iep, when accented, always has falling intonation, whereas the ending of the
+Nominative Plural os always has rising intonation.
+
+The following examples illustrate the four main types of Lithuanian accent paradigms (they
+look somewhat different in modern Lithuanian, but this is irrelevant for the problem):
+
+Paradigm 1 2 3 4
+Example linden hand head winter
+Nom. Sg. — liepo ranko galvo Ziemo
+Gen. Sg. liepos ratkos — galvos_ — Ziemos
+Nom. PI. liepos ratkos — gdlvos —_ziémos
+Acc. PI. liepaNs_ rankaNs_ galvaNs_ ziemaNs
+
+In the late 19&quot; century, the great Swiss linguist Ferdinand de Saussure studied the accent
+paradigms of Lithuanian nouns and came to the conclusion that at an earlier stage of the
+development of Lithuanian there were not four, but only two accent paradigms. Later, as a
+result of a specific rule, which is now known as Saussure&#x27;s Law, the accent moved under certain
+conditions, and each paradigm split in two.
+
+Assignment 1. Determine which accent paradigms originally belonged together.
+
+Assignment 2. Determine what the initial accent paradigms looked like.
+
+Assignment 3. Formulate Saussure’s Law.
+
+Note. z is a specific Lithuanian consonant, N shows a specific (nasal) pronunciation of the
+preceding vowel.
+
+Good luck!
+Authors: Boris L. Iomdin (#1), Xenia A. Guiliarova (#2), Ivan A. Derzhanski (#3, #4),
+Alexander M. Lubotsky (#5).
+Editors: Alexander S. Berdichevsky, Dmitry V. Gerasimov, Xenia A. Guiliarova (editor-in-chief),
+Stanislav B. Gurevich, Ivan A. Derzhanski, Boris L. Iomdin, Leonid I. Kulikov,
+Alexander B. Letuchiy, Alexander M. Lubotsky, Elena V. Muravenko, Maria L. Rubinstein.
+English translation: Ivan A. Derzhanski, Boris L. Iomdin.
+5 The Lithuanian language is of the Baltic branch of the Indo-European language family. It is spoken by 3 million
+people in Lithuania and some other countries.</pre>
+  <h3>Solution PDF Pages</h3>
+  <div class="images"><img src="iol-2005-individual-p5-solution-p4.png" alt="iol-2005-individual-p5-solution-p4.png"></div>
+  <h3>Parsed Canonical Answer</h3>
+  <pre>Problem 5
+First of all, for every syllable we must determine to which of the two types it belongs. This
+is easy to do, since, according to the problem statement, the syllable type always remains the
+same within the same root or the same ending. Hence, the root has falling intonation in
+paradigms 1 and 3, and rising intonation in paradigms 2 and 4. The endings in Nom.Sg. and
+Acc.PI. always have falling intonation, whereas in Gen.Sg. and Nom.PI. they always have rising
+intonation (the last one is explicated in the problem).
+
+Let us represent these data in a table (where a designates any vowel, and stressed
+syllables are set in boldface):
+
+Paradigm 1 2 3 4
+Nom.Sg. da aa aa aa
+Gen.Sg. aa aa aa aa
+Nom.PIl. aa aa aa aa
+Acc.PI. da aa da aa
+
+Which paradigms belonged together?
+
+In 1 and 3 the root has falling intonation, but the accent patterns are different. This means
+that 1 and 3 must have been different from the outset. The same is true for paradigms 2 u 4.
+Therefore only two options remain:
+
+A.1+2and3+4
+
+or
+
+B.1+4and3+2
+
+Option B would have us explain more differences in the place of the accent than option A
+(5 versus 3), so we start with option A. Comparing 1 and 2, we notice that the places of the
+accent are only different in Nom.Sg. and Acc.Pl., dd in 1 corresponding to da in 2 in both cases.
+Comparing 3 and 4, we notice that the places of the accent are only different in Acc.Pl; in this
+case, too, dda in 3 corresponds to da in 4. Option B does not yield an acceptable solution. For
+instance, the different places of the accent in Gen.Sg (aa) and Nom.Pl. (aa) in 4, the intonation
+of both syllables being the same, and their céincidence in 1 cannot be accounted for. We
+conclude that option A is correct.
+
+Assignment 1. Paradigms 1 and 2, on the one hand, and paradigms 3 and 4, on the
+other hand, originally belonged together.
+
+In order to determine what the two initial paradigms looked like, we have to answer the
+question why dd and dd have different places of accent. Maybe da changed to ad? But we can
+see the sequence dda in 3 (Nom.Sg.), and ad does not occur in any of the paradigms 1-4.
+Therefore ad always changed to aa, and not the other way around.
+
+Assignment 2. Paradigm 1 + 2 looked as 1 looks now (the root was always accented),
+and the paradigm 3 + 4 looked as 3 looks now (the endings were accented in the Singular and
+the root was accented in the Plural).
+
+Assignment 3. Saussure’s Law says that in the sequence syllable with rising intonation —
+syllable with falling intonation (ad) the accent shifted from the first syllable to the second one
+(aa).</pre>
+</section>
+
+<section>
+  <h2>iol-2014-team-p1</h2>
+  <p>2014 team P1 · team_full_document / team_full_document</p>
+  <h3>Problem PDF Pages</h3>
+  <div class="images"><img src="iol-2014-team-p1-problem-p1.png" alt="iol-2014-team-p1-problem-p1.png">
+<img src="iol-2014-team-p1-problem-p2.png" alt="iol-2014-team-p1-problem-p2.png">
+<img src="iol-2014-team-p1-problem-p3.png" alt="iol-2014-team-p1-problem-p3.png">
+<img src="iol-2014-team-p1-problem-p4.png" alt="iol-2014-team-p1-problem-p4.png"></div>
+  <h3>Parsed Prompt</h3>
+  <pre>Solve this International Linguistics Olympiad problem. Reason step by step, then give the final answer inside \boxed{}.
+
+en(B)
+
+        Twelfth International Olympiad in Linguistics
+
+                    Beijing (China), 21 25 July 2014
+
+                       Team Contest Problem
+
+   Here is the text of The Universal Declaration of Human Rights in English and Armenian.
+The Armenian sentences are given in Roman transcription, in alphabetical order.
+   Determine the correspondences between the Armenian and the English sentences.
+                                                          Boris Iomdin, Ivan Derzhanski
+
+   1.  (a) All human beings are born free and equal in dignity and rights.
+       (b) They are endowed with reason and conscience and should act towards one another
+           in a spirit of brotherhood.
+
+   2.  (a) Everyone is entitled to all the rights and freedoms set forth in this declaration,
+          without distinction of any kind, such as race, colour, sex, language, religion, polit-
+             ical or other opinion, national or social origin, property, birth or other status.
+       (b) Furthermore, no distinction shall be made on the basis of the political, jurisdic-
+           tional or international status of the country or territory to which a person belongs,
+         whether it be independent, trust, non-self-governing or under any other limitation
+            of sovereignty.
+
+   3. Everyone has the right to life, liberty and security of person.
+
+   4. No one shall be held in slavery or servitude; slavery and the slave trade shall be prohibited
+      in all their forms.
+
+   5. No one shall be subjected to torture or to cruel, inhuman or degrading treatment or
+     punishment.
+
+   6. Everyone has the right to recognition everywhere as a person before the law.
+
+   7.  (a) All are equal before the law and are entitled without any discrimination to equal
+           protection of the law.
+       (b) All are entitled to equal protection against any discrimination in violation of this
+           declaration and against any incitement to such discrimination.
+
+   8. Everyone has the right to an e ective remedy by the competent national tribunals for
+      acts violating the fundamental rights granted him by the constitution or by law.
+
+   9. No one shall be subjected to arbitrary arrest, detention or exile.
+
+  10. Everyone is entitled in full equality to a fair and public hearing by an independent and
+      impartial tribunal, in the determination of his rights and obligations and of any criminal
+     charge against him.
+
+Twelfth International Olympiad in Linguistics (2014).                                     2
+Team Contest Problem
+
+  11.  (a) Everyone charged with a penal o ence has the right to be presumed innocent until
+          proved guilty according to law in a public trial at which he has had all the guarantees
+          necessary for his defence.
+       (b) No one shall be held guilty on account of any act or omission which did not consti-
+           tute a penal o ence, under national or international law, at the time when it was
+          committed.
+       (c) Nor shall a heavier penalty be imposed than the one that was applicable at the
+          time the penal o ence was committed.
+
+  12.  (a) No one shall be subjected to arbitrary interference with his privacy, family, home
+           or correspondence, nor to attacks upon his honour and reputation.
+       (b) Everyone has the right to the protection of the law against such interference or
+           attacks.
+
+  13.  (a) Everyone has the right to freedom of movement and residence within the borders
+            of each state.
+       (b) Everyone has the right to leave any country, including his own, and to return to
+            his country.
+
+  14.  (a) Everyone has the right to seek and to enjoy in other countries asylum from perse-
+           cution.
+       (b) This right may not be invoked in the case of prosecutions genuinely arising from
+           non-political crimes or from acts contrary to the purposes and principles of the
+          United Nations.
+
+  15.  (a) Everyone has the right to a nationality.
+       (b) No one shall be arbitrarily deprived of his nationality nor denied the right to change
+            his nationality.
+
+  16.  (a) Men and women of full age, without any limitation due to race, nationality or
+            religion, have the right to marry and to found a family.
+       (b) They are entitled to equal rights as to marriage, during marriage and at its dissol-
+           ution.
+       (c) Marriage shall be entered into only with the free and full consent of the intending
+           spouses.
+       (d) The family is the natural and fundamental group unit of society and is entitled to
+           protection by society and the State.
+
+  17.  (a) Everyone has the right to own property alone as well as in association with others.
+       (b) No one shall be arbitrarily deprived of his property.
+
+Twelfth Intern</pre>
+  <h3>Solution PDF Pages</h3>
+  <div class="images"><img src="iol-2014-team-p1-solution-p1.png" alt="iol-2014-team-p1-solution-p1.png">
+<img src="iol-2014-team-p1-solution-p2.png" alt="iol-2014-team-p1-solution-p2.png">
+<img src="iol-2014-team-p1-solution-p3.png" alt="iol-2014-team-p1-solution-p3.png"></div>
+  <h3>Parsed Canonical Answer</h3>
+  <pre>1.  Bolor mardik c’nvum en azat u havasar irenc aržanapatvut‘jamb u iravunk‘nerov.
+   Nrank‘ unen banakanut‘jun u xiłč’ ev mimjanc petk‘ ē ełbajrabar veraberven.
+2. Amen ok‘ uni ajs hřčakagrum bervac’ bolor iravunk‘nern u azatut‘junnerǝ ařanc orevē xtrut‘jan,
+   himnvac’ cełajin, maški gujni, seři, lezvi, kroni, k‘ałak‘akan kam ajl hamozmunk‘neri, azgajin kam
+    socialakan c’agman, unecvac’k‘i, dasajin patkanelut‘jan kam orevē ajl kargavič’aki vra.
+    Avelin, oč mi xtrakanut‘jun čpetk‘ ē lini himnvac’ erkri kam tarac’k‘i, k‘ałak‘akan, iravakan, kam
+    miǯazgajin kargavič’aki vra, lini da ankax, xnamarkjal, očink‘nakařavarvoł kam ink‘nišxanut‘jan orevē
+     ajl sahmanap‘akumov petakan kazmavorum, orin patkanum ē mardǝ.
+3.  Jurak‘ančjur ok‘ uni aprelu, azatut‘jan u anʒi anʒeřnmxeliut‘jan iravunk‘.
+4.  Oč ok‘ čpetk‘ ē lini strkut‘jan kam anazat vič’akum; petk‘ ē argelven strkatirut‘jan u strukneri ařuc’axi
+    bolor ʒeverǝ.
+5.  Oč ok‘ čpetk‘ ē ent‘arkvi kttank‘neri, dažan, anmardkajin kam storacucič verabermunk‘i kam patži.
+6. Amen ok‘, ur ēl or lini, iravunk‘ uni č’anačvel orpes iravasubjekt.
+7.  Bolorǝ havasar en ōrenk‘i ařǯev, ařanc orevē xtrakanut‘jan, unen ōrenk‘ov havasar paštpanvelu
+    iravunk‘.
+   Bolorǝ unen havasar paštpanut‘jan iravunk‘ ǝnddem cankacac’ xtrakanut‘jan, orov xaxtvum ē ajs
+    hřčakagirǝ, ev ǝnddem nman xtrakanut‘jan młoł cankacac’ sadrank‘i.
+8. Amen ok‘ uni azgajin liazor dataranneri miǯocov ir iravunk‘neri ardjunavet verakangnman iravunk‘,
+    et‘e xaxtvum en sahmanadrut‘jamb kam ōrenk‘ov sahmanvac’ nra himnakan iravunk‘nerǝ.
+9.  Oč ok‘ či karoł ent‘arkvel kamajakan kalank‘i, bantarkut‘jan kam artak‘sman.
+10. Amen ok‘  ir iravunk‘neri u partakanut‘junneri sahmanman hamar ev  ir dem cankacac’ k‘reakan
+   meładrank‘i depk‘um,  liakatar  havasarut‘jan himan  vra,  iravunk‘  uni, or  ir  gorc’ǝ  ardaraci u
+   hraparakajnoren lsvi ankax u ankoł mnapah datarani kołmic.
+11. Hancagorc’ut‘jan hamar meładrvoł jurak‘ančjur mard iravunk‘ uni anmeł hamarvel k‘ani deř nra mełk‘ǝ
+     či apacucvac’ ōrenk‘ov  naxatesvac’  hraparakajin  datak‘nnut‘jamb,  ori žamanak apahovven nra
+    paštpanut‘jan hamar bolor anhražešt erašxik‘nerǝ.
+   Oč ok‘ či karoł datapartvel orevicē arark‘i kam bac t‘ołman hamar, orǝ katarvel ē ajnpisi žamanak, erb
+    ajn azgajin kam miǯazgajin ōrenk‘ov hancagorc’ut‘jun či hamarvel.
+    Či karoł sahmanvel naev aveli c’anr patiž, k‘an ajn, orǝ kirařvum ēr hancagorc’ut‘jan katarman
+   žamanak.
+12. Oč ok‘ či karoł ent‘arkvel kamajakan miǯamtut‘jan ir anʒnakan u ǝntanekan kjank‘i nkatmamb, ir tan,
+    t‘łt‘akcut‘jan ev kam ir patvi u hambavi dem kamajakan otnʒgut‘jan.
+   Amen ok‘ uni ōrenk‘ov paštpanvelu iravunk‘ ǝnddem nman miǯamtut‘jan kam otnʒgut‘jan.
+13. Amen ok‘ uni tełašaržvelu ev bnakut‘jan vajr ǝntrelu azatut‘jan iravunk‘ jurak‘ančjur petut‘jan
+   sahmannerum.
+   Amen ok‘ uni cankacac’ erkric, ajd t‘vum ir erkric heřanalu ev ir erkir veradařnalu iravunk‘.
+14. Amen ok‘ iravunk‘ uni ajl erkrnerum hetapndumic apastan oronel ev apastanic ōgtvel.
+    Ajs iravunk‘ǝ či karoł gorc’adrvel ajnpisi očk‘ałak‘akan hancagorc’ut‘junneri kam arark‘neri hamar
+    harucvac’  hetapndut‘junneri  depk‘um,  oronk‘  hakasum  en  Miavorvac’  azgeri  npataknerin u
+    skzbunk‘nerin.
+15. Amen ok‘ uni k‘ałak‘aciut‘jan iravunk‘.
+   Oč ok‘ či karoł kamajakanoren zrkvel ir k‘ałak‘aciut‘junic kam k‘ałak‘aciut‘junǝ p‘oxelu iravunk‘ic.
+16. Čap‘ahas tłamardik u kanajk‘, ařanc orevē cełajin, azgajin kam kronakan sahmanap‘akman, iravunk‘
+   unen amusnanal ev ǝntanik‘ himnel.
+   Nrank‘ havasar iravunk‘ner unen amusnanalis, amusnut‘jan ǝnt‘ack‘um ev amusnaluc’ut‘jan žamanak.
+   Amusnut‘junǝ karoł ē kajanal miajn amusnacoł kołmeri liaržek‘ u azat hamaʒajnut‘jan depk‘um.
+    Əntanik‘ǝ hasarakut‘jan bnakan u himnakan bǯiǯn ē ev paštpanvelu iravunk‘ uni hasarakut‘jan u
+    petut‘jan kołmic.
+17. Amen ok‘ uni sep‘akanut‘jun unenalu iravunk‘, inčpes menak, ajnpes ēl urišneri het miasin.
+   Oč ok‘ či karoł kamajakanoren zrkvel ir sep‘</pre>
+</section>
+</body></html>