Python4 overall-hard benchmark build 20260827T161031Z-hard-datagen

.gitattributes

@@ -60,3 +60,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
 data_generation/20260813T162500Z-datagen/full/selection.json filter=lfs diff=lfs merge=lfs -text
 data_generation/20260813T162500Z-datagen/full/teacher_calls.jsonl filter=lfs diff=lfs merge=lfs -text
+hard_benchmark/20260827T161031Z-hard-datagen/full/teacher_calls.jsonl filter=lfs diff=lfs merge=lfs -text

hard_benchmark/20260827T161031Z-hard-datagen/full/boa_conformance.log

@@ -0,0 +1,6 @@
+........................................................................ [ 25%]
+........................................................................ [ 51%]
+........................................................................ [ 77%]
+..............................................................           [100%]
+278 passed in 13.58s
+warning: `VIRTUAL_ENV=.venv` does not match the project environment path `/workspace/boa/.venv` and will be ignored; use `--active` to target the active environment instead

hard_benchmark/20260827T161031Z-hard-datagen/full/data/overall_hard_manifest.json

@@ -0,0 +1,76 @@
+{
+  "schema_version": "python4_overall_hard_benchmark_v1",
+  "run_id": "20260827T161031Z-hard-datagen",
+  "created_at": "2026-08-27T16:57:27Z",
+  "source_commit": "2690bcc6b398dcaa59185b781c12a13f5b2b9569",
+  "file": "overall_hard_benchmark.jsonl",
+  "items": 256,
+  "sha256": "c62ea1d8aa4e5b0633dbf0dc042728763931592158bfbe26648c0349768d3ebc",
+  "json_hash": "651b5b3d903f1ad592ee71104925f68d70719df2ed950c56ad82cbf06edd4473",
+  "upstream": {
+    "repo_id": "newfacade/LeetCodeDataset",
+    "repo_type": "dataset",
+    "revision": "215604aeed660029df7de2fea5a4d7b6ed476a08",
+    "train_file": "LeetCodeDataset-train.jsonl",
+    "test_file": "LeetCodeDataset-test.jsonl",
+    "license": "apache-2.0"
+  },
+  "eft_exclusion": {
+    "repo_id": "arcadia-impact/python4-leetcode-eft",
+    "revision": "3877dd099e11bfa7aa3968f5a45dbd78bb2d18d0",
+    "file": "aft.jsonl",
+    "problem_ids": 1024
+  },
+  "selection": {
+    "candidates": 548,
+    "candidate_difficulty": {
+      "Hard": 175,
+      "Medium": 373
+    },
+    "attempted_candidates_this_run": 0,
+    "battery_candidate_depth": 290,
+    "battery_difficulty": {
+      "hard": 162,
+      "medium": 94
+    }
+  },
+  "teacher": {
+    "model": "claude-fable-5",
+    "effort": "low",
+    "repairs": {
+      "0": 121,
+      "1": 92,
+      "2": 36,
+      "3": 7
+    }
+  },
+  "usage": {
+    "call_records": 666,
+    "billed_calls": 665,
+    "tokens": {
+      "input": 2154063,
+      "cache_write": 136292,
+      "cache_read": 3349638,
+      "output": 967287
+    },
+    "rates_usd_per_mtok": {
+      "input": 10.0,
+      "cache_write": 12.5,
+      "cache_read": 1.0,
+      "output": 50.0
+    },
+    "estimated_cost_usd": 74.96
+  },
+  "certification": {
+    "tasks": 256,
+    "difficulty_counts": {
+      "hard": 162,
+      "medium": 94
+    },
+    "tests_per_task_min": 3,
+    "tests_per_task_max": 20,
+    "benchmark_sha256": "e1823be021a06ece1a55f4997b11bce51f39f0b0f479f4343fb0cfab72c9afef",
+    "certified": true
+  },
+  "boa_revision": "a215d2d1875f3d3d986185597c7f12a1d0258568"
+}

hard_benchmark/20260827T161031Z-hard-datagen/full/pilot_summary.json

@@ -0,0 +1,39 @@
+{
+  "run_id": "20260827T161031Z-hard-datagen",
+  "requested": 12,
+  "passed": 12,
+  "pass_fraction": 1.0,
+  "minimum_pass_fraction": 0.8,
+  "keys": [
+    "hard:count-different-palindromic-subsequences",
+    "hard:count-number-of-special-subsequences",
+    "hard:distribute-elements-into-two-arrays-ii",
+    "hard:find-the-maximum-length-of-a-good-subsequence-ii",
+    "hard:find-the-shortest-superstring",
+    "hard:maximum-score-from-grid-operations",
+    "hard:minimum-changes-to-make-k-semi-palindromes",
+    "hard:minimum-distance-to-type-a-word-using-two-fingers",
+    "hard:minimum-edge-weight-equilibrium-queries-in-a-tree",
+    "hard:odd-even-jump",
+    "hard:paint-house-iii",
+    "hard:the-most-similar-path-in-a-graph"
+  ],
+  "accepted": true,
+  "usage": {
+    "call_records": 16,
+    "billed_calls": 16,
+    "tokens": {
+      "input": 64615,
+      "cache_write": 52420,
+      "cache_read": 31452,
+      "output": 31003
+    },
+    "rates_usd_per_mtok": {
+      "input": 10.0,
+      "cache_write": 12.5,
+      "cache_read": 1.0,
+      "output": 50.0
+    },
+    "estimated_cost_usd": 2.88
+  }
+}

hard_benchmark/20260827T161031Z-hard-datagen/full/resolved_config.yaml

@@ -0,0 +1,131 @@
+schema_version: python4_aft_v2
+seed: 424242
+sources:
+  leetcode:
+    repo_id: newfacade/LeetCodeDataset
+    repo_type: dataset
+    revision: 215604aeed660029df7de2fea5a4d7b6ed476a08
+    train_file: LeetCodeDataset-train.jsonl
+    test_file: LeetCodeDataset-test.jsonl
+    license: apache-2.0
+  boa:
+    repo_id: ArcadiaImpact/boa
+    revision: a215d2d1875f3d3d986185597c7f12a1d0258568
+  parents:
+    repo_id: arcadia-impact/python4-gemma3-27b
+    revision: 415ce4d73de6ed42b1cb3ee196909655dda8138d
+  dolci:
+    repo_id: allenai/Dolci-Instruct-SFT
+    repo_type: dataset
+    revision: bd3c8f3a9b2cc5a9682e44b96ddd0bb2ff027221
+    split: train
+  tokenizer:
+    repo_id: unsloth/gemma-3-27b-pt
+    revision: eb493e07419db4938e915c619689bb513181aebb
+teacher:
+  provider: anthropic
+  model: claude-fable-5
+  pilot_min_pass_fraction: 0.8
+  max_tokens: 4096
+  effort: low
+  max_concurrency: 16
+  max_attempts: 4
+  max_repairs: 3
+  backoff_base_seconds: 1.0
+  backoff_max_seconds: 60.0
+parents:
+- arm: control
+  subfolder: control/sft/end
+- arm: mixed_1ep
+  subfolder: dose_1ep_70m/sft/end
+- arm: ordered_1ep
+  subfolder: sdf_ordered_1ep/dolci_10m/end
+- arm: mixed_4ep
+  subfolder: experimental/sft/end
+- arm: ordered_4ep
+  subfolder: sdf_ordered/dolci_10m/end
+rules:
+  held_in:
+  - statement_terminators
+  - out_parameter
+  - manual_allocation
+  - one_based_positive_indexing
+  held_out:
+  - end_inclusive_slice
+  - negative_exclusion
+  - uppercase_boolean
+  - grouped_large_integer
+  - matrix_multiplication
+  min_positive_index_fraction: 0.8
+dataset:
+  aft_rows: 1024
+  min_tests_per_problem: 3
+  max_tests_per_problem: 20
+evaluation:
+  python_timeout_seconds: 5
+replay_aft:
+  dolci_token_fraction: 0.1
+  rows: 1024
+  epochs: 4
+  optimizer_steps: 128
+  sequence_len: 4096
+  candidate_pool_rows: 8192
+  token_fraction_tolerance: 0.001
+  total_token_drift_tolerance: 0.01
+  dataset_file: aft_dolci10.jsonl
+  manifest_file: aft_dolci10_manifest.json
+  dolci_surface_filter: true
+training:
+  stage: aft_python4_gemma3_27b
+  model: gemma3_27b
+  rows: 1024
+  epochs: 4
+  sequence_len: 4096
+  micro_batch_size: 4
+  gradient_accumulation_steps: 8
+  global_batch_size: 32
+  optimizer_steps: 128
+  learning_rate: 0.0001
+  warmup_ratio: 0.05
+  lora:
+    r: 64
+    alpha: 128
+    dropout: 0.0
+    target_layers: 62
+    target_projections:
+    - self_attn.q_proj
+    - self_attn.k_proj
+    - self_attn.v_proj
+    - self_attn.o_proj
+    - mlp.gate_proj
+    - mlp.up_proj
+    - mlp.down_proj
+hub:
+  dataset_repo: arcadia-impact/python4-leetcode-eft
+  dataset_revision: 3877dd099e11bfa7aa3968f5a45dbd78bb2d18d0
+  adapter_repo: arcadia-impact/python4-gemma3-27b-eft
+  logs_repo: arcadia-impact/python4-gemma3-27b-eft-v2-logs
+  private: false
+improved_eval:
+  logs_repo: arcadia-impact/python4-gemma3-27b-eft-v2-eval
+  adapter_revision: 2f1085d7ee918b7750e4a9428a6567105d6f14ed
+  training_run_id: 20260813T154138Z
+  adapter_subfolder_template: runs/{run_id}/arms/{arm}/adapter
+  boa_executable: /workspace/boa/.venv/bin/python4
+  overall_seed: 424242
+  generation:
+    temperature: 0.0
+    samples_per_prompt: 1
+    rule_max_new_tokens: 1024
+    overall_max_new_tokens: 2048
+runtime:
+  gpu: H200
+  cloud: SECURE
+  disk_gb: 600
+  max_hours: 8
+  max_parallel_arms: 5
+  provision_stagger_seconds: 10
+  image: runpod/pytorch:0.7.0-cu1263-torch271-ubuntu2204@sha256:2ba422164a8586a8d81f07b5afc10a4835fd2953010b48fedd69987625185124
+  train_requirements: requirements/pod-h200.txt
+  eval_requirements: requirements/pod-vllm.txt
+scale: 27b

hard_benchmark/20260827T161031Z-hard-datagen/full/teacher_calls.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ce27ceb0dafa697dcaac495f65175822b925851402b31e40f34edf0e024288c
+size 17346236