Organize HLE results and provenance under each model
Browse filesPreserve all trajectories and scores. Fold model-specific HLE summaries and provenance into seven model folders; remove obsolete standalone comparison metadata.
- Gemma-4-12B-it/README.md +14 -0
- Gemma-4-12B-it/hle-verified-provenance.json +384 -0
- Gemma-4-12B-it/summary.json +524 -1
- NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/README.md +14 -0
- NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-provenance.json +398 -0
- NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/summary.json +536 -1
- Olmo-3-32B-Think/README.md +14 -0
- Olmo-3-32B-Think/hle-verified-provenance.json +396 -0
- Olmo-3-32B-Think/summary.json +400 -1
- Qwen3.5-35B-A3B/README.md +14 -0
- Qwen3.5-35B-A3B/hle-verified-provenance.json +400 -0
- Qwen3.5-35B-A3B/summary.json +514 -1
- Qwen3.5-9B/README.md +14 -0
- Qwen3.5-9B/hle-verified-provenance.json +395 -0
- Qwen3.5-9B/summary.json +193 -1
- README.md +14 -2
- gpt-oss-20b/README.md +14 -0
- gpt-oss-20b/hle-verified-provenance.json +406 -0
- gpt-oss-20b/summary.json +467 -1
- hle-verified-20261030/README.md +0 -29
- hle-verified-20261030/manifest.json +0 -1
- hle-verified-20261030/provenance.json +0 -1
- hle-verified-20261030/source-manifest.json +0 -1
- hle-verified-20261030/summary.json +0 -1
- hle-verified-20261030/sweep.json +0 -1
- hle-verified-20261030/token-preflight.json +0 -1
- sft-agentic-stage2/README.md +14 -0
- sft-agentic-stage2/hle-verified-provenance.json +394 -0
- sft-agentic-stage2/summary.json +472 -1
Gemma-4-12B-it/README.md
CHANGED
|
@@ -20,3 +20,17 @@ Reasoning output cap: 98,304 tokens. FrontierScience judge: GPT-5.6 Luna. ArXivM
|
|
| 20 |
- Earlier infrastructure-failed retries are documented in `recovery_provenance.tar.gz`; originals and replacement attempts are preserved. They do not add attempts to the denominator. Native zero-score attempts were not selectively rerun.
|
| 21 |
|
| 22 |
Every benchmark has trajectories, aligned judge/verifier records, summary, and compressed native/event artifacts. Model-level manifest gives SHA-256 and size. Publication made no new model/judge calls and did not change scores.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
- Earlier infrastructure-failed retries are documented in `recovery_provenance.tar.gz`; originals and replacement attempts are preserved. They do not add attempts to the denominator. Native zero-score attempts were not selectively rerun.
|
| 21 |
|
| 22 |
Every benchmark has trajectories, aligned judge/verifier records, summary, and compressed native/event artifacts. Model-level manifest gives SHA-256 and size. Publication made no new model/judge calls and did not change scores.
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
## HLE-Verified (2026-10-30)
|
| 26 |
+
|
| 27 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **466/1,600 (29.12%)**.
|
| 28 |
+
|
| 29 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 30 |
+
|---|---:|---:|---|
|
| 31 |
+
| Gold | 159/575 | 27.65% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 32 |
+
| Revision | 307/1025 | 29.95% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 33 |
+
|
| 34 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 98,304 tokens; context 131,072 tokens. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 35 |
+
|
| 36 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
Gemma-4-12B-it/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,384 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "gemma4-12b-it",
|
| 7 |
+
"folder": "Gemma-4-12B-it",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 159,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.27652173913043476,
|
| 13 |
+
"cap_hits": 3
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 307,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.2995121951219512,
|
| 19 |
+
"cap_hits": 10
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 466,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.29125
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "gemma4-12b-it",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "gemma4-12b",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "gemma",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "gemma4_unified_text",
|
| 35 |
+
"hidden_size": 3840,
|
| 36 |
+
"num_hidden_layers": 48,
|
| 37 |
+
"vocab_size": 262144
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 98304,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"top_k": -1,
|
| 46 |
+
"presence_penalty": 0.0,
|
| 47 |
+
"chat_template_kwargs": {
|
| 48 |
+
"enable_thinking": true
|
| 49 |
+
}
|
| 50 |
+
}
|
| 51 |
+
},
|
| 52 |
+
"serving": {
|
| 53 |
+
"context_length": 131072,
|
| 54 |
+
"dtype": "bfloat16",
|
| 55 |
+
"attention_backend": "triton",
|
| 56 |
+
"mem_fraction_static": 0.85,
|
| 57 |
+
"reasoning_parser": "gemma4",
|
| 58 |
+
"external_model_package": "gemma4_compat",
|
| 59 |
+
"extra_args": [
|
| 60 |
+
"--chunked-prefill-size",
|
| 61 |
+
"4096"
|
| 62 |
+
],
|
| 63 |
+
"tp": 2,
|
| 64 |
+
"replicas_per_node": 4,
|
| 65 |
+
"max_running_requests": 32
|
| 66 |
+
},
|
| 67 |
+
"reasoning": {
|
| 68 |
+
"requests_per_engine": 32,
|
| 69 |
+
"benchmarks": {}
|
| 70 |
+
},
|
| 71 |
+
"agentic_tasks_per_engine": 1
|
| 72 |
+
},
|
| 73 |
+
"node": 2,
|
| 74 |
+
"checkpoint": "/workspace/data/raw_model/gemma-4-12B-it-707f0a3b8a3c",
|
| 75 |
+
"repo": "google/gemma-4-12B-it",
|
| 76 |
+
"revision": "707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7",
|
| 77 |
+
"checkpoint_config_sha256": "478c46e8d2c52d5c2d85bf67e3b3e8c90e7c9d91086cee27e3c267907e936bd9",
|
| 78 |
+
"index_sha256": null,
|
| 79 |
+
"shard_sizes": {
|
| 80 |
+
"model.safetensors": 23919549408
|
| 81 |
+
},
|
| 82 |
+
"endpoints": [
|
| 83 |
+
{
|
| 84 |
+
"id": "node-2-replica-0",
|
| 85 |
+
"api_base": "http://10.4.70.214:8700/v1"
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"id": "node-2-replica-1",
|
| 89 |
+
"api_base": "http://10.4.70.214:8701/v1"
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"id": "node-2-replica-2",
|
| 93 |
+
"api_base": "http://10.4.70.214:8702/v1"
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"id": "node-2-replica-3",
|
| 97 |
+
"api_base": "http://10.4.70.214:8703/v1"
|
| 98 |
+
}
|
| 99 |
+
],
|
| 100 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gemma4-12b-it",
|
| 101 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gemma4-12b-it/configs/hle-verified-gold-revision-96k.yaml",
|
| 102 |
+
"concurrency": 128,
|
| 103 |
+
"judge_concurrency": 64,
|
| 104 |
+
"tracking": {
|
| 105 |
+
"enabled": false,
|
| 106 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 107 |
+
},
|
| 108 |
+
"config_sha256": {
|
| 109 |
+
"model.yaml": "ef319aecd370202f03b6bf612bfb2507a92d17be361615b0abe65b7ba9222beb",
|
| 110 |
+
"configs/hle-verified-gold-revision-96k.yaml": "42ffa0f22d18b4e04a4321d5904208c87583ffccca5de393b690ccc29893dab3",
|
| 111 |
+
"configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4",
|
| 112 |
+
"configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"
|
| 113 |
+
}
|
| 114 |
+
},
|
| 115 |
+
"dataset": {
|
| 116 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 117 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 118 |
+
"counts": {
|
| 119 |
+
"Gold": {
|
| 120 |
+
"total": 668,
|
| 121 |
+
"text": 575,
|
| 122 |
+
"image_excluded": 93
|
| 123 |
+
},
|
| 124 |
+
"Revision": {
|
| 125 |
+
"total": 1143,
|
| 126 |
+
"text": 1025,
|
| 127 |
+
"image_excluded": 118
|
| 128 |
+
}
|
| 129 |
+
},
|
| 130 |
+
"sha256": {
|
| 131 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 132 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 133 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 134 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 135 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 136 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 137 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 138 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 139 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 140 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 141 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 142 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 143 |
+
}
|
| 144 |
+
},
|
| 145 |
+
"token_preflight": {
|
| 146 |
+
"model": "gemma4-12b-it",
|
| 147 |
+
"questions": 1600,
|
| 148 |
+
"max_prompt_tokens": 10151,
|
| 149 |
+
"output_cap": 98304,
|
| 150 |
+
"context": 131072,
|
| 151 |
+
"dynamic_output_budget": false,
|
| 152 |
+
"affected_questions": 0,
|
| 153 |
+
"min_effective_output_cap": 98304
|
| 154 |
+
},
|
| 155 |
+
"shared_recovery_provenance": {
|
| 156 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 157 |
+
"publication_reruns": false,
|
| 158 |
+
"all_questions_scored": true,
|
| 159 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 160 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 161 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 162 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 163 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 164 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 165 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 166 |
+
},
|
| 167 |
+
"source_manifest": {
|
| 168 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 169 |
+
"files": {
|
| 170 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 171 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 172 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 173 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 174 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 175 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 176 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 177 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 178 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 179 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 180 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 181 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 182 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 183 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 184 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 185 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 186 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 187 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 188 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 189 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 190 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 191 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 192 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 193 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 194 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 195 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 196 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 197 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 198 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 199 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 200 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 201 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 202 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 203 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 204 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 205 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 206 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 207 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 208 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 209 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 210 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 211 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 212 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 213 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 214 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 215 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 216 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 217 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 218 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 219 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 220 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 221 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 222 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 223 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 224 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 225 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 226 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 227 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 228 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 229 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 230 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 231 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 232 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 233 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 234 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 235 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 236 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 237 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 238 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 239 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 240 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 241 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 242 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 243 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 244 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 245 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 246 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 247 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 248 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 249 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 250 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 251 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 252 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 253 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 254 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 255 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 256 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 257 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 258 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 259 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 260 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 261 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 262 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 263 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 264 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 265 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 266 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 267 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 268 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 269 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 270 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 271 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 272 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 273 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 274 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 275 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 276 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 277 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 278 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 279 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 280 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 281 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 282 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 283 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 284 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 285 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 286 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 287 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 288 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 289 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 290 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 291 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 292 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 293 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 294 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 295 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 296 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 297 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 298 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 299 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 300 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 301 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 302 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 303 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 304 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 305 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 306 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 307 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 308 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 309 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 310 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 311 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 312 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 313 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 314 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 315 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 316 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 317 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 318 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 319 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 320 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 321 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 322 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 323 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 324 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 325 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 326 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 327 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 328 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 329 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 330 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 331 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 332 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 333 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 334 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 335 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 336 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 337 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 338 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 339 |
+
},
|
| 340 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 341 |
+
},
|
| 342 |
+
"artifact_manifest": {
|
| 343 |
+
"Gemma-4-12B-it/hle-verified-gold/configs.json": {
|
| 344 |
+
"size": 95876,
|
| 345 |
+
"sha256": "25a009e6c6226da7d37495e9a923c6519228bd06c0abef92f8185fffd54d97c8"
|
| 346 |
+
},
|
| 347 |
+
"Gemma-4-12B-it/hle-verified-gold/events.jsonl.gz": {
|
| 348 |
+
"size": 12394882,
|
| 349 |
+
"sha256": "372211276b94cd99f2d455bda1aee8ca8f96f9461143b850d3f582bf05c47e74"
|
| 350 |
+
},
|
| 351 |
+
"Gemma-4-12B-it/hle-verified-gold/judge_trajectories.jsonl": {
|
| 352 |
+
"size": 808737,
|
| 353 |
+
"sha256": "8deb1da2f9285ac2e39b151e074785844f16f1da4bc64788032ad9e932fda6d7"
|
| 354 |
+
},
|
| 355 |
+
"Gemma-4-12B-it/hle-verified-gold/summary.json": {
|
| 356 |
+
"size": 1495,
|
| 357 |
+
"sha256": "6e5cffccacef2f8eb03cbcf45da90f97468cefb6153e8fdede1d217a32341e0e"
|
| 358 |
+
},
|
| 359 |
+
"Gemma-4-12B-it/hle-verified-gold/trajectories.jsonl": {
|
| 360 |
+
"size": 30459259,
|
| 361 |
+
"sha256": "3b943c4da59d980a98c8e027c7b4a5c46893825ba6e1c6facfb0f0dcb5dfc2aa"
|
| 362 |
+
},
|
| 363 |
+
"Gemma-4-12B-it/hle-verified-revision/configs.json": {
|
| 364 |
+
"size": 164308,
|
| 365 |
+
"sha256": "35c646cf84d89f08772cd52ea897a63040166e7e8355965f007f3af05a5bc483"
|
| 366 |
+
},
|
| 367 |
+
"Gemma-4-12B-it/hle-verified-revision/events.jsonl.gz": {
|
| 368 |
+
"size": 24343432,
|
| 369 |
+
"sha256": "f25a4183fe986f31821b984244964c866449fb5280d91ffb9d0673684a6eaab0"
|
| 370 |
+
},
|
| 371 |
+
"Gemma-4-12B-it/hle-verified-revision/judge_trajectories.jsonl": {
|
| 372 |
+
"size": 1461032,
|
| 373 |
+
"sha256": "2bd13649258614bf6b9ae852dedec6de9affb24a70d68b0f748b1fb5003e5844"
|
| 374 |
+
},
|
| 375 |
+
"Gemma-4-12B-it/hle-verified-revision/summary.json": {
|
| 376 |
+
"size": 1516,
|
| 377 |
+
"sha256": "77a100d4185ae9069742c37fd316297355723476fd44ced24b535dd6ad8f1960"
|
| 378 |
+
},
|
| 379 |
+
"Gemma-4-12B-it/hle-verified-revision/trajectories.jsonl": {
|
| 380 |
+
"size": 57776284,
|
| 381 |
+
"sha256": "943f338e784a0e4677818a0148748b65e64c059cf7ca35b6730abd1171f7c840"
|
| 382 |
+
}
|
| 383 |
+
}
|
| 384 |
+
}
|
Gemma-4-12B-it/summary.json
CHANGED
|
@@ -1 +1,524 @@
|
|
| 1 |
-
{"frontierscience-olympiad": {"total_records": 3000, "completed": 3000, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.439, "accuracy": 0.439, "mean_latency_seconds": 318.5555908056985, "finish_reason_counts": {"length": 13, "stop": 2987}, "length_limited_response_count": 13, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 3000, "request_count": 3000, "retry_request_count": 0, "prompt_tokens": 5792582, "cached_prompt_tokens": 65971, "cache_write_prompt_tokens": 5429239, "completion_tokens": 774881, "reasoning_tokens": 500420, "total_tokens": 6567463}, "judge_api_usage": {"request_count": 3056, "prompt_tokens": 5884031, "cached_prompt_tokens": 65971, "cache_write_prompt_tokens": 5520520, "completion_tokens": 784014, "reasoning_tokens": 504141, "total_tokens": 6668045}, "benchmark_metrics": {"track": "olympiad", "n_problems_scored": 100, "n_trajectories_scored": 3000, "aggregation": "avg@30", "reported_accuracy": 0.439, "mean_rubric_score": null, "success_threshold": 0.5, "community_profile": null}, "benchmark": "frontierscience-olympiad", "model": "gemma4-12b-thinking-default-20261028", "execution_mode": "non_agent", "generation_hash": "f01fff53861607b6cc1f4c64b16e92d801cb2f9c56119f567df6947449d905b8", "scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016", "selected": 3000, "incomplete": 0, "generated_now": 2944, "scored_now": 2944, "cached_completed": 56, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 3000, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-28T20:30:53.226445+00:00", "provisional": false, "grading_complete": true, "source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run", "publication_date": "2026-10-29"}, "frontierscience-research": {"total_records": 1800, "completed": 1800, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.23008833333333334, "accuracy": 0.02388888888888889, "mean_latency_seconds": 130.2966016948043, "finish_reason_counts": {"length": 1, "stop": 1799}, "length_limited_response_count": 1, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 1800, "request_count": 1825, "retry_request_count": 25, "prompt_tokens": 6184390, "cached_prompt_tokens": 360921, "cache_write_prompt_tokens": 5817333, "completion_tokens": 2486028, "reasoning_tokens": 1375147, "total_tokens": 8670418}, "judge_api_usage": {"request_count": 1856, "prompt_tokens": 6427080, "cached_prompt_tokens": 360921, "cache_write_prompt_tokens": 6059930, "completion_tokens": 2566034, "reasoning_tokens": 1418153, "total_tokens": 8993114}, "benchmark_metrics": {"track": "research", "n_problems_scored": 60, "n_trajectories_scored": 1800, "aggregation": "avg@30", "reported_accuracy": 0.02388888888888889, "mean_rubric_score": 0.23008833333333334, "success_threshold": 0.7, "community_profile": null}, "benchmark": "frontierscience-research", "model": "gemma4-12b-thinking-default-20261028", "execution_mode": "non_agent", "generation_hash": "b8c3e91ede078b65e6e97c1c7b4f910a42136f9ed83eb6c15cf16644b962c094", "scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750", "selected": 1800, "incomplete": 0, "generated_now": 1591, "scored_now": 1604, "cached_completed": 196, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 1800, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-28T20:08:35.019803+00:00", "provisional": false, "grading_complete": true, "source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run", "publication_date": "2026-10-29"}, "arxivmath": {"benchmark": "arxivmath", "releases": [{"total_records": 68, "completed": 68, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.3382352941176471, "accuracy": 0.3382352941176471, "mean_latency_seconds": 367.5543368255894, "finish_reason_counts": {"stop": 68}, "length_limited_response_count": 0, "observed_max_completion_tokens": 50699, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.3382352941176471, "competition_metrics": {"arxiv/december": {"n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.3382352941176471}}}, "benchmark": "matharena-arxivmath-december", "model": "gemma4-12b-thinking-default-20261028", "execution_mode": "non_agent", "generation_hash": "51a354515c17fdc7b137943ef8ca3a43e130b1b33f830556f22c75ced8245521", "scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661", "selected": 68, "incomplete": 0, "generated_now": 61, "scored_now": 61, "cached_completed": 7, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 68, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-28T19:39:02.350344+00:00"}, {"total_records": 128, "completed": 128, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.3046875, "accuracy": 0.3046875, "mean_latency_seconds": 346.9827236642159, "finish_reason_counts": {"stop": 128}, "length_limited_response_count": 0, "observed_max_completion_tokens": 38984, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.3046875, "competition_metrics": {"arxiv/february": {"n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.3046875}}}, "benchmark": "matharena-arxivmath-february", "model": "gemma4-12b-thinking-default-20261028", "execution_mode": "non_agent", "generation_hash": "ada70051681516a4d657df07247e3cc3627b2ae9668704bde2017ba9868b1b2c", "scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5", "selected": 128, "incomplete": 0, "generated_now": 125, "scored_now": 125, "cached_completed": 3, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 128, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-28T19:39:35.480294+00:00"}, {"total_records": 92, "completed": 92, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.391304347826087, "accuracy": 0.391304347826087, "mean_latency_seconds": 384.78772876479246, "finish_reason_counts": {"stop": 92}, "length_limited_response_count": 0, "observed_max_completion_tokens": 47044, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.391304347826087, "competition_metrics": {"arxiv/january": {"n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.391304347826087}}}, "benchmark": "matharena-arxivmath-january", "model": "gemma4-12b-thinking-default-20261028", "execution_mode": "non_agent", "generation_hash": "8444f820a02c4a0f2522b7788d6a7b4b51411062b5283c56b97ae486e94388cd", "scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8", "selected": 92, "incomplete": 0, "generated_now": 92, "scored_now": 92, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 92, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-28T19:40:31.309145+00:00"}, {"total_records": 124, "completed": 124, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.25806451612903225, "accuracy": 0.25806451612903225, "mean_latency_seconds": 327.9356012846641, "finish_reason_counts": {"stop": 124}, "length_limited_response_count": 0, "observed_max_completion_tokens": 41506, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.25806451612903225, "competition_metrics": {"arxiv/march": {"n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.25806451612903225}}}, "benchmark": "matharena-arxivmath-march", "model": "gemma4-12b-thinking-default-20261028", "execution_mode": "non_agent", "generation_hash": "3981d7dfa0cc9c30ffa6838accc875e14d491450732c85562b4885c7427aaa7f", "scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8", "selected": 124, "incomplete": 0, "generated_now": 119, "scored_now": 119, "cached_completed": 5, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 124, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-28T19:37:10.757171+00:00"}], "total_records": 412, "completed": 412, "accuracy": 0.3155339805825243, "provisional": false, "grading_complete": true, "source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run", "publication_date": "2026-10-29"}, "scicode-verified": {"updated_at": 1793218906.0395374, "expected_problems": 64, "expected_steps": 287, "generated_steps": 287, "scored_steps": 287, "passed_steps": 189, "completed_problems": 64, "passed_problems": 31, "subproblem_accuracy": 0.6585365853658537, "main_problem_accuracy": 0.484375, "complete": true, "cap_hits": 0, "mean_output_tokens": 6563.003484320558, "grading_infrastructure_errors": 0, "errors": [], "provisional": false, "grading_complete": true, "source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run", "publication_date": "2026-10-29"}, "scicoding-agent": {"tasks": 50, "attempts": 100, "attempts_per_task": 2, "metric": "avg@2", "passed": 30, "avg_at_2": 0.3, "unscored": 0, "agent_timeouts": 0, "reward_positive_with_exception": 0, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1", "provisional": false, "grading_complete": true, "source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run", "publication_date": "2026-10-29"}, "terminal-bench-2.1": {"tasks": 89, "attempts": 178, "attempts_per_task": 2, "metric": "avg@2", "passed": 41, "avg_at_2": 0.2303370786516854, "unscored": 0, "agent_timeouts": 73, "reward_positive_with_exception": 0, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1", "provisional": false, "grading_complete": true, "source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run", "publication_date": "2026-10-29", "scoring_adjudications": {"terminal-bench-2.1--pytorch-model-recovery--attempt-002": {"timestamp": "2026-10-29T02:54:23.209570+00:00", "reward": 0.0, "method": "user_authorized_zero_for_verifier_setup_timeout", "authorization": "[REDACTED_CREDENTIAL]", "reason": "Agent timed out; subsequent verifier exhausted its budget downloading dependencies before recorded test execution. Zero is a user adjudication, not a native test verdict.", "native_reward": null, "original_verifier_result": null, "native_tests_executed": "not evidenced", "generation_or_verifier_rerun": false, "backup": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/archive/user-zero-adjudication-20261029/terminal-bench-2.1--pytorch-model-recovery--attempt-002", "evidence_sha256": {"result.json": "39fcbacd95efc60c6206c392dc10b1042e360d52ab01dfc9da3f3b5db7c01aeb", "trial.log": "f4079b102eb2e030b59db97150da9e79e62888739af519694df2443f9fb7b01d", "verifier/test-stdout.txt": "7327b6f99e09223054e707efa9e9e25dec6b37e7f9e6f2ea4ccf1ff14e0b7edf"}}, "terminal-bench-2.1--torch-pipeline-parallelism--attempt-002": {"timestamp": "2026-10-29T02:50:58.603775+00:00", "reward": 0, "method": "adopt_existing_native_verifier_artifacts", "authorization": "[REDACTED_CREDENTIAL]", "reason": "Native reward.txt is zero; CTRF and test stdout evidence failed correctness tests. Harbor timed out before recording the native reward.", "original_verifier_result": null, "original_exception_type": "VerifierTimeoutError", "evidence": {"verifier/reward.txt": "9a271f2a916b0b6ee6cecb2426f0b3206ef074578be55d9bc94f6f3fe3ab86aa", "verifier/ctrf.json": "806ed353196f76e1e04e9bb9c306cf062098f0c3d5d8c5993693e18ac663fdfd", "verifier/test-stdout.txt": "08a57104364c6fb08ae9ab7a1a4f7a42759c705c37a9dc1c3fe01f693685a848", "result.json": "8b76029fb1a0a222a12fb83a80ada203a18598a3793d840ed93c4d59f603c655"}, "backup": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/archive/native-zero-adoption-20261029/terminal-bench-2.1--torch-pipeline-parallelism--attempt-002", "generation_or_verifier_rerun": false}}, "scoring_note": "One explicit user-authorized zero after verifier setup timeout; one adopted native zero. Original exceptions, null original rewards and backups retained. Not all scores are completed native test verdicts."}}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"frontierscience-olympiad": {
|
| 3 |
+
"total_records": 3000,
|
| 4 |
+
"completed": 3000,
|
| 5 |
+
"query_errors": 0,
|
| 6 |
+
"score_errors": 0,
|
| 7 |
+
"generated_unscored": 0,
|
| 8 |
+
"mean_score": 0.439,
|
| 9 |
+
"accuracy": 0.439,
|
| 10 |
+
"mean_latency_seconds": 318.5555908056985,
|
| 11 |
+
"finish_reason_counts": {
|
| 12 |
+
"length": 13,
|
| 13 |
+
"stop": 2987
|
| 14 |
+
},
|
| 15 |
+
"length_limited_response_count": 13,
|
| 16 |
+
"observed_max_completion_tokens": 98304,
|
| 17 |
+
"judge_usage": {
|
| 18 |
+
"judged_record_count": 3000,
|
| 19 |
+
"request_count": 3000,
|
| 20 |
+
"retry_request_count": 0,
|
| 21 |
+
"prompt_tokens": 5792582,
|
| 22 |
+
"cached_prompt_tokens": 65971,
|
| 23 |
+
"cache_write_prompt_tokens": 5429239,
|
| 24 |
+
"completion_tokens": 774881,
|
| 25 |
+
"reasoning_tokens": 500420,
|
| 26 |
+
"total_tokens": 6567463
|
| 27 |
+
},
|
| 28 |
+
"judge_api_usage": {
|
| 29 |
+
"request_count": 3056,
|
| 30 |
+
"prompt_tokens": 5884031,
|
| 31 |
+
"cached_prompt_tokens": 65971,
|
| 32 |
+
"cache_write_prompt_tokens": 5520520,
|
| 33 |
+
"completion_tokens": 784014,
|
| 34 |
+
"reasoning_tokens": 504141,
|
| 35 |
+
"total_tokens": 6668045
|
| 36 |
+
},
|
| 37 |
+
"benchmark_metrics": {
|
| 38 |
+
"track": "olympiad",
|
| 39 |
+
"n_problems_scored": 100,
|
| 40 |
+
"n_trajectories_scored": 3000,
|
| 41 |
+
"aggregation": "avg@30",
|
| 42 |
+
"reported_accuracy": 0.439,
|
| 43 |
+
"mean_rubric_score": null,
|
| 44 |
+
"success_threshold": 0.5,
|
| 45 |
+
"community_profile": null
|
| 46 |
+
},
|
| 47 |
+
"benchmark": "frontierscience-olympiad",
|
| 48 |
+
"model": "gemma4-12b-thinking-default-20261028",
|
| 49 |
+
"execution_mode": "non_agent",
|
| 50 |
+
"generation_hash": "f01fff53861607b6cc1f4c64b16e92d801cb2f9c56119f567df6947449d905b8",
|
| 51 |
+
"scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016",
|
| 52 |
+
"selected": 3000,
|
| 53 |
+
"incomplete": 0,
|
| 54 |
+
"generated_now": 2944,
|
| 55 |
+
"scored_now": 2944,
|
| 56 |
+
"cached_completed": 56,
|
| 57 |
+
"stage": "all",
|
| 58 |
+
"worker_count": 1,
|
| 59 |
+
"worker_index": 0,
|
| 60 |
+
"worker_owned_predictions": 3000,
|
| 61 |
+
"generation_concurrency": 512,
|
| 62 |
+
"judge_concurrency": 256,
|
| 63 |
+
"updated_at": "2026-10-28T20:30:53.226445+00:00",
|
| 64 |
+
"provisional": false,
|
| 65 |
+
"grading_complete": true,
|
| 66 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run",
|
| 67 |
+
"publication_date": "2026-10-29"
|
| 68 |
+
},
|
| 69 |
+
"frontierscience-research": {
|
| 70 |
+
"total_records": 1800,
|
| 71 |
+
"completed": 1800,
|
| 72 |
+
"query_errors": 0,
|
| 73 |
+
"score_errors": 0,
|
| 74 |
+
"generated_unscored": 0,
|
| 75 |
+
"mean_score": 0.23008833333333334,
|
| 76 |
+
"accuracy": 0.02388888888888889,
|
| 77 |
+
"mean_latency_seconds": 130.2966016948043,
|
| 78 |
+
"finish_reason_counts": {
|
| 79 |
+
"length": 1,
|
| 80 |
+
"stop": 1799
|
| 81 |
+
},
|
| 82 |
+
"length_limited_response_count": 1,
|
| 83 |
+
"observed_max_completion_tokens": 98304,
|
| 84 |
+
"judge_usage": {
|
| 85 |
+
"judged_record_count": 1800,
|
| 86 |
+
"request_count": 1825,
|
| 87 |
+
"retry_request_count": 25,
|
| 88 |
+
"prompt_tokens": 6184390,
|
| 89 |
+
"cached_prompt_tokens": 360921,
|
| 90 |
+
"cache_write_prompt_tokens": 5817333,
|
| 91 |
+
"completion_tokens": 2486028,
|
| 92 |
+
"reasoning_tokens": 1375147,
|
| 93 |
+
"total_tokens": 8670418
|
| 94 |
+
},
|
| 95 |
+
"judge_api_usage": {
|
| 96 |
+
"request_count": 1856,
|
| 97 |
+
"prompt_tokens": 6427080,
|
| 98 |
+
"cached_prompt_tokens": 360921,
|
| 99 |
+
"cache_write_prompt_tokens": 6059930,
|
| 100 |
+
"completion_tokens": 2566034,
|
| 101 |
+
"reasoning_tokens": 1418153,
|
| 102 |
+
"total_tokens": 8993114
|
| 103 |
+
},
|
| 104 |
+
"benchmark_metrics": {
|
| 105 |
+
"track": "research",
|
| 106 |
+
"n_problems_scored": 60,
|
| 107 |
+
"n_trajectories_scored": 1800,
|
| 108 |
+
"aggregation": "avg@30",
|
| 109 |
+
"reported_accuracy": 0.02388888888888889,
|
| 110 |
+
"mean_rubric_score": 0.23008833333333334,
|
| 111 |
+
"success_threshold": 0.7,
|
| 112 |
+
"community_profile": null
|
| 113 |
+
},
|
| 114 |
+
"benchmark": "frontierscience-research",
|
| 115 |
+
"model": "gemma4-12b-thinking-default-20261028",
|
| 116 |
+
"execution_mode": "non_agent",
|
| 117 |
+
"generation_hash": "b8c3e91ede078b65e6e97c1c7b4f910a42136f9ed83eb6c15cf16644b962c094",
|
| 118 |
+
"scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750",
|
| 119 |
+
"selected": 1800,
|
| 120 |
+
"incomplete": 0,
|
| 121 |
+
"generated_now": 1591,
|
| 122 |
+
"scored_now": 1604,
|
| 123 |
+
"cached_completed": 196,
|
| 124 |
+
"stage": "all",
|
| 125 |
+
"worker_count": 1,
|
| 126 |
+
"worker_index": 0,
|
| 127 |
+
"worker_owned_predictions": 1800,
|
| 128 |
+
"generation_concurrency": 512,
|
| 129 |
+
"judge_concurrency": 256,
|
| 130 |
+
"updated_at": "2026-10-28T20:08:35.019803+00:00",
|
| 131 |
+
"provisional": false,
|
| 132 |
+
"grading_complete": true,
|
| 133 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run",
|
| 134 |
+
"publication_date": "2026-10-29"
|
| 135 |
+
},
|
| 136 |
+
"arxivmath": {
|
| 137 |
+
"benchmark": "arxivmath",
|
| 138 |
+
"releases": [
|
| 139 |
+
{
|
| 140 |
+
"total_records": 68,
|
| 141 |
+
"completed": 68,
|
| 142 |
+
"query_errors": 0,
|
| 143 |
+
"score_errors": 0,
|
| 144 |
+
"generated_unscored": 0,
|
| 145 |
+
"mean_score": 0.3382352941176471,
|
| 146 |
+
"accuracy": 0.3382352941176471,
|
| 147 |
+
"mean_latency_seconds": 367.5543368255894,
|
| 148 |
+
"finish_reason_counts": {
|
| 149 |
+
"stop": 68
|
| 150 |
+
},
|
| 151 |
+
"length_limited_response_count": 0,
|
| 152 |
+
"observed_max_completion_tokens": 50699,
|
| 153 |
+
"judge_usage": {
|
| 154 |
+
"judged_record_count": 0,
|
| 155 |
+
"request_count": 0,
|
| 156 |
+
"retry_request_count": 0,
|
| 157 |
+
"prompt_tokens": 0,
|
| 158 |
+
"cached_prompt_tokens": 0,
|
| 159 |
+
"cache_write_prompt_tokens": 0,
|
| 160 |
+
"completion_tokens": 0,
|
| 161 |
+
"reasoning_tokens": 0,
|
| 162 |
+
"total_tokens": 0
|
| 163 |
+
},
|
| 164 |
+
"judge_api_usage": {
|
| 165 |
+
"request_count": 0,
|
| 166 |
+
"prompt_tokens": 0,
|
| 167 |
+
"cached_prompt_tokens": 0,
|
| 168 |
+
"cache_write_prompt_tokens": 0,
|
| 169 |
+
"completion_tokens": 0,
|
| 170 |
+
"reasoning_tokens": 0,
|
| 171 |
+
"total_tokens": 0
|
| 172 |
+
},
|
| 173 |
+
"benchmark_metrics": {
|
| 174 |
+
"competition": "arxivmath",
|
| 175 |
+
"aggregation": "avg@4",
|
| 176 |
+
"n_problems_scored": 17,
|
| 177 |
+
"n_trajectories_scored": 68,
|
| 178 |
+
"reported_accuracy": 0.3382352941176471,
|
| 179 |
+
"competition_metrics": {
|
| 180 |
+
"arxiv/december": {
|
| 181 |
+
"n_problems_scored": 17,
|
| 182 |
+
"n_trajectories_scored": 68,
|
| 183 |
+
"reported_accuracy": 0.3382352941176471
|
| 184 |
+
}
|
| 185 |
+
}
|
| 186 |
+
},
|
| 187 |
+
"benchmark": "matharena-arxivmath-december",
|
| 188 |
+
"model": "gemma4-12b-thinking-default-20261028",
|
| 189 |
+
"execution_mode": "non_agent",
|
| 190 |
+
"generation_hash": "51a354515c17fdc7b137943ef8ca3a43e130b1b33f830556f22c75ced8245521",
|
| 191 |
+
"scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661",
|
| 192 |
+
"selected": 68,
|
| 193 |
+
"incomplete": 0,
|
| 194 |
+
"generated_now": 61,
|
| 195 |
+
"scored_now": 61,
|
| 196 |
+
"cached_completed": 7,
|
| 197 |
+
"stage": "all",
|
| 198 |
+
"worker_count": 1,
|
| 199 |
+
"worker_index": 0,
|
| 200 |
+
"worker_owned_predictions": 68,
|
| 201 |
+
"generation_concurrency": 512,
|
| 202 |
+
"judge_concurrency": 256,
|
| 203 |
+
"updated_at": "2026-10-28T19:39:02.350344+00:00"
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"total_records": 128,
|
| 207 |
+
"completed": 128,
|
| 208 |
+
"query_errors": 0,
|
| 209 |
+
"score_errors": 0,
|
| 210 |
+
"generated_unscored": 0,
|
| 211 |
+
"mean_score": 0.3046875,
|
| 212 |
+
"accuracy": 0.3046875,
|
| 213 |
+
"mean_latency_seconds": 346.9827236642159,
|
| 214 |
+
"finish_reason_counts": {
|
| 215 |
+
"stop": 128
|
| 216 |
+
},
|
| 217 |
+
"length_limited_response_count": 0,
|
| 218 |
+
"observed_max_completion_tokens": 38984,
|
| 219 |
+
"judge_usage": {
|
| 220 |
+
"judged_record_count": 0,
|
| 221 |
+
"request_count": 0,
|
| 222 |
+
"retry_request_count": 0,
|
| 223 |
+
"prompt_tokens": 0,
|
| 224 |
+
"cached_prompt_tokens": 0,
|
| 225 |
+
"cache_write_prompt_tokens": 0,
|
| 226 |
+
"completion_tokens": 0,
|
| 227 |
+
"reasoning_tokens": 0,
|
| 228 |
+
"total_tokens": 0
|
| 229 |
+
},
|
| 230 |
+
"judge_api_usage": {
|
| 231 |
+
"request_count": 0,
|
| 232 |
+
"prompt_tokens": 0,
|
| 233 |
+
"cached_prompt_tokens": 0,
|
| 234 |
+
"cache_write_prompt_tokens": 0,
|
| 235 |
+
"completion_tokens": 0,
|
| 236 |
+
"reasoning_tokens": 0,
|
| 237 |
+
"total_tokens": 0
|
| 238 |
+
},
|
| 239 |
+
"benchmark_metrics": {
|
| 240 |
+
"competition": "arxivmath",
|
| 241 |
+
"aggregation": "avg@4",
|
| 242 |
+
"n_problems_scored": 32,
|
| 243 |
+
"n_trajectories_scored": 128,
|
| 244 |
+
"reported_accuracy": 0.3046875,
|
| 245 |
+
"competition_metrics": {
|
| 246 |
+
"arxiv/february": {
|
| 247 |
+
"n_problems_scored": 32,
|
| 248 |
+
"n_trajectories_scored": 128,
|
| 249 |
+
"reported_accuracy": 0.3046875
|
| 250 |
+
}
|
| 251 |
+
}
|
| 252 |
+
},
|
| 253 |
+
"benchmark": "matharena-arxivmath-february",
|
| 254 |
+
"model": "gemma4-12b-thinking-default-20261028",
|
| 255 |
+
"execution_mode": "non_agent",
|
| 256 |
+
"generation_hash": "ada70051681516a4d657df07247e3cc3627b2ae9668704bde2017ba9868b1b2c",
|
| 257 |
+
"scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5",
|
| 258 |
+
"selected": 128,
|
| 259 |
+
"incomplete": 0,
|
| 260 |
+
"generated_now": 125,
|
| 261 |
+
"scored_now": 125,
|
| 262 |
+
"cached_completed": 3,
|
| 263 |
+
"stage": "all",
|
| 264 |
+
"worker_count": 1,
|
| 265 |
+
"worker_index": 0,
|
| 266 |
+
"worker_owned_predictions": 128,
|
| 267 |
+
"generation_concurrency": 512,
|
| 268 |
+
"judge_concurrency": 256,
|
| 269 |
+
"updated_at": "2026-10-28T19:39:35.480294+00:00"
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"total_records": 92,
|
| 273 |
+
"completed": 92,
|
| 274 |
+
"query_errors": 0,
|
| 275 |
+
"score_errors": 0,
|
| 276 |
+
"generated_unscored": 0,
|
| 277 |
+
"mean_score": 0.391304347826087,
|
| 278 |
+
"accuracy": 0.391304347826087,
|
| 279 |
+
"mean_latency_seconds": 384.78772876479246,
|
| 280 |
+
"finish_reason_counts": {
|
| 281 |
+
"stop": 92
|
| 282 |
+
},
|
| 283 |
+
"length_limited_response_count": 0,
|
| 284 |
+
"observed_max_completion_tokens": 47044,
|
| 285 |
+
"judge_usage": {
|
| 286 |
+
"judged_record_count": 0,
|
| 287 |
+
"request_count": 0,
|
| 288 |
+
"retry_request_count": 0,
|
| 289 |
+
"prompt_tokens": 0,
|
| 290 |
+
"cached_prompt_tokens": 0,
|
| 291 |
+
"cache_write_prompt_tokens": 0,
|
| 292 |
+
"completion_tokens": 0,
|
| 293 |
+
"reasoning_tokens": 0,
|
| 294 |
+
"total_tokens": 0
|
| 295 |
+
},
|
| 296 |
+
"judge_api_usage": {
|
| 297 |
+
"request_count": 0,
|
| 298 |
+
"prompt_tokens": 0,
|
| 299 |
+
"cached_prompt_tokens": 0,
|
| 300 |
+
"cache_write_prompt_tokens": 0,
|
| 301 |
+
"completion_tokens": 0,
|
| 302 |
+
"reasoning_tokens": 0,
|
| 303 |
+
"total_tokens": 0
|
| 304 |
+
},
|
| 305 |
+
"benchmark_metrics": {
|
| 306 |
+
"competition": "arxivmath",
|
| 307 |
+
"aggregation": "avg@4",
|
| 308 |
+
"n_problems_scored": 23,
|
| 309 |
+
"n_trajectories_scored": 92,
|
| 310 |
+
"reported_accuracy": 0.391304347826087,
|
| 311 |
+
"competition_metrics": {
|
| 312 |
+
"arxiv/january": {
|
| 313 |
+
"n_problems_scored": 23,
|
| 314 |
+
"n_trajectories_scored": 92,
|
| 315 |
+
"reported_accuracy": 0.391304347826087
|
| 316 |
+
}
|
| 317 |
+
}
|
| 318 |
+
},
|
| 319 |
+
"benchmark": "matharena-arxivmath-january",
|
| 320 |
+
"model": "gemma4-12b-thinking-default-20261028",
|
| 321 |
+
"execution_mode": "non_agent",
|
| 322 |
+
"generation_hash": "8444f820a02c4a0f2522b7788d6a7b4b51411062b5283c56b97ae486e94388cd",
|
| 323 |
+
"scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8",
|
| 324 |
+
"selected": 92,
|
| 325 |
+
"incomplete": 0,
|
| 326 |
+
"generated_now": 92,
|
| 327 |
+
"scored_now": 92,
|
| 328 |
+
"cached_completed": 0,
|
| 329 |
+
"stage": "all",
|
| 330 |
+
"worker_count": 1,
|
| 331 |
+
"worker_index": 0,
|
| 332 |
+
"worker_owned_predictions": 92,
|
| 333 |
+
"generation_concurrency": 512,
|
| 334 |
+
"judge_concurrency": 256,
|
| 335 |
+
"updated_at": "2026-10-28T19:40:31.309145+00:00"
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"total_records": 124,
|
| 339 |
+
"completed": 124,
|
| 340 |
+
"query_errors": 0,
|
| 341 |
+
"score_errors": 0,
|
| 342 |
+
"generated_unscored": 0,
|
| 343 |
+
"mean_score": 0.25806451612903225,
|
| 344 |
+
"accuracy": 0.25806451612903225,
|
| 345 |
+
"mean_latency_seconds": 327.9356012846641,
|
| 346 |
+
"finish_reason_counts": {
|
| 347 |
+
"stop": 124
|
| 348 |
+
},
|
| 349 |
+
"length_limited_response_count": 0,
|
| 350 |
+
"observed_max_completion_tokens": 41506,
|
| 351 |
+
"judge_usage": {
|
| 352 |
+
"judged_record_count": 0,
|
| 353 |
+
"request_count": 0,
|
| 354 |
+
"retry_request_count": 0,
|
| 355 |
+
"prompt_tokens": 0,
|
| 356 |
+
"cached_prompt_tokens": 0,
|
| 357 |
+
"cache_write_prompt_tokens": 0,
|
| 358 |
+
"completion_tokens": 0,
|
| 359 |
+
"reasoning_tokens": 0,
|
| 360 |
+
"total_tokens": 0
|
| 361 |
+
},
|
| 362 |
+
"judge_api_usage": {
|
| 363 |
+
"request_count": 0,
|
| 364 |
+
"prompt_tokens": 0,
|
| 365 |
+
"cached_prompt_tokens": 0,
|
| 366 |
+
"cache_write_prompt_tokens": 0,
|
| 367 |
+
"completion_tokens": 0,
|
| 368 |
+
"reasoning_tokens": 0,
|
| 369 |
+
"total_tokens": 0
|
| 370 |
+
},
|
| 371 |
+
"benchmark_metrics": {
|
| 372 |
+
"competition": "arxivmath",
|
| 373 |
+
"aggregation": "avg@4",
|
| 374 |
+
"n_problems_scored": 31,
|
| 375 |
+
"n_trajectories_scored": 124,
|
| 376 |
+
"reported_accuracy": 0.25806451612903225,
|
| 377 |
+
"competition_metrics": {
|
| 378 |
+
"arxiv/march": {
|
| 379 |
+
"n_problems_scored": 31,
|
| 380 |
+
"n_trajectories_scored": 124,
|
| 381 |
+
"reported_accuracy": 0.25806451612903225
|
| 382 |
+
}
|
| 383 |
+
}
|
| 384 |
+
},
|
| 385 |
+
"benchmark": "matharena-arxivmath-march",
|
| 386 |
+
"model": "gemma4-12b-thinking-default-20261028",
|
| 387 |
+
"execution_mode": "non_agent",
|
| 388 |
+
"generation_hash": "3981d7dfa0cc9c30ffa6838accc875e14d491450732c85562b4885c7427aaa7f",
|
| 389 |
+
"scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8",
|
| 390 |
+
"selected": 124,
|
| 391 |
+
"incomplete": 0,
|
| 392 |
+
"generated_now": 119,
|
| 393 |
+
"scored_now": 119,
|
| 394 |
+
"cached_completed": 5,
|
| 395 |
+
"stage": "all",
|
| 396 |
+
"worker_count": 1,
|
| 397 |
+
"worker_index": 0,
|
| 398 |
+
"worker_owned_predictions": 124,
|
| 399 |
+
"generation_concurrency": 512,
|
| 400 |
+
"judge_concurrency": 256,
|
| 401 |
+
"updated_at": "2026-10-28T19:37:10.757171+00:00"
|
| 402 |
+
}
|
| 403 |
+
],
|
| 404 |
+
"total_records": 412,
|
| 405 |
+
"completed": 412,
|
| 406 |
+
"accuracy": 0.3155339805825243,
|
| 407 |
+
"provisional": false,
|
| 408 |
+
"grading_complete": true,
|
| 409 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run",
|
| 410 |
+
"publication_date": "2026-10-29"
|
| 411 |
+
},
|
| 412 |
+
"scicode-verified": {
|
| 413 |
+
"updated_at": 1793218906.0395374,
|
| 414 |
+
"expected_problems": 64,
|
| 415 |
+
"expected_steps": 287,
|
| 416 |
+
"generated_steps": 287,
|
| 417 |
+
"scored_steps": 287,
|
| 418 |
+
"passed_steps": 189,
|
| 419 |
+
"completed_problems": 64,
|
| 420 |
+
"passed_problems": 31,
|
| 421 |
+
"subproblem_accuracy": 0.6585365853658537,
|
| 422 |
+
"main_problem_accuracy": 0.484375,
|
| 423 |
+
"complete": true,
|
| 424 |
+
"cap_hits": 0,
|
| 425 |
+
"mean_output_tokens": 6563.003484320558,
|
| 426 |
+
"grading_infrastructure_errors": 0,
|
| 427 |
+
"errors": [],
|
| 428 |
+
"provisional": false,
|
| 429 |
+
"grading_complete": true,
|
| 430 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run",
|
| 431 |
+
"publication_date": "2026-10-29"
|
| 432 |
+
},
|
| 433 |
+
"scicoding-agent": {
|
| 434 |
+
"tasks": 50,
|
| 435 |
+
"attempts": 100,
|
| 436 |
+
"attempts_per_task": 2,
|
| 437 |
+
"metric": "avg@2",
|
| 438 |
+
"passed": 30,
|
| 439 |
+
"avg_at_2": 0.3,
|
| 440 |
+
"unscored": 0,
|
| 441 |
+
"agent_timeouts": 0,
|
| 442 |
+
"reward_positive_with_exception": 0,
|
| 443 |
+
"scoring": "native_verifier",
|
| 444 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 445 |
+
"provisional": false,
|
| 446 |
+
"grading_complete": true,
|
| 447 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run",
|
| 448 |
+
"publication_date": "2026-10-29"
|
| 449 |
+
},
|
| 450 |
+
"terminal-bench-2.1": {
|
| 451 |
+
"tasks": 89,
|
| 452 |
+
"attempts": 178,
|
| 453 |
+
"attempts_per_task": 2,
|
| 454 |
+
"metric": "avg@2",
|
| 455 |
+
"passed": 41,
|
| 456 |
+
"avg_at_2": 0.2303370786516854,
|
| 457 |
+
"unscored": 0,
|
| 458 |
+
"agent_timeouts": 73,
|
| 459 |
+
"reward_positive_with_exception": 0,
|
| 460 |
+
"scoring": "native_verifier",
|
| 461 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 462 |
+
"provisional": false,
|
| 463 |
+
"grading_complete": true,
|
| 464 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/run",
|
| 465 |
+
"publication_date": "2026-10-29",
|
| 466 |
+
"scoring_adjudications": {
|
| 467 |
+
"terminal-bench-2.1--pytorch-model-recovery--attempt-002": {
|
| 468 |
+
"timestamp": "2026-10-29T02:54:23.209570+00:00",
|
| 469 |
+
"reward": 0.0,
|
| 470 |
+
"method": "user_authorized_zero_for_verifier_setup_timeout",
|
| 471 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 472 |
+
"reason": "Agent timed out; subsequent verifier exhausted its budget downloading dependencies before recorded test execution. Zero is a user adjudication, not a native test verdict.",
|
| 473 |
+
"native_reward": null,
|
| 474 |
+
"original_verifier_result": null,
|
| 475 |
+
"native_tests_executed": "not evidenced",
|
| 476 |
+
"generation_or_verifier_rerun": false,
|
| 477 |
+
"backup": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/archive/user-zero-adjudication-20261029/terminal-bench-2.1--pytorch-model-recovery--attempt-002",
|
| 478 |
+
"evidence_sha256": {
|
| 479 |
+
"result.json": "39fcbacd95efc60c6206c392dc10b1042e360d52ab01dfc9da3f3b5db7c01aeb",
|
| 480 |
+
"trial.log": "f4079b102eb2e030b59db97150da9e79e62888739af519694df2443f9fb7b01d",
|
| 481 |
+
"verifier/test-stdout.txt": "7327b6f99e09223054e707efa9e9e25dec6b37e7f9e6f2ea4ccf1ff14e0b7edf"
|
| 482 |
+
}
|
| 483 |
+
},
|
| 484 |
+
"terminal-bench-2.1--torch-pipeline-parallelism--attempt-002": {
|
| 485 |
+
"timestamp": "2026-10-29T02:50:58.603775+00:00",
|
| 486 |
+
"reward": 0,
|
| 487 |
+
"method": "adopt_existing_native_verifier_artifacts",
|
| 488 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 489 |
+
"reason": "Native reward.txt is zero; CTRF and test stdout evidence failed correctness tests. Harbor timed out before recording the native reward.",
|
| 490 |
+
"original_verifier_result": null,
|
| 491 |
+
"original_exception_type": "VerifierTimeoutError",
|
| 492 |
+
"evidence": {
|
| 493 |
+
"verifier/reward.txt": "9a271f2a916b0b6ee6cecb2426f0b3206ef074578be55d9bc94f6f3fe3ab86aa",
|
| 494 |
+
"verifier/ctrf.json": "806ed353196f76e1e04e9bb9c306cf062098f0c3d5d8c5993693e18ac663fdfd",
|
| 495 |
+
"verifier/test-stdout.txt": "08a57104364c6fb08ae9ab7a1a4f7a42759c705c37a9dc1c3fe01f693685a848",
|
| 496 |
+
"result.json": "8b76029fb1a0a222a12fb83a80ada203a18598a3793d840ed93c4d59f603c655"
|
| 497 |
+
},
|
| 498 |
+
"backup": "/workspace/data/sci-llm-training/runtime/gemma4-12b-full-suite-controller-20261027/gemma4-12b-full-suite-20261028/archive/native-zero-adoption-20261029/terminal-bench-2.1--torch-pipeline-parallelism--attempt-002",
|
| 499 |
+
"generation_or_verifier_rerun": false
|
| 500 |
+
}
|
| 501 |
+
},
|
| 502 |
+
"scoring_note": "One explicit user-authorized zero after verifier setup timeout; one adopted native zero. Original exceptions, null original rewards and backups retained. Not all scores are completed native test verdicts."
|
| 503 |
+
},
|
| 504 |
+
"hle-verified": {
|
| 505 |
+
"gold": {
|
| 506 |
+
"correct": 159,
|
| 507 |
+
"total": 575,
|
| 508 |
+
"accuracy": 0.27652173913043476,
|
| 509 |
+
"cap_hits": 3
|
| 510 |
+
},
|
| 511 |
+
"revision": {
|
| 512 |
+
"correct": 307,
|
| 513 |
+
"total": 1025,
|
| 514 |
+
"accuracy": 0.2995121951219512,
|
| 515 |
+
"cap_hits": 10
|
| 516 |
+
},
|
| 517 |
+
"correct": 466,
|
| 518 |
+
"total": 1600,
|
| 519 |
+
"overall_accuracy": 0.29125,
|
| 520 |
+
"all_complete": true,
|
| 521 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 522 |
+
"provenance": "hle-verified-provenance.json"
|
| 523 |
+
}
|
| 524 |
+
}
|
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/README.md
CHANGED
|
@@ -14,3 +14,17 @@ Completed full suite evaluation, published 2026-10-29.
|
|
| 14 |
Reasoning: 98,304 output tokens, temperature 1.0, top-p 1.0. FS uses GPT-5.6 Luna; arXiv Math uses deterministic grading; SciCode and agentic tasks use native executable verifiers. Exact frozen settings are in provenance.json. All model reasoning and available native artifacts are preserved; credentials are redacted.
|
| 15 |
|
| 16 |
TB avg@2: 43/178 = 24.16%, with 174 native scores plus four explicitly user-adjudicated zeros. Null original verifier results and timeout evidence are retained, not represented as successful native verification. The nine replaced attempts are archived for audit and do not add to the denominator.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
Reasoning: 98,304 output tokens, temperature 1.0, top-p 1.0. FS uses GPT-5.6 Luna; arXiv Math uses deterministic grading; SciCode and agentic tasks use native executable verifiers. Exact frozen settings are in provenance.json. All model reasoning and available native artifacts are preserved; credentials are redacted.
|
| 15 |
|
| 16 |
TB avg@2: 43/178 = 24.16%, with 174 native scores plus four explicitly user-adjudicated zeros. Null original verifier results and timeout evidence are retained, not represented as successful native verification. The nine replaced attempts are archived for audit and do not add to the denominator.
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
## HLE-Verified (2026-10-30)
|
| 20 |
+
|
| 21 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **426/1,600 (26.62%)**.
|
| 22 |
+
|
| 23 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 24 |
+
|---|---:|---:|---|
|
| 25 |
+
| Gold | 134/575 | 23.30% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 26 |
+
| Revision | 292/1025 | 28.49% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 27 |
+
|
| 28 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 98,304 tokens; context 131,072 tokens. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 29 |
+
|
| 30 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,398 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "nemotron35-lightning",
|
| 7 |
+
"folder": "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 134,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.23304347826086957,
|
| 13 |
+
"cap_hits": 57
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 292,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.2848780487804878,
|
| 19 |
+
"cap_hits": 95
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 426,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.26625
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "nemotron35-lightning",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "nemotron35-lightning-30b-a3b",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "nemotron",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "nemotron_h",
|
| 35 |
+
"hidden_size": 2688,
|
| 36 |
+
"num_hidden_layers": 52,
|
| 37 |
+
"vocab_size": 131072
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 98304,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"chat_template_kwargs": {
|
| 46 |
+
"enable_thinking": true
|
| 47 |
+
}
|
| 48 |
+
}
|
| 49 |
+
},
|
| 50 |
+
"serving": {
|
| 51 |
+
"context_length": 131072,
|
| 52 |
+
"dtype": "bfloat16",
|
| 53 |
+
"attention_backend": "fa3",
|
| 54 |
+
"mem_fraction_static": 0.85,
|
| 55 |
+
"reasoning_parser": "nemotron_3",
|
| 56 |
+
"extra_args": [
|
| 57 |
+
"--mamba-ssm-dtype",
|
| 58 |
+
"float16",
|
| 59 |
+
"--cuda-graph-max-bs-decode",
|
| 60 |
+
"32",
|
| 61 |
+
"--tool-call-parser",
|
| 62 |
+
"qwen3_coder"
|
| 63 |
+
],
|
| 64 |
+
"tp": 2,
|
| 65 |
+
"replicas_per_node": 4,
|
| 66 |
+
"max_running_requests": 32
|
| 67 |
+
},
|
| 68 |
+
"reasoning": {
|
| 69 |
+
"requests_per_engine": 32,
|
| 70 |
+
"benchmarks": {}
|
| 71 |
+
},
|
| 72 |
+
"agentic_tasks_per_engine": 1
|
| 73 |
+
},
|
| 74 |
+
"node": 0,
|
| 75 |
+
"checkpoint": "/workspace/data/raw_model/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16-a9904d24bcc1",
|
| 76 |
+
"repo": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16",
|
| 77 |
+
"revision": "a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 78 |
+
"checkpoint_config_sha256": "a3827a0f5e311547b40943dc081e3ff2f8a277466e8c1a3df2291e8db8a7617c",
|
| 79 |
+
"index_sha256": "67f21da80ce245a3e24967f54eef3fa10e67a63eba16ef19a2d0569f06103f50",
|
| 80 |
+
"shard_sizes": {
|
| 81 |
+
"model-00001-of-00014.safetensors": 4991210024,
|
| 82 |
+
"model-00002-of-00014.safetensors": 4992601016,
|
| 83 |
+
"model-00003-of-00014.safetensors": 4992601368,
|
| 84 |
+
"model-00004-of-00014.safetensors": 4995692800,
|
| 85 |
+
"model-00005-of-00014.safetensors": 4980553832,
|
| 86 |
+
"model-00006-of-00014.safetensors": 4989423536,
|
| 87 |
+
"model-00007-of-00014.safetensors": 4992601488,
|
| 88 |
+
"model-00008-of-00014.safetensors": 4992601520,
|
| 89 |
+
"model-00009-of-00014.safetensors": 4995692800,
|
| 90 |
+
"model-00010-of-00014.safetensors": 4992601520,
|
| 91 |
+
"model-00011-of-00014.safetensors": 4995692800,
|
| 92 |
+
"model-00012-of-00014.safetensors": 4995692816,
|
| 93 |
+
"model-00013-of-00014.safetensors": 3249723504,
|
| 94 |
+
"model-00014-of-00014.safetensors": 2670685240
|
| 95 |
+
},
|
| 96 |
+
"endpoints": [
|
| 97 |
+
{
|
| 98 |
+
"id": "node-0-replica-0",
|
| 99 |
+
"api_base": "http://10.2.5.84:8700/v1"
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"id": "node-0-replica-1",
|
| 103 |
+
"api_base": "http://10.2.5.84:8701/v1"
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"id": "node-0-replica-2",
|
| 107 |
+
"api_base": "http://10.2.5.84:8702/v1"
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"id": "node-0-replica-3",
|
| 111 |
+
"api_base": "http://10.2.5.84:8703/v1"
|
| 112 |
+
}
|
| 113 |
+
],
|
| 114 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/nemotron35-lightning",
|
| 115 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/nemotron35-lightning/configs/hle-verified-gold-revision-96k.yaml",
|
| 116 |
+
"concurrency": 128,
|
| 117 |
+
"judge_concurrency": 64,
|
| 118 |
+
"tracking": {
|
| 119 |
+
"enabled": false,
|
| 120 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 121 |
+
},
|
| 122 |
+
"config_sha256": {
|
| 123 |
+
"model.yaml": "2021a2aa4f08953902ccd9e33ff9ab2ad754f301c3e66e4735df32591127b0ed",
|
| 124 |
+
"configs/hle-verified-gold-revision-96k.yaml": "eaa36f2db209bce2f9de06724792d4d03f0ad0139c88b50ccd0e9a780126117c",
|
| 125 |
+
"configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4",
|
| 126 |
+
"configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"
|
| 127 |
+
}
|
| 128 |
+
},
|
| 129 |
+
"dataset": {
|
| 130 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 131 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 132 |
+
"counts": {
|
| 133 |
+
"Gold": {
|
| 134 |
+
"total": 668,
|
| 135 |
+
"text": 575,
|
| 136 |
+
"image_excluded": 93
|
| 137 |
+
},
|
| 138 |
+
"Revision": {
|
| 139 |
+
"total": 1143,
|
| 140 |
+
"text": 1025,
|
| 141 |
+
"image_excluded": 118
|
| 142 |
+
}
|
| 143 |
+
},
|
| 144 |
+
"sha256": {
|
| 145 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 146 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 147 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 148 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 149 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 150 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 151 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 152 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 153 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 154 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 155 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 156 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 157 |
+
}
|
| 158 |
+
},
|
| 159 |
+
"token_preflight": {
|
| 160 |
+
"model": "nemotron35-lightning",
|
| 161 |
+
"questions": 1600,
|
| 162 |
+
"max_prompt_tokens": 13776,
|
| 163 |
+
"output_cap": 98304,
|
| 164 |
+
"context": 131072,
|
| 165 |
+
"dynamic_output_budget": false,
|
| 166 |
+
"affected_questions": 0,
|
| 167 |
+
"min_effective_output_cap": 98304
|
| 168 |
+
},
|
| 169 |
+
"shared_recovery_provenance": {
|
| 170 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 171 |
+
"publication_reruns": false,
|
| 172 |
+
"all_questions_scored": true,
|
| 173 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 174 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 175 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 176 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 177 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 178 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 179 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 180 |
+
},
|
| 181 |
+
"source_manifest": {
|
| 182 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 183 |
+
"files": {
|
| 184 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 185 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 186 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 187 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 188 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 189 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 190 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 191 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 192 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 193 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 194 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 195 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 196 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 197 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 198 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 199 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 200 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 201 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 202 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 203 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 204 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 205 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 206 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 207 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 208 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 209 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 210 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 211 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 212 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 213 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 214 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 215 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 216 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 217 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 218 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 219 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 220 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 221 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 222 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 223 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 224 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 225 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 226 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 227 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 228 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 229 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 230 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 231 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 232 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 233 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 234 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 235 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 236 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 237 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 238 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 239 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 240 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 241 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 242 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 243 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 244 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 245 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 246 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 247 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 248 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 249 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 250 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 251 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 252 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 253 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 254 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 255 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 256 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 257 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 258 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 259 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 260 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 261 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 262 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 263 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 264 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 265 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 266 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 267 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 268 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 269 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 270 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 271 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 272 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 273 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 274 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 275 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 276 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 277 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 278 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 279 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 280 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 281 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 282 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 283 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 284 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 285 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 286 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 287 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 288 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 289 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 290 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 291 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 292 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 293 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 294 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 295 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 296 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 297 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 298 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 299 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 300 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 301 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 302 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 303 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 304 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 305 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 306 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 307 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 308 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 309 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 310 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 311 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 312 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 313 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 314 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 315 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 316 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 317 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 318 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 319 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 320 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 321 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 322 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 323 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 324 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 325 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 326 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 327 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 328 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 329 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 330 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 331 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 332 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 333 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 334 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 335 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 336 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 337 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 338 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 339 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 340 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 341 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 342 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 343 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 344 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 345 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 346 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 347 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 348 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 349 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 350 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 351 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 352 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 353 |
+
},
|
| 354 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 355 |
+
},
|
| 356 |
+
"artifact_manifest": {
|
| 357 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/configs.json": {
|
| 358 |
+
"size": 96540,
|
| 359 |
+
"sha256": "63416cb9660523e05f57d41dba3d99dac1a41bdfcb901896daf3074a3ba6af9c"
|
| 360 |
+
},
|
| 361 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/events.jsonl.gz": {
|
| 362 |
+
"size": 43697680,
|
| 363 |
+
"sha256": "8b4f3ddd8cdd092dd94e45ecbf2bd058dfc778025dfacc5564174d3f22bf3d6d"
|
| 364 |
+
},
|
| 365 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/judge_trajectories.jsonl": {
|
| 366 |
+
"size": 807987,
|
| 367 |
+
"sha256": "194d663838cd73be2ca90824ddbf75386937fa67b0b98bc07b399f54a3f6560d"
|
| 368 |
+
},
|
| 369 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/summary.json": {
|
| 370 |
+
"size": 1504,
|
| 371 |
+
"sha256": "69aa51f1a177ae8489ec03df54b4b82b0940cb4927c2217aaad75fd1fe4ff3a9"
|
| 372 |
+
},
|
| 373 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/trajectories.jsonl": {
|
| 374 |
+
"size": 86865668,
|
| 375 |
+
"sha256": "a03274a1b9b7c94f4250161e93343c41747022bdd35c9e14bd660d0574b273fd"
|
| 376 |
+
},
|
| 377 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/configs.json": {
|
| 378 |
+
"size": 164972,
|
| 379 |
+
"sha256": "9110a79193b19841f399f51c14b824c2e4c15a2dfc745a8ca08805b4ce69d037"
|
| 380 |
+
},
|
| 381 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/events.jsonl.gz": {
|
| 382 |
+
"size": 81111047,
|
| 383 |
+
"sha256": "83707e1fd3752c6828b5f53d57761141bc6c05f942a577bf3cbe20a18bcaf26c"
|
| 384 |
+
},
|
| 385 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/judge_trajectories.jsonl": {
|
| 386 |
+
"size": 1460099,
|
| 387 |
+
"sha256": "d3bc211cc10b3443ed1baafbb45336c8bb961e176b31c31998864240ad47f6f8"
|
| 388 |
+
},
|
| 389 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/summary.json": {
|
| 390 |
+
"size": 1519,
|
| 391 |
+
"sha256": "737e73ccadc0e159ef4caae8a78565cf668f74ff41889f30c10719014acab918"
|
| 392 |
+
},
|
| 393 |
+
"NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/trajectories.jsonl": {
|
| 394 |
+
"size": 158052659,
|
| 395 |
+
"sha256": "a1a76fdad16f0182616ca18263ddb1d4e6c4f3d47b5466ffeca42e87749932ba"
|
| 396 |
+
}
|
| 397 |
+
}
|
| 398 |
+
}
|
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/summary.json
CHANGED
|
@@ -1 +1,536 @@
|
|
| 1 |
-
{"arxivmath": {"benchmark": "arxivmath", "releases": [{"total_records": 68, "completed": 68, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.35294117647058826, "accuracy": 0.35294117647058826, "mean_latency_seconds": 836.4844917968303, "finish_reason_counts": {"length": 9, "stop": 59}, "length_limited_response_count": 9, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.35294117647058826, "competition_metrics": {"arxiv/december": {"n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.35294117647058826}}}, "benchmark": "matharena-arxivmath-december", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "execution_mode": "non_agent", "generation_hash": "743b679a458bb461b649916e6a8f4b51ed814f6d79c63609036126177f3310bc", "scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661", "selected": 68, "incomplete": 0, "generated_now": 55, "scored_now": 55, "cached_completed": 13, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 68, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-29T05:56:09.635826+00:00"}, {"total_records": 128, "completed": 128, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.1796875, "accuracy": 0.1796875, "mean_latency_seconds": 1516.842542529219, "finish_reason_counts": {"length": 6, "stop": 122}, "length_limited_response_count": 6, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.1796875, "competition_metrics": {"arxiv/february": {"n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.1796875}}}, "benchmark": "matharena-arxivmath-february", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "execution_mode": "non_agent", "generation_hash": "ac454530ae6d74e65bca96b6ad35052f78882e30531bb988e903c4df4ff9757a", "scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5", "selected": 128, "incomplete": 0, "generated_now": 48, "scored_now": 48, "cached_completed": 80, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 128, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-29T05:54:58.573288+00:00"}, {"total_records": 92, "completed": 92, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.34782608695652173, "accuracy": 0.34782608695652173, "mean_latency_seconds": 1037.4007359575075, "finish_reason_counts": {"length": 7, "stop": 85}, "length_limited_response_count": 7, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.34782608695652173, "competition_metrics": {"arxiv/january": {"n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.34782608695652173}}}, "benchmark": "matharena-arxivmath-january", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "execution_mode": "non_agent", "generation_hash": "20caa2f614cc612500566e949fb8e60f7addccdface08b8c97a489098db37741", "scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8", "selected": 92, "incomplete": 0, "generated_now": 74, "scored_now": 74, "cached_completed": 18, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 92, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-29T05:57:16.061403+00:00"}, {"total_records": 124, "completed": 124, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.24193548387096775, "accuracy": 0.24193548387096775, "mean_latency_seconds": 765.4728525543685, "finish_reason_counts": {"length": 6, "stop": 118}, "length_limited_response_count": 6, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.24193548387096775, "competition_metrics": {"arxiv/march": {"n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.24193548387096775}}}, "benchmark": "matharena-arxivmath-march", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "execution_mode": "non_agent", "generation_hash": "06fad0949d70d5b686c16d8acb8610bd1277b13ba5fb5e0a894ab0132c1d885a", "scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8", "selected": 124, "incomplete": 0, "generated_now": 93, "scored_now": 93, "cached_completed": 31, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 124, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-29T05:58:44.434614+00:00"}], "total_records": 412, "completed": 412, "accuracy": 0.2645631067961165, "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"}, "frontierscience-olympiad": {"total_records": 3000, "completed": 3000, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.542, "accuracy": 0.542, "mean_latency_seconds": 530.9761690581157, "finish_reason_counts": {"length": 301, "stop": 2699}, "length_limited_response_count": 301, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 3000, "request_count": 3000, "retry_request_count": 0, "prompt_tokens": 2802503, "cached_prompt_tokens": 27218, "cache_write_prompt_tokens": 1235799, "completion_tokens": 674208, "reasoning_tokens": 448074, "total_tokens": 3476711}, "judge_api_usage": {"request_count": 3000, "prompt_tokens": 2802503, "cached_prompt_tokens": 27218, "cache_write_prompt_tokens": 1235799, "completion_tokens": 674208, "reasoning_tokens": 448074, "total_tokens": 3476711}, "benchmark_metrics": {"track": "olympiad", "n_problems_scored": 100, "n_trajectories_scored": 3000, "aggregation": "avg@30", "reported_accuracy": 0.542, "mean_rubric_score": null, "success_threshold": 0.5, "community_profile": null}, "benchmark": "frontierscience-olympiad", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "execution_mode": "non_agent", "generation_hash": "093a6888a7b2df2153f1c00b78389a167c22fd3a4c039f79240683d24a27ed5e", "scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016", "selected": 3000, "incomplete": 0, "generated_now": 2841, "scored_now": 2841, "cached_completed": 159, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 3000, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-29T06:45:00.048121+00:00", "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"}, "frontierscience-research": {"total_records": 1800, "completed": 1800, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.29168638888888887, "accuracy": 0.07166666666666667, "mean_latency_seconds": 549.7299430483823, "finish_reason_counts": {"length": 79, "stop": 1721}, "length_limited_response_count": 79, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 1800, "request_count": 1832, "retry_request_count": 32, "prompt_tokens": 11125895, "cached_prompt_tokens": 246486, "cache_write_prompt_tokens": 10836821, "completion_tokens": 3061669, "reasoning_tokens": 1817948, "total_tokens": 14187564}, "judge_api_usage": {"request_count": 1832, "prompt_tokens": 11125895, "cached_prompt_tokens": 246486, "cache_write_prompt_tokens": 10836821, "completion_tokens": 3061669, "reasoning_tokens": 1817948, "total_tokens": 14187564}, "benchmark_metrics": {"track": "research", "n_problems_scored": 60, "n_trajectories_scored": 1800, "aggregation": "avg@30", "reported_accuracy": 0.07166666666666667, "mean_rubric_score": 0.29168638888888887, "success_threshold": 0.7, "community_profile": null}, "benchmark": "frontierscience-research", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "execution_mode": "non_agent", "generation_hash": "3749020dcb028aa6fd38943de4c92c32672113762d2745830e10358371eea801", "scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750", "selected": 1800, "incomplete": 0, "generated_now": 1561, "scored_now": 1564, "cached_completed": 236, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 1800, "generation_concurrency": 512, "judge_concurrency": 256, "updated_at": "2026-10-29T06:34:29.975558+00:00", "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"}, "scicode-verified": {"updated_at": 1793255967.5510898, "expected_problems": 64, "expected_steps": 287, "generated_steps": 287, "scored_steps": 287, "passed_steps": 155, "completed_problems": 64, "passed_problems": 21, "subproblem_accuracy": 0.5400696864111498, "main_problem_accuracy": 0.328125, "complete": true, "cap_hits": 0, "mean_output_tokens": 9008.062717770035, "grading_infrastructure_errors": 0, "errors": [], "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"}, "scicoding-agent": {"tasks": 50, "attempts": 100, "attempts_per_task": 2, "metric": "avg@2", "passed": 26, "avg_at_2": 0.26, "unscored": 0, "agent_timeouts": 2, "reward_positive_with_exception": 1, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1", "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"}, "terminal-bench-2.1": {"tasks": 89, "attempts": 178, "attempts_per_task": 2, "metric": "avg@2", "passed": 43, "avg_at_2": 0.24157303370786518, "unscored": 0, "agent_timeouts": 58, "reward_positive_with_exception": 3, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1", "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run", "scoring_adjudications": {"mailman--attempt-002": {"timestamp": "2026-10-29T19:11:17.460325+00:00", "reward": 0.0, "native_reward": null, "original_verifier_result": null, "method": "explicit_user_authorized_zero_for_unscored_attempt", "authorization": "[REDACTED_CREDENTIAL]", "reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.", "exception_type": "VerifierTimeoutError", "backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--mailman--attempt-002", "original_sha256": "ef5726e0fa911c30c412c7f7cbe08181541c85351260634745e65e0d147b7ab4", "generation_or_verifier_rerun": false}, "pytorch-model-recovery--attempt-001": {"timestamp": "2026-10-29T19:11:17.511954+00:00", "reward": 0.0, "native_reward": null, "original_verifier_result": null, "method": "explicit_user_authorized_zero_for_unscored_attempt", "authorization": "[REDACTED_CREDENTIAL]", "reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.", "exception_type": "AgentTimeoutError", "backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--pytorch-model-recovery--attempt-001", "original_sha256": "68430013dc4d101ddd43f9dbacb6a17c5b772d507b8301ef3b18c6f30904340c", "generation_or_verifier_rerun": false}, "pytorch-model-recovery--attempt-002": {"timestamp": "2026-10-29T19:11:17.565699+00:00", "reward": 0.0, "native_reward": null, "original_verifier_result": null, "method": "explicit_user_authorized_zero_for_unscored_attempt", "authorization": "[REDACTED_CREDENTIAL]", "reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.", "exception_type": "AgentTimeoutError", "backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--pytorch-model-recovery--attempt-002", "original_sha256": "7fae00d6c6402beb1764a220d36240d2234c301adbcb73a2386afc5c69e971a5", "generation_or_verifier_rerun": false}, "torch-pipeline-parallelism--attempt-002": {"timestamp": "2026-10-29T19:11:17.623946+00:00", "reward": 0.0, "native_reward": null, "original_verifier_result": null, "method": "explicit_user_authorized_zero_for_unscored_attempt", "authorization": "[REDACTED_CREDENTIAL]", "reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.", "exception_type": "VerifierTimeoutError", "backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--torch-pipeline-parallelism--attempt-002", "original_sha256": "67322cae95590fb5ba48bdf823a06f30d0861c161d2cd86c15adfad997ee3ac7", "generation_or_verifier_rerun": false}}, "native_scored": 174, "adjudicated_zero": 4, "scoring_note": "Four missing native verdicts assigned zero by explicit user instruction; original exceptions and null rewards preserved in native_artifacts.tar.gz and recovery_provenance.tar.gz."}}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"arxivmath": {
|
| 3 |
+
"benchmark": "arxivmath",
|
| 4 |
+
"releases": [
|
| 5 |
+
{
|
| 6 |
+
"total_records": 68,
|
| 7 |
+
"completed": 68,
|
| 8 |
+
"query_errors": 0,
|
| 9 |
+
"score_errors": 0,
|
| 10 |
+
"generated_unscored": 0,
|
| 11 |
+
"mean_score": 0.35294117647058826,
|
| 12 |
+
"accuracy": 0.35294117647058826,
|
| 13 |
+
"mean_latency_seconds": 836.4844917968303,
|
| 14 |
+
"finish_reason_counts": {
|
| 15 |
+
"length": 9,
|
| 16 |
+
"stop": 59
|
| 17 |
+
},
|
| 18 |
+
"length_limited_response_count": 9,
|
| 19 |
+
"observed_max_completion_tokens": 98304,
|
| 20 |
+
"judge_usage": {
|
| 21 |
+
"judged_record_count": 0,
|
| 22 |
+
"request_count": 0,
|
| 23 |
+
"retry_request_count": 0,
|
| 24 |
+
"prompt_tokens": 0,
|
| 25 |
+
"cached_prompt_tokens": 0,
|
| 26 |
+
"cache_write_prompt_tokens": 0,
|
| 27 |
+
"completion_tokens": 0,
|
| 28 |
+
"reasoning_tokens": 0,
|
| 29 |
+
"total_tokens": 0
|
| 30 |
+
},
|
| 31 |
+
"judge_api_usage": {
|
| 32 |
+
"request_count": 0,
|
| 33 |
+
"prompt_tokens": 0,
|
| 34 |
+
"cached_prompt_tokens": 0,
|
| 35 |
+
"cache_write_prompt_tokens": 0,
|
| 36 |
+
"completion_tokens": 0,
|
| 37 |
+
"reasoning_tokens": 0,
|
| 38 |
+
"total_tokens": 0
|
| 39 |
+
},
|
| 40 |
+
"benchmark_metrics": {
|
| 41 |
+
"competition": "arxivmath",
|
| 42 |
+
"aggregation": "avg@4",
|
| 43 |
+
"n_problems_scored": 17,
|
| 44 |
+
"n_trajectories_scored": 68,
|
| 45 |
+
"reported_accuracy": 0.35294117647058826,
|
| 46 |
+
"competition_metrics": {
|
| 47 |
+
"arxiv/december": {
|
| 48 |
+
"n_problems_scored": 17,
|
| 49 |
+
"n_trajectories_scored": 68,
|
| 50 |
+
"reported_accuracy": 0.35294117647058826
|
| 51 |
+
}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"benchmark": "matharena-arxivmath-december",
|
| 55 |
+
"model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 56 |
+
"execution_mode": "non_agent",
|
| 57 |
+
"generation_hash": "743b679a458bb461b649916e6a8f4b51ed814f6d79c63609036126177f3310bc",
|
| 58 |
+
"scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661",
|
| 59 |
+
"selected": 68,
|
| 60 |
+
"incomplete": 0,
|
| 61 |
+
"generated_now": 55,
|
| 62 |
+
"scored_now": 55,
|
| 63 |
+
"cached_completed": 13,
|
| 64 |
+
"stage": "all",
|
| 65 |
+
"worker_count": 1,
|
| 66 |
+
"worker_index": 0,
|
| 67 |
+
"worker_owned_predictions": 68,
|
| 68 |
+
"generation_concurrency": 512,
|
| 69 |
+
"judge_concurrency": 256,
|
| 70 |
+
"updated_at": "2026-10-29T05:56:09.635826+00:00"
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"total_records": 128,
|
| 74 |
+
"completed": 128,
|
| 75 |
+
"query_errors": 0,
|
| 76 |
+
"score_errors": 0,
|
| 77 |
+
"generated_unscored": 0,
|
| 78 |
+
"mean_score": 0.1796875,
|
| 79 |
+
"accuracy": 0.1796875,
|
| 80 |
+
"mean_latency_seconds": 1516.842542529219,
|
| 81 |
+
"finish_reason_counts": {
|
| 82 |
+
"length": 6,
|
| 83 |
+
"stop": 122
|
| 84 |
+
},
|
| 85 |
+
"length_limited_response_count": 6,
|
| 86 |
+
"observed_max_completion_tokens": 98304,
|
| 87 |
+
"judge_usage": {
|
| 88 |
+
"judged_record_count": 0,
|
| 89 |
+
"request_count": 0,
|
| 90 |
+
"retry_request_count": 0,
|
| 91 |
+
"prompt_tokens": 0,
|
| 92 |
+
"cached_prompt_tokens": 0,
|
| 93 |
+
"cache_write_prompt_tokens": 0,
|
| 94 |
+
"completion_tokens": 0,
|
| 95 |
+
"reasoning_tokens": 0,
|
| 96 |
+
"total_tokens": 0
|
| 97 |
+
},
|
| 98 |
+
"judge_api_usage": {
|
| 99 |
+
"request_count": 0,
|
| 100 |
+
"prompt_tokens": 0,
|
| 101 |
+
"cached_prompt_tokens": 0,
|
| 102 |
+
"cache_write_prompt_tokens": 0,
|
| 103 |
+
"completion_tokens": 0,
|
| 104 |
+
"reasoning_tokens": 0,
|
| 105 |
+
"total_tokens": 0
|
| 106 |
+
},
|
| 107 |
+
"benchmark_metrics": {
|
| 108 |
+
"competition": "arxivmath",
|
| 109 |
+
"aggregation": "avg@4",
|
| 110 |
+
"n_problems_scored": 32,
|
| 111 |
+
"n_trajectories_scored": 128,
|
| 112 |
+
"reported_accuracy": 0.1796875,
|
| 113 |
+
"competition_metrics": {
|
| 114 |
+
"arxiv/february": {
|
| 115 |
+
"n_problems_scored": 32,
|
| 116 |
+
"n_trajectories_scored": 128,
|
| 117 |
+
"reported_accuracy": 0.1796875
|
| 118 |
+
}
|
| 119 |
+
}
|
| 120 |
+
},
|
| 121 |
+
"benchmark": "matharena-arxivmath-february",
|
| 122 |
+
"model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 123 |
+
"execution_mode": "non_agent",
|
| 124 |
+
"generation_hash": "ac454530ae6d74e65bca96b6ad35052f78882e30531bb988e903c4df4ff9757a",
|
| 125 |
+
"scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5",
|
| 126 |
+
"selected": 128,
|
| 127 |
+
"incomplete": 0,
|
| 128 |
+
"generated_now": 48,
|
| 129 |
+
"scored_now": 48,
|
| 130 |
+
"cached_completed": 80,
|
| 131 |
+
"stage": "all",
|
| 132 |
+
"worker_count": 1,
|
| 133 |
+
"worker_index": 0,
|
| 134 |
+
"worker_owned_predictions": 128,
|
| 135 |
+
"generation_concurrency": 512,
|
| 136 |
+
"judge_concurrency": 256,
|
| 137 |
+
"updated_at": "2026-10-29T05:54:58.573288+00:00"
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"total_records": 92,
|
| 141 |
+
"completed": 92,
|
| 142 |
+
"query_errors": 0,
|
| 143 |
+
"score_errors": 0,
|
| 144 |
+
"generated_unscored": 0,
|
| 145 |
+
"mean_score": 0.34782608695652173,
|
| 146 |
+
"accuracy": 0.34782608695652173,
|
| 147 |
+
"mean_latency_seconds": 1037.4007359575075,
|
| 148 |
+
"finish_reason_counts": {
|
| 149 |
+
"length": 7,
|
| 150 |
+
"stop": 85
|
| 151 |
+
},
|
| 152 |
+
"length_limited_response_count": 7,
|
| 153 |
+
"observed_max_completion_tokens": 98304,
|
| 154 |
+
"judge_usage": {
|
| 155 |
+
"judged_record_count": 0,
|
| 156 |
+
"request_count": 0,
|
| 157 |
+
"retry_request_count": 0,
|
| 158 |
+
"prompt_tokens": 0,
|
| 159 |
+
"cached_prompt_tokens": 0,
|
| 160 |
+
"cache_write_prompt_tokens": 0,
|
| 161 |
+
"completion_tokens": 0,
|
| 162 |
+
"reasoning_tokens": 0,
|
| 163 |
+
"total_tokens": 0
|
| 164 |
+
},
|
| 165 |
+
"judge_api_usage": {
|
| 166 |
+
"request_count": 0,
|
| 167 |
+
"prompt_tokens": 0,
|
| 168 |
+
"cached_prompt_tokens": 0,
|
| 169 |
+
"cache_write_prompt_tokens": 0,
|
| 170 |
+
"completion_tokens": 0,
|
| 171 |
+
"reasoning_tokens": 0,
|
| 172 |
+
"total_tokens": 0
|
| 173 |
+
},
|
| 174 |
+
"benchmark_metrics": {
|
| 175 |
+
"competition": "arxivmath",
|
| 176 |
+
"aggregation": "avg@4",
|
| 177 |
+
"n_problems_scored": 23,
|
| 178 |
+
"n_trajectories_scored": 92,
|
| 179 |
+
"reported_accuracy": 0.34782608695652173,
|
| 180 |
+
"competition_metrics": {
|
| 181 |
+
"arxiv/january": {
|
| 182 |
+
"n_problems_scored": 23,
|
| 183 |
+
"n_trajectories_scored": 92,
|
| 184 |
+
"reported_accuracy": 0.34782608695652173
|
| 185 |
+
}
|
| 186 |
+
}
|
| 187 |
+
},
|
| 188 |
+
"benchmark": "matharena-arxivmath-january",
|
| 189 |
+
"model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 190 |
+
"execution_mode": "non_agent",
|
| 191 |
+
"generation_hash": "20caa2f614cc612500566e949fb8e60f7addccdface08b8c97a489098db37741",
|
| 192 |
+
"scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8",
|
| 193 |
+
"selected": 92,
|
| 194 |
+
"incomplete": 0,
|
| 195 |
+
"generated_now": 74,
|
| 196 |
+
"scored_now": 74,
|
| 197 |
+
"cached_completed": 18,
|
| 198 |
+
"stage": "all",
|
| 199 |
+
"worker_count": 1,
|
| 200 |
+
"worker_index": 0,
|
| 201 |
+
"worker_owned_predictions": 92,
|
| 202 |
+
"generation_concurrency": 512,
|
| 203 |
+
"judge_concurrency": 256,
|
| 204 |
+
"updated_at": "2026-10-29T05:57:16.061403+00:00"
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"total_records": 124,
|
| 208 |
+
"completed": 124,
|
| 209 |
+
"query_errors": 0,
|
| 210 |
+
"score_errors": 0,
|
| 211 |
+
"generated_unscored": 0,
|
| 212 |
+
"mean_score": 0.24193548387096775,
|
| 213 |
+
"accuracy": 0.24193548387096775,
|
| 214 |
+
"mean_latency_seconds": 765.4728525543685,
|
| 215 |
+
"finish_reason_counts": {
|
| 216 |
+
"length": 6,
|
| 217 |
+
"stop": 118
|
| 218 |
+
},
|
| 219 |
+
"length_limited_response_count": 6,
|
| 220 |
+
"observed_max_completion_tokens": 98304,
|
| 221 |
+
"judge_usage": {
|
| 222 |
+
"judged_record_count": 0,
|
| 223 |
+
"request_count": 0,
|
| 224 |
+
"retry_request_count": 0,
|
| 225 |
+
"prompt_tokens": 0,
|
| 226 |
+
"cached_prompt_tokens": 0,
|
| 227 |
+
"cache_write_prompt_tokens": 0,
|
| 228 |
+
"completion_tokens": 0,
|
| 229 |
+
"reasoning_tokens": 0,
|
| 230 |
+
"total_tokens": 0
|
| 231 |
+
},
|
| 232 |
+
"judge_api_usage": {
|
| 233 |
+
"request_count": 0,
|
| 234 |
+
"prompt_tokens": 0,
|
| 235 |
+
"cached_prompt_tokens": 0,
|
| 236 |
+
"cache_write_prompt_tokens": 0,
|
| 237 |
+
"completion_tokens": 0,
|
| 238 |
+
"reasoning_tokens": 0,
|
| 239 |
+
"total_tokens": 0
|
| 240 |
+
},
|
| 241 |
+
"benchmark_metrics": {
|
| 242 |
+
"competition": "arxivmath",
|
| 243 |
+
"aggregation": "avg@4",
|
| 244 |
+
"n_problems_scored": 31,
|
| 245 |
+
"n_trajectories_scored": 124,
|
| 246 |
+
"reported_accuracy": 0.24193548387096775,
|
| 247 |
+
"competition_metrics": {
|
| 248 |
+
"arxiv/march": {
|
| 249 |
+
"n_problems_scored": 31,
|
| 250 |
+
"n_trajectories_scored": 124,
|
| 251 |
+
"reported_accuracy": 0.24193548387096775
|
| 252 |
+
}
|
| 253 |
+
}
|
| 254 |
+
},
|
| 255 |
+
"benchmark": "matharena-arxivmath-march",
|
| 256 |
+
"model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 257 |
+
"execution_mode": "non_agent",
|
| 258 |
+
"generation_hash": "06fad0949d70d5b686c16d8acb8610bd1277b13ba5fb5e0a894ab0132c1d885a",
|
| 259 |
+
"scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8",
|
| 260 |
+
"selected": 124,
|
| 261 |
+
"incomplete": 0,
|
| 262 |
+
"generated_now": 93,
|
| 263 |
+
"scored_now": 93,
|
| 264 |
+
"cached_completed": 31,
|
| 265 |
+
"stage": "all",
|
| 266 |
+
"worker_count": 1,
|
| 267 |
+
"worker_index": 0,
|
| 268 |
+
"worker_owned_predictions": 124,
|
| 269 |
+
"generation_concurrency": 512,
|
| 270 |
+
"judge_concurrency": 256,
|
| 271 |
+
"updated_at": "2026-10-29T05:58:44.434614+00:00"
|
| 272 |
+
}
|
| 273 |
+
],
|
| 274 |
+
"total_records": 412,
|
| 275 |
+
"completed": 412,
|
| 276 |
+
"accuracy": 0.2645631067961165,
|
| 277 |
+
"provisional": false,
|
| 278 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"
|
| 279 |
+
},
|
| 280 |
+
"frontierscience-olympiad": {
|
| 281 |
+
"total_records": 3000,
|
| 282 |
+
"completed": 3000,
|
| 283 |
+
"query_errors": 0,
|
| 284 |
+
"score_errors": 0,
|
| 285 |
+
"generated_unscored": 0,
|
| 286 |
+
"mean_score": 0.542,
|
| 287 |
+
"accuracy": 0.542,
|
| 288 |
+
"mean_latency_seconds": 530.9761690581157,
|
| 289 |
+
"finish_reason_counts": {
|
| 290 |
+
"length": 301,
|
| 291 |
+
"stop": 2699
|
| 292 |
+
},
|
| 293 |
+
"length_limited_response_count": 301,
|
| 294 |
+
"observed_max_completion_tokens": 98304,
|
| 295 |
+
"judge_usage": {
|
| 296 |
+
"judged_record_count": 3000,
|
| 297 |
+
"request_count": 3000,
|
| 298 |
+
"retry_request_count": 0,
|
| 299 |
+
"prompt_tokens": 2802503,
|
| 300 |
+
"cached_prompt_tokens": 27218,
|
| 301 |
+
"cache_write_prompt_tokens": 1235799,
|
| 302 |
+
"completion_tokens": 674208,
|
| 303 |
+
"reasoning_tokens": 448074,
|
| 304 |
+
"total_tokens": 3476711
|
| 305 |
+
},
|
| 306 |
+
"judge_api_usage": {
|
| 307 |
+
"request_count": 3000,
|
| 308 |
+
"prompt_tokens": 2802503,
|
| 309 |
+
"cached_prompt_tokens": 27218,
|
| 310 |
+
"cache_write_prompt_tokens": 1235799,
|
| 311 |
+
"completion_tokens": 674208,
|
| 312 |
+
"reasoning_tokens": 448074,
|
| 313 |
+
"total_tokens": 3476711
|
| 314 |
+
},
|
| 315 |
+
"benchmark_metrics": {
|
| 316 |
+
"track": "olympiad",
|
| 317 |
+
"n_problems_scored": 100,
|
| 318 |
+
"n_trajectories_scored": 3000,
|
| 319 |
+
"aggregation": "avg@30",
|
| 320 |
+
"reported_accuracy": 0.542,
|
| 321 |
+
"mean_rubric_score": null,
|
| 322 |
+
"success_threshold": 0.5,
|
| 323 |
+
"community_profile": null
|
| 324 |
+
},
|
| 325 |
+
"benchmark": "frontierscience-olympiad",
|
| 326 |
+
"model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 327 |
+
"execution_mode": "non_agent",
|
| 328 |
+
"generation_hash": "093a6888a7b2df2153f1c00b78389a167c22fd3a4c039f79240683d24a27ed5e",
|
| 329 |
+
"scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016",
|
| 330 |
+
"selected": 3000,
|
| 331 |
+
"incomplete": 0,
|
| 332 |
+
"generated_now": 2841,
|
| 333 |
+
"scored_now": 2841,
|
| 334 |
+
"cached_completed": 159,
|
| 335 |
+
"stage": "all",
|
| 336 |
+
"worker_count": 1,
|
| 337 |
+
"worker_index": 0,
|
| 338 |
+
"worker_owned_predictions": 3000,
|
| 339 |
+
"generation_concurrency": 512,
|
| 340 |
+
"judge_concurrency": 256,
|
| 341 |
+
"updated_at": "2026-10-29T06:45:00.048121+00:00",
|
| 342 |
+
"provisional": false,
|
| 343 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"
|
| 344 |
+
},
|
| 345 |
+
"frontierscience-research": {
|
| 346 |
+
"total_records": 1800,
|
| 347 |
+
"completed": 1800,
|
| 348 |
+
"query_errors": 0,
|
| 349 |
+
"score_errors": 0,
|
| 350 |
+
"generated_unscored": 0,
|
| 351 |
+
"mean_score": 0.29168638888888887,
|
| 352 |
+
"accuracy": 0.07166666666666667,
|
| 353 |
+
"mean_latency_seconds": 549.7299430483823,
|
| 354 |
+
"finish_reason_counts": {
|
| 355 |
+
"length": 79,
|
| 356 |
+
"stop": 1721
|
| 357 |
+
},
|
| 358 |
+
"length_limited_response_count": 79,
|
| 359 |
+
"observed_max_completion_tokens": 98304,
|
| 360 |
+
"judge_usage": {
|
| 361 |
+
"judged_record_count": 1800,
|
| 362 |
+
"request_count": 1832,
|
| 363 |
+
"retry_request_count": 32,
|
| 364 |
+
"prompt_tokens": 11125895,
|
| 365 |
+
"cached_prompt_tokens": 246486,
|
| 366 |
+
"cache_write_prompt_tokens": 10836821,
|
| 367 |
+
"completion_tokens": 3061669,
|
| 368 |
+
"reasoning_tokens": 1817948,
|
| 369 |
+
"total_tokens": 14187564
|
| 370 |
+
},
|
| 371 |
+
"judge_api_usage": {
|
| 372 |
+
"request_count": 1832,
|
| 373 |
+
"prompt_tokens": 11125895,
|
| 374 |
+
"cached_prompt_tokens": 246486,
|
| 375 |
+
"cache_write_prompt_tokens": 10836821,
|
| 376 |
+
"completion_tokens": 3061669,
|
| 377 |
+
"reasoning_tokens": 1817948,
|
| 378 |
+
"total_tokens": 14187564
|
| 379 |
+
},
|
| 380 |
+
"benchmark_metrics": {
|
| 381 |
+
"track": "research",
|
| 382 |
+
"n_problems_scored": 60,
|
| 383 |
+
"n_trajectories_scored": 1800,
|
| 384 |
+
"aggregation": "avg@30",
|
| 385 |
+
"reported_accuracy": 0.07166666666666667,
|
| 386 |
+
"mean_rubric_score": 0.29168638888888887,
|
| 387 |
+
"success_threshold": 0.7,
|
| 388 |
+
"community_profile": null
|
| 389 |
+
},
|
| 390 |
+
"benchmark": "frontierscience-research",
|
| 391 |
+
"model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@a9904d24bcc1d289a1950fa9d2b978c47cf903b9",
|
| 392 |
+
"execution_mode": "non_agent",
|
| 393 |
+
"generation_hash": "3749020dcb028aa6fd38943de4c92c32672113762d2745830e10358371eea801",
|
| 394 |
+
"scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750",
|
| 395 |
+
"selected": 1800,
|
| 396 |
+
"incomplete": 0,
|
| 397 |
+
"generated_now": 1561,
|
| 398 |
+
"scored_now": 1564,
|
| 399 |
+
"cached_completed": 236,
|
| 400 |
+
"stage": "all",
|
| 401 |
+
"worker_count": 1,
|
| 402 |
+
"worker_index": 0,
|
| 403 |
+
"worker_owned_predictions": 1800,
|
| 404 |
+
"generation_concurrency": 512,
|
| 405 |
+
"judge_concurrency": 256,
|
| 406 |
+
"updated_at": "2026-10-29T06:34:29.975558+00:00",
|
| 407 |
+
"provisional": false,
|
| 408 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"
|
| 409 |
+
},
|
| 410 |
+
"scicode-verified": {
|
| 411 |
+
"updated_at": 1793255967.5510898,
|
| 412 |
+
"expected_problems": 64,
|
| 413 |
+
"expected_steps": 287,
|
| 414 |
+
"generated_steps": 287,
|
| 415 |
+
"scored_steps": 287,
|
| 416 |
+
"passed_steps": 155,
|
| 417 |
+
"completed_problems": 64,
|
| 418 |
+
"passed_problems": 21,
|
| 419 |
+
"subproblem_accuracy": 0.5400696864111498,
|
| 420 |
+
"main_problem_accuracy": 0.328125,
|
| 421 |
+
"complete": true,
|
| 422 |
+
"cap_hits": 0,
|
| 423 |
+
"mean_output_tokens": 9008.062717770035,
|
| 424 |
+
"grading_infrastructure_errors": 0,
|
| 425 |
+
"errors": [],
|
| 426 |
+
"provisional": false,
|
| 427 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"
|
| 428 |
+
},
|
| 429 |
+
"scicoding-agent": {
|
| 430 |
+
"tasks": 50,
|
| 431 |
+
"attempts": 100,
|
| 432 |
+
"attempts_per_task": 2,
|
| 433 |
+
"metric": "avg@2",
|
| 434 |
+
"passed": 26,
|
| 435 |
+
"avg_at_2": 0.26,
|
| 436 |
+
"unscored": 0,
|
| 437 |
+
"agent_timeouts": 2,
|
| 438 |
+
"reward_positive_with_exception": 1,
|
| 439 |
+
"scoring": "native_verifier",
|
| 440 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 441 |
+
"provisional": false,
|
| 442 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run"
|
| 443 |
+
},
|
| 444 |
+
"terminal-bench-2.1": {
|
| 445 |
+
"tasks": 89,
|
| 446 |
+
"attempts": 178,
|
| 447 |
+
"attempts_per_task": 2,
|
| 448 |
+
"metric": "avg@2",
|
| 449 |
+
"passed": 43,
|
| 450 |
+
"avg_at_2": 0.24157303370786518,
|
| 451 |
+
"unscored": 0,
|
| 452 |
+
"agent_timeouts": 58,
|
| 453 |
+
"reward_positive_with_exception": 3,
|
| 454 |
+
"scoring": "native_verifier",
|
| 455 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 456 |
+
"provisional": false,
|
| 457 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/run",
|
| 458 |
+
"scoring_adjudications": {
|
| 459 |
+
"mailman--attempt-002": {
|
| 460 |
+
"timestamp": "2026-10-29T19:11:17.460325+00:00",
|
| 461 |
+
"reward": 0.0,
|
| 462 |
+
"native_reward": null,
|
| 463 |
+
"original_verifier_result": null,
|
| 464 |
+
"method": "explicit_user_authorized_zero_for_unscored_attempt",
|
| 465 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 466 |
+
"reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.",
|
| 467 |
+
"exception_type": "VerifierTimeoutError",
|
| 468 |
+
"backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--mailman--attempt-002",
|
| 469 |
+
"original_sha256": "ef5726e0fa911c30c412c7f7cbe08181541c85351260634745e65e0d147b7ab4",
|
| 470 |
+
"generation_or_verifier_rerun": false
|
| 471 |
+
},
|
| 472 |
+
"pytorch-model-recovery--attempt-001": {
|
| 473 |
+
"timestamp": "2026-10-29T19:11:17.511954+00:00",
|
| 474 |
+
"reward": 0.0,
|
| 475 |
+
"native_reward": null,
|
| 476 |
+
"original_verifier_result": null,
|
| 477 |
+
"method": "explicit_user_authorized_zero_for_unscored_attempt",
|
| 478 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 479 |
+
"reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.",
|
| 480 |
+
"exception_type": "AgentTimeoutError",
|
| 481 |
+
"backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--pytorch-model-recovery--attempt-001",
|
| 482 |
+
"original_sha256": "68430013dc4d101ddd43f9dbacb6a17c5b772d507b8301ef3b18c6f30904340c",
|
| 483 |
+
"generation_or_verifier_rerun": false
|
| 484 |
+
},
|
| 485 |
+
"pytorch-model-recovery--attempt-002": {
|
| 486 |
+
"timestamp": "2026-10-29T19:11:17.565699+00:00",
|
| 487 |
+
"reward": 0.0,
|
| 488 |
+
"native_reward": null,
|
| 489 |
+
"original_verifier_result": null,
|
| 490 |
+
"method": "explicit_user_authorized_zero_for_unscored_attempt",
|
| 491 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 492 |
+
"reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.",
|
| 493 |
+
"exception_type": "AgentTimeoutError",
|
| 494 |
+
"backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--pytorch-model-recovery--attempt-002",
|
| 495 |
+
"original_sha256": "7fae00d6c6402beb1764a220d36240d2234c301adbcb73a2386afc5c69e971a5",
|
| 496 |
+
"generation_or_verifier_rerun": false
|
| 497 |
+
},
|
| 498 |
+
"torch-pipeline-parallelism--attempt-002": {
|
| 499 |
+
"timestamp": "2026-10-29T19:11:17.623946+00:00",
|
| 500 |
+
"reward": 0.0,
|
| 501 |
+
"native_reward": null,
|
| 502 |
+
"original_verifier_result": null,
|
| 503 |
+
"method": "explicit_user_authorized_zero_for_unscored_attempt",
|
| 504 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 505 |
+
"reason": "Original attempt ended without a native verifier reward; this is an adjudicated zero, not a completed native test verdict.",
|
| 506 |
+
"exception_type": "VerifierTimeoutError",
|
| 507 |
+
"backup": "/workspace/data/sci-llm-training/runtime/nemotron35-lightning-full-suite-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--torch-pipeline-parallelism--attempt-002",
|
| 508 |
+
"original_sha256": "67322cae95590fb5ba48bdf823a06f30d0861c161d2cd86c15adfad997ee3ac7",
|
| 509 |
+
"generation_or_verifier_rerun": false
|
| 510 |
+
}
|
| 511 |
+
},
|
| 512 |
+
"native_scored": 174,
|
| 513 |
+
"adjudicated_zero": 4,
|
| 514 |
+
"scoring_note": "Four missing native verdicts assigned zero by explicit user instruction; original exceptions and null rewards preserved in native_artifacts.tar.gz and recovery_provenance.tar.gz."
|
| 515 |
+
},
|
| 516 |
+
"hle-verified": {
|
| 517 |
+
"gold": {
|
| 518 |
+
"correct": 134,
|
| 519 |
+
"total": 575,
|
| 520 |
+
"accuracy": 0.23304347826086957,
|
| 521 |
+
"cap_hits": 57
|
| 522 |
+
},
|
| 523 |
+
"revision": {
|
| 524 |
+
"correct": 292,
|
| 525 |
+
"total": 1025,
|
| 526 |
+
"accuracy": 0.2848780487804878,
|
| 527 |
+
"cap_hits": 95
|
| 528 |
+
},
|
| 529 |
+
"correct": 426,
|
| 530 |
+
"total": 1600,
|
| 531 |
+
"overall_accuracy": 0.26625,
|
| 532 |
+
"all_complete": true,
|
| 533 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 534 |
+
"provenance": "hle-verified-provenance.json"
|
| 535 |
+
}
|
| 536 |
+
}
|
Olmo-3-32B-Think/README.md
CHANGED
|
@@ -22,3 +22,17 @@ Research mean rubric and trajectory success are separate metrics. Agentic avg@2
|
|
| 22 |
| frontierscience-olympiad | avg@30 | 28.03% | 3000/3000 scored |
|
| 23 |
| frontierscience-research | Mean rubric (scored trajectories) | 19.72% | 1800/1800 scored |
|
| 24 |
| scicode-verified | Main-problem accuracy | 25.00% | 64/64 problems |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
| frontierscience-olympiad | avg@30 | 28.03% | 3000/3000 scored |
|
| 23 |
| frontierscience-research | Mean rubric (scored trajectories) | 19.72% | 1800/1800 scored |
|
| 24 |
| scicode-verified | Main-problem accuracy | 25.00% | 64/64 problems |
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
## HLE-Verified (2026-10-30)
|
| 28 |
+
|
| 29 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **289/1,600 (18.06%)**.
|
| 30 |
+
|
| 31 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 32 |
+
|---|---:|---:|---|
|
| 33 |
+
| Gold | 69/575 | 12.00% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 34 |
+
| Revision | 220/1025 | 21.46% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 35 |
+
|
| 36 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 61,440 tokens; context 65,536 tokens. Output allowance is reduced for long prompts to fit the native context; prompts are not truncated. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 37 |
+
|
| 38 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
Olmo-3-32B-Think/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,396 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "olmo3-32b-think",
|
| 7 |
+
"folder": "Olmo-3-32B-Think",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 69,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.12,
|
| 13 |
+
"cap_hits": 0
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 220,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.2146341463414634,
|
| 19 |
+
"cap_hits": 0
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 289,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.180625
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "olmo3-32b-think",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "olmo3-32b-think",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "olmo",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "olmo3",
|
| 35 |
+
"hidden_size": 5120,
|
| 36 |
+
"num_hidden_layers": 64,
|
| 37 |
+
"vocab_size": 100278
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 61440,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"top_k": -1,
|
| 46 |
+
"min_p": 0.0,
|
| 47 |
+
"presence_penalty": 0.0,
|
| 48 |
+
"repetition_penalty": 1.0,
|
| 49 |
+
"_max_total_tokens": 65536
|
| 50 |
+
}
|
| 51 |
+
},
|
| 52 |
+
"serving": {
|
| 53 |
+
"context_length": 65536,
|
| 54 |
+
"dtype": "bfloat16",
|
| 55 |
+
"attention_backend": "fa3",
|
| 56 |
+
"mem_fraction_static": 0.85,
|
| 57 |
+
"reasoning_parser": "deepseek-r1",
|
| 58 |
+
"extra_args": [
|
| 59 |
+
"--cuda-graph-max-bs-decode",
|
| 60 |
+
"32"
|
| 61 |
+
],
|
| 62 |
+
"tp": 2,
|
| 63 |
+
"replicas_per_node": 4,
|
| 64 |
+
"max_running_requests": 32
|
| 65 |
+
},
|
| 66 |
+
"reasoning": {
|
| 67 |
+
"requests_per_engine": 32,
|
| 68 |
+
"benchmarks": {}
|
| 69 |
+
},
|
| 70 |
+
"agentic_tasks_per_engine": 1
|
| 71 |
+
},
|
| 72 |
+
"node": 3,
|
| 73 |
+
"checkpoint": "/workspace/data/raw_model/Olmo-3-32B-Think-f2edda15216e",
|
| 74 |
+
"repo": "allenai/Olmo-3-32B-Think",
|
| 75 |
+
"revision": "f2edda15216e738ef2bb73771e11890e152b2112",
|
| 76 |
+
"checkpoint_config_sha256": "d47a120cb2669cd46482d5c2da9488903b67759ca167cd85c1127b2e05755f95",
|
| 77 |
+
"index_sha256": "3f639e0b434e75af1bc14942b8ad0e4ab100923f2ca071c6985208d50293a269",
|
| 78 |
+
"shard_sizes": {
|
| 79 |
+
"model-00001-of-00014.safetensors": 4990600552,
|
| 80 |
+
"model-00002-of-00014.safetensors": 4938975944,
|
| 81 |
+
"model-00003-of-00014.safetensors": 4876048688,
|
| 82 |
+
"model-00004-of-00014.safetensors": 4876048704,
|
| 83 |
+
"model-00005-of-00014.safetensors": 4876048704,
|
| 84 |
+
"model-00006-of-00014.safetensors": 4876048704,
|
| 85 |
+
"model-00007-of-00014.safetensors": 4876048704,
|
| 86 |
+
"model-00008-of-00014.safetensors": 4876048704,
|
| 87 |
+
"model-00009-of-00014.safetensors": 4876048704,
|
| 88 |
+
"model-00010-of-00014.safetensors": 4876048704,
|
| 89 |
+
"model-00011-of-00014.safetensors": 4876048704,
|
| 90 |
+
"model-00012-of-00014.safetensors": 4876048704,
|
| 91 |
+
"model-00013-of-00014.safetensors": 4750216928,
|
| 92 |
+
"model-00014-of-00014.safetensors": 1026846848
|
| 93 |
+
},
|
| 94 |
+
"endpoints": [
|
| 95 |
+
{
|
| 96 |
+
"id": "node-3-replica-0",
|
| 97 |
+
"api_base": "http://10.0.154.99:8700/v1"
|
| 98 |
+
},
|
| 99 |
+
{
|
| 100 |
+
"id": "node-3-replica-1",
|
| 101 |
+
"api_base": "http://10.0.154.99:8701/v1"
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"id": "node-3-replica-2",
|
| 105 |
+
"api_base": "http://10.0.154.99:8702/v1"
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"id": "node-3-replica-3",
|
| 109 |
+
"api_base": "http://10.0.154.99:8703/v1"
|
| 110 |
+
}
|
| 111 |
+
],
|
| 112 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/olmo3-32b-think",
|
| 113 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/olmo3-32b-think/configs/hle-verified-gold-revision-96k.yaml",
|
| 114 |
+
"concurrency": 128,
|
| 115 |
+
"judge_concurrency": 64,
|
| 116 |
+
"tracking": {
|
| 117 |
+
"enabled": false,
|
| 118 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 119 |
+
},
|
| 120 |
+
"config_sha256": {
|
| 121 |
+
"model.yaml": "8ce9f0fbf7af64b5c1367f60a31ecceb4573cffaaaabadb341fac66e4f65ef7b",
|
| 122 |
+
"configs/hle-verified-gold-revision-96k.yaml": "1cb6a3f9c82920a2b19d4b3eb8da385fb30a494981bbbe357d0d1417b34c3275",
|
| 123 |
+
"configs/hle-verified-gold.yaml": "d9d0b773707980d1259e5b4500a454e3da651a47132e1bbb52c28661f804221d",
|
| 124 |
+
"configs/hle-verified-revision.yaml": "b3f43ef4b948a0a719e1914755df24ad4164c26e312c86603b7efaab1bac36da"
|
| 125 |
+
}
|
| 126 |
+
},
|
| 127 |
+
"dataset": {
|
| 128 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 129 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 130 |
+
"counts": {
|
| 131 |
+
"Gold": {
|
| 132 |
+
"total": 668,
|
| 133 |
+
"text": 575,
|
| 134 |
+
"image_excluded": 93
|
| 135 |
+
},
|
| 136 |
+
"Revision": {
|
| 137 |
+
"total": 1143,
|
| 138 |
+
"text": 1025,
|
| 139 |
+
"image_excluded": 118
|
| 140 |
+
}
|
| 141 |
+
},
|
| 142 |
+
"sha256": {
|
| 143 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 144 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 145 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 146 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 147 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 148 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 149 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 150 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 151 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 152 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 153 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 154 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 155 |
+
}
|
| 156 |
+
},
|
| 157 |
+
"token_preflight": {
|
| 158 |
+
"model": "olmo3-32b-think",
|
| 159 |
+
"questions": 1600,
|
| 160 |
+
"max_prompt_tokens": 13591,
|
| 161 |
+
"output_cap": 61440,
|
| 162 |
+
"context": 65536,
|
| 163 |
+
"dynamic_output_budget": true,
|
| 164 |
+
"affected_questions": 5,
|
| 165 |
+
"min_effective_output_cap": 51945
|
| 166 |
+
},
|
| 167 |
+
"shared_recovery_provenance": {
|
| 168 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 169 |
+
"publication_reruns": false,
|
| 170 |
+
"all_questions_scored": true,
|
| 171 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 172 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 173 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 174 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 175 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 176 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 177 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 178 |
+
},
|
| 179 |
+
"source_manifest": {
|
| 180 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 181 |
+
"files": {
|
| 182 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 183 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 184 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 185 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 186 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 187 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 188 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 189 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 190 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 191 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 192 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 193 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 194 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 195 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 196 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 197 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 198 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 199 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 200 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 201 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 202 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 203 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 204 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 205 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 206 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 207 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 208 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 209 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 210 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 211 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 212 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 213 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 214 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 215 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 216 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 217 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 218 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 219 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 220 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 221 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 222 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 223 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 224 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 225 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 226 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 227 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 228 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 229 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 230 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 231 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 232 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 233 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 234 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 235 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 236 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 237 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 238 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 239 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 240 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 241 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 242 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 243 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 244 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 245 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 246 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 247 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 248 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 249 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 250 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 251 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 252 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 253 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 254 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 255 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 256 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 257 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 258 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 259 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 260 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 261 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 262 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 263 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 264 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 265 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 266 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 267 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 268 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 269 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 270 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 271 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 272 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 273 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 274 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 275 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 276 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 277 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 278 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 279 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 280 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 281 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 282 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 283 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 284 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 285 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 286 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 287 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 288 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 289 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 290 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 291 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 292 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 293 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 294 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 295 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 296 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 297 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 298 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 299 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 300 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 301 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 302 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 303 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 304 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 305 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 306 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 307 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 308 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 309 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 310 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 311 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 312 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 313 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 314 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 315 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 316 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 317 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 318 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 319 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 320 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 321 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 322 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 323 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 324 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 325 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 326 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 327 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 328 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 329 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 330 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 331 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 332 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 333 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 334 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 335 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 336 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 337 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 338 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 339 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 340 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 341 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 342 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 343 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 344 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 345 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 346 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 347 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 348 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 349 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 350 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 351 |
+
},
|
| 352 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 353 |
+
},
|
| 354 |
+
"artifact_manifest": {
|
| 355 |
+
"Olmo-3-32B-Think/hle-verified-gold/configs.json": {
|
| 356 |
+
"size": 96624,
|
| 357 |
+
"sha256": "83cf37950d455c9c52a1a69e4f0e772602083e0257d6f04b7d6f456698391bd0"
|
| 358 |
+
},
|
| 359 |
+
"Olmo-3-32B-Think/hle-verified-gold/events.jsonl.gz": {
|
| 360 |
+
"size": 20232761,
|
| 361 |
+
"sha256": "7e32e5885184e09eab0c7f96a15dc3fa3ff5ea9a0da174ed279f7cd0f3a1dee6"
|
| 362 |
+
},
|
| 363 |
+
"Olmo-3-32B-Think/hle-verified-gold/judge_trajectories.jsonl": {
|
| 364 |
+
"size": 802265,
|
| 365 |
+
"sha256": "b04d68535d47e5c4746fb2320355f3daeb86c5eb078063bfb1334ace5b2542e9"
|
| 366 |
+
},
|
| 367 |
+
"Olmo-3-32B-Think/hle-verified-gold/summary.json": {
|
| 368 |
+
"size": 1440,
|
| 369 |
+
"sha256": "c4d000570412503eaa8c8e3e7cca4e285209e73a1a3916d03768c917f3a25dd6"
|
| 370 |
+
},
|
| 371 |
+
"Olmo-3-32B-Think/hle-verified-gold/trajectories.jsonl": {
|
| 372 |
+
"size": 37719102,
|
| 373 |
+
"sha256": "21e99af8b23fc0814a7b79aacb5ab4c8ff6b19f051e28ae09ca1f6846a6a4937"
|
| 374 |
+
},
|
| 375 |
+
"Olmo-3-32B-Think/hle-verified-revision/configs.json": {
|
| 376 |
+
"size": 165056,
|
| 377 |
+
"sha256": "db5e7b7d094dbffde70da87a3bec9e21dfe83a2791898de9bb6b11723ed24e9e"
|
| 378 |
+
},
|
| 379 |
+
"Olmo-3-32B-Think/hle-verified-revision/events.jsonl.gz": {
|
| 380 |
+
"size": 37941813,
|
| 381 |
+
"sha256": "b7b3aaca63e47ad81212f580ebcdd253d26c022fff10fdb0ef6dd2c0ffc523c8"
|
| 382 |
+
},
|
| 383 |
+
"Olmo-3-32B-Think/hle-verified-revision/judge_trajectories.jsonl": {
|
| 384 |
+
"size": 1446837,
|
| 385 |
+
"sha256": "1fe962a321092f2a6a162be42ac90905b68f2dccea08f3a0e53ddab220d193ee"
|
| 386 |
+
},
|
| 387 |
+
"Olmo-3-32B-Think/hle-verified-revision/summary.json": {
|
| 388 |
+
"size": 1497,
|
| 389 |
+
"sha256": "46fb3a59e8f35fb903dfb0ee2d572838fdea1f7f08692f0f69f08440dfa234e7"
|
| 390 |
+
},
|
| 391 |
+
"Olmo-3-32B-Think/hle-verified-revision/trajectories.jsonl": {
|
| 392 |
+
"size": 68910479,
|
| 393 |
+
"sha256": "57b1f5dbdfd8e8c35db47bedcb63d93d36075530d2bcf336cfb9ee203e98416f"
|
| 394 |
+
}
|
| 395 |
+
}
|
| 396 |
+
}
|
Olmo-3-32B-Think/summary.json
CHANGED
|
@@ -1 +1,400 @@
|
|
| 1 |
-
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"arxivmath": {
|
| 3 |
+
"benchmark": "arxivmath",
|
| 4 |
+
"releases": [
|
| 5 |
+
{
|
| 6 |
+
"total_records": 68,
|
| 7 |
+
"completed": 68,
|
| 8 |
+
"query_errors": 0,
|
| 9 |
+
"score_errors": 0,
|
| 10 |
+
"generated_unscored": 0,
|
| 11 |
+
"mean_score": 0.29411764705882354,
|
| 12 |
+
"accuracy": 0.29411764705882354,
|
| 13 |
+
"mean_latency_seconds": 728.9124951505561,
|
| 14 |
+
"finish_reason_counts": {
|
| 15 |
+
"stop": 68
|
| 16 |
+
},
|
| 17 |
+
"length_limited_response_count": 0,
|
| 18 |
+
"observed_max_completion_tokens": 44361,
|
| 19 |
+
"judge_usage": {
|
| 20 |
+
"judged_record_count": 0,
|
| 21 |
+
"request_count": 0,
|
| 22 |
+
"retry_request_count": 0,
|
| 23 |
+
"prompt_tokens": 0,
|
| 24 |
+
"cached_prompt_tokens": 0,
|
| 25 |
+
"cache_write_prompt_tokens": 0,
|
| 26 |
+
"completion_tokens": 0,
|
| 27 |
+
"reasoning_tokens": 0,
|
| 28 |
+
"total_tokens": 0
|
| 29 |
+
},
|
| 30 |
+
"benchmark_metrics": {
|
| 31 |
+
"competition": "arxivmath",
|
| 32 |
+
"aggregation": "avg@4",
|
| 33 |
+
"n_problems_scored": 17,
|
| 34 |
+
"n_trajectories_scored": 68,
|
| 35 |
+
"reported_accuracy": 0.29411764705882354,
|
| 36 |
+
"competition_metrics": {
|
| 37 |
+
"arxiv/december": {
|
| 38 |
+
"n_problems_scored": 17,
|
| 39 |
+
"n_trajectories_scored": 68,
|
| 40 |
+
"reported_accuracy": 0.29411764705882354
|
| 41 |
+
}
|
| 42 |
+
}
|
| 43 |
+
},
|
| 44 |
+
"benchmark": "matharena-arxivmath-december",
|
| 45 |
+
"model": "allenai/Olmo-3-32B-Think@f2edda15216e738ef2bb73771e11890e152b2112",
|
| 46 |
+
"execution_mode": "non_agent",
|
| 47 |
+
"generation_hash": "72274964f292792296b83f730cdd8b300c82da4440725ba99fcbef1330babd4c",
|
| 48 |
+
"scoring_hash": "c88da5c711f18d54ee3696adb31a301556f46896bc882437d15acc1fefe05048",
|
| 49 |
+
"selected": 68,
|
| 50 |
+
"incomplete": 0,
|
| 51 |
+
"generated_now": 68,
|
| 52 |
+
"scored_now": 68,
|
| 53 |
+
"cached_completed": 0,
|
| 54 |
+
"stage": "all",
|
| 55 |
+
"worker_count": 1,
|
| 56 |
+
"worker_index": 0,
|
| 57 |
+
"worker_owned_predictions": 68,
|
| 58 |
+
"generation_concurrency": 256,
|
| 59 |
+
"judge_concurrency": 256,
|
| 60 |
+
"updated_at": "2026-10-30T10:20:28.273085+00:00"
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"total_records": 128,
|
| 64 |
+
"completed": 128,
|
| 65 |
+
"query_errors": 0,
|
| 66 |
+
"score_errors": 0,
|
| 67 |
+
"generated_unscored": 0,
|
| 68 |
+
"mean_score": 0.1640625,
|
| 69 |
+
"accuracy": 0.1640625,
|
| 70 |
+
"mean_latency_seconds": 648.4357359755604,
|
| 71 |
+
"finish_reason_counts": {
|
| 72 |
+
"stop": 128
|
| 73 |
+
},
|
| 74 |
+
"length_limited_response_count": 0,
|
| 75 |
+
"observed_max_completion_tokens": 60615,
|
| 76 |
+
"judge_usage": {
|
| 77 |
+
"judged_record_count": 0,
|
| 78 |
+
"request_count": 0,
|
| 79 |
+
"retry_request_count": 0,
|
| 80 |
+
"prompt_tokens": 0,
|
| 81 |
+
"cached_prompt_tokens": 0,
|
| 82 |
+
"cache_write_prompt_tokens": 0,
|
| 83 |
+
"completion_tokens": 0,
|
| 84 |
+
"reasoning_tokens": 0,
|
| 85 |
+
"total_tokens": 0
|
| 86 |
+
},
|
| 87 |
+
"benchmark_metrics": {
|
| 88 |
+
"competition": "arxivmath",
|
| 89 |
+
"aggregation": "avg@4",
|
| 90 |
+
"n_problems_scored": 32,
|
| 91 |
+
"n_trajectories_scored": 128,
|
| 92 |
+
"reported_accuracy": 0.1640625,
|
| 93 |
+
"competition_metrics": {
|
| 94 |
+
"arxiv/february": {
|
| 95 |
+
"n_problems_scored": 32,
|
| 96 |
+
"n_trajectories_scored": 128,
|
| 97 |
+
"reported_accuracy": 0.1640625
|
| 98 |
+
}
|
| 99 |
+
}
|
| 100 |
+
},
|
| 101 |
+
"benchmark": "matharena-arxivmath-february",
|
| 102 |
+
"model": "allenai/Olmo-3-32B-Think@f2edda15216e738ef2bb73771e11890e152b2112",
|
| 103 |
+
"execution_mode": "non_agent",
|
| 104 |
+
"generation_hash": "03b1765e135a3558975afc438ab526f3f651d3650963c1a3d8d7455ae217aa49",
|
| 105 |
+
"scoring_hash": "ddffb44da8eb67bcb9d81749c4bb55324fdf1b95fe20752e334a57d2c83dcefd",
|
| 106 |
+
"selected": 128,
|
| 107 |
+
"incomplete": 0,
|
| 108 |
+
"generated_now": 128,
|
| 109 |
+
"scored_now": 128,
|
| 110 |
+
"cached_completed": 0,
|
| 111 |
+
"stage": "all",
|
| 112 |
+
"worker_count": 1,
|
| 113 |
+
"worker_index": 0,
|
| 114 |
+
"worker_owned_predictions": 128,
|
| 115 |
+
"generation_concurrency": 256,
|
| 116 |
+
"judge_concurrency": 256,
|
| 117 |
+
"updated_at": "2026-10-30T10:16:01.546758+00:00"
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"total_records": 92,
|
| 121 |
+
"completed": 92,
|
| 122 |
+
"query_errors": 0,
|
| 123 |
+
"score_errors": 0,
|
| 124 |
+
"generated_unscored": 0,
|
| 125 |
+
"mean_score": 0.20652173913043478,
|
| 126 |
+
"accuracy": 0.20652173913043478,
|
| 127 |
+
"mean_latency_seconds": 818.9604985704949,
|
| 128 |
+
"finish_reason_counts": {
|
| 129 |
+
"stop": 92
|
| 130 |
+
},
|
| 131 |
+
"length_limited_response_count": 0,
|
| 132 |
+
"observed_max_completion_tokens": 45130,
|
| 133 |
+
"judge_usage": {
|
| 134 |
+
"judged_record_count": 0,
|
| 135 |
+
"request_count": 0,
|
| 136 |
+
"retry_request_count": 0,
|
| 137 |
+
"prompt_tokens": 0,
|
| 138 |
+
"cached_prompt_tokens": 0,
|
| 139 |
+
"cache_write_prompt_tokens": 0,
|
| 140 |
+
"completion_tokens": 0,
|
| 141 |
+
"reasoning_tokens": 0,
|
| 142 |
+
"total_tokens": 0
|
| 143 |
+
},
|
| 144 |
+
"benchmark_metrics": {
|
| 145 |
+
"competition": "arxivmath",
|
| 146 |
+
"aggregation": "avg@4",
|
| 147 |
+
"n_problems_scored": 23,
|
| 148 |
+
"n_trajectories_scored": 92,
|
| 149 |
+
"reported_accuracy": 0.20652173913043478,
|
| 150 |
+
"competition_metrics": {
|
| 151 |
+
"arxiv/january": {
|
| 152 |
+
"n_problems_scored": 23,
|
| 153 |
+
"n_trajectories_scored": 92,
|
| 154 |
+
"reported_accuracy": 0.20652173913043478
|
| 155 |
+
}
|
| 156 |
+
}
|
| 157 |
+
},
|
| 158 |
+
"benchmark": "matharena-arxivmath-january",
|
| 159 |
+
"model": "allenai/Olmo-3-32B-Think@f2edda15216e738ef2bb73771e11890e152b2112",
|
| 160 |
+
"execution_mode": "non_agent",
|
| 161 |
+
"generation_hash": "2bc5b7ecc7074ece6c922ce98ec4da64ffff6c63afabf9fc59dc3a83090335d4",
|
| 162 |
+
"scoring_hash": "63d5e87d33e71105b239be11efa62d25454fccbfc3e0073d82b31e8e73712c4a",
|
| 163 |
+
"selected": 92,
|
| 164 |
+
"incomplete": 0,
|
| 165 |
+
"generated_now": 92,
|
| 166 |
+
"scored_now": 92,
|
| 167 |
+
"cached_completed": 0,
|
| 168 |
+
"stage": "all",
|
| 169 |
+
"worker_count": 1,
|
| 170 |
+
"worker_index": 0,
|
| 171 |
+
"worker_owned_predictions": 92,
|
| 172 |
+
"generation_concurrency": 256,
|
| 173 |
+
"judge_concurrency": 256,
|
| 174 |
+
"updated_at": "2026-10-30T10:21:39.581508+00:00"
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"total_records": 124,
|
| 178 |
+
"completed": 124,
|
| 179 |
+
"query_errors": 0,
|
| 180 |
+
"score_errors": 0,
|
| 181 |
+
"generated_unscored": 0,
|
| 182 |
+
"mean_score": 0.12096774193548387,
|
| 183 |
+
"accuracy": 0.12096774193548387,
|
| 184 |
+
"mean_latency_seconds": 675.0631301865637,
|
| 185 |
+
"finish_reason_counts": {
|
| 186 |
+
"stop": 124
|
| 187 |
+
},
|
| 188 |
+
"length_limited_response_count": 0,
|
| 189 |
+
"observed_max_completion_tokens": 39459,
|
| 190 |
+
"judge_usage": {
|
| 191 |
+
"judged_record_count": 0,
|
| 192 |
+
"request_count": 0,
|
| 193 |
+
"retry_request_count": 0,
|
| 194 |
+
"prompt_tokens": 0,
|
| 195 |
+
"cached_prompt_tokens": 0,
|
| 196 |
+
"cache_write_prompt_tokens": 0,
|
| 197 |
+
"completion_tokens": 0,
|
| 198 |
+
"reasoning_tokens": 0,
|
| 199 |
+
"total_tokens": 0
|
| 200 |
+
},
|
| 201 |
+
"benchmark_metrics": {
|
| 202 |
+
"competition": "arxivmath",
|
| 203 |
+
"aggregation": "avg@4",
|
| 204 |
+
"n_problems_scored": 31,
|
| 205 |
+
"n_trajectories_scored": 124,
|
| 206 |
+
"reported_accuracy": 0.12096774193548387,
|
| 207 |
+
"competition_metrics": {
|
| 208 |
+
"arxiv/march": {
|
| 209 |
+
"n_problems_scored": 31,
|
| 210 |
+
"n_trajectories_scored": 124,
|
| 211 |
+
"reported_accuracy": 0.12096774193548387
|
| 212 |
+
}
|
| 213 |
+
}
|
| 214 |
+
},
|
| 215 |
+
"benchmark": "matharena-arxivmath-march",
|
| 216 |
+
"model": "allenai/Olmo-3-32B-Think@f2edda15216e738ef2bb73771e11890e152b2112",
|
| 217 |
+
"execution_mode": "non_agent",
|
| 218 |
+
"generation_hash": "06b5a49e13919c0840105de1d832accb805300039da5caff80f2d8f4d39d3ef7",
|
| 219 |
+
"scoring_hash": "5f75d45c2d67e020a73cfe93539d7be0b0647c07fc20f13d1e73985c65300434",
|
| 220 |
+
"selected": 124,
|
| 221 |
+
"incomplete": 0,
|
| 222 |
+
"generated_now": 124,
|
| 223 |
+
"scored_now": 124,
|
| 224 |
+
"cached_completed": 0,
|
| 225 |
+
"stage": "all",
|
| 226 |
+
"worker_count": 1,
|
| 227 |
+
"worker_index": 0,
|
| 228 |
+
"worker_owned_predictions": 124,
|
| 229 |
+
"generation_concurrency": 256,
|
| 230 |
+
"judge_concurrency": 256,
|
| 231 |
+
"updated_at": "2026-10-30T10:17:19.880084+00:00"
|
| 232 |
+
}
|
| 233 |
+
],
|
| 234 |
+
"total_records": 412,
|
| 235 |
+
"completed": 412,
|
| 236 |
+
"accuracy": 0.1820388349514563
|
| 237 |
+
},
|
| 238 |
+
"frontierscience-olympiad": {
|
| 239 |
+
"total_records": 3000,
|
| 240 |
+
"completed": 3000,
|
| 241 |
+
"query_errors": 0,
|
| 242 |
+
"score_errors": 0,
|
| 243 |
+
"generated_unscored": 0,
|
| 244 |
+
"mean_score": 0.2803333333333333,
|
| 245 |
+
"accuracy": 0.2803333333333333,
|
| 246 |
+
"mean_latency_seconds": 655.6978648639127,
|
| 247 |
+
"finish_reason_counts": {
|
| 248 |
+
"length": 1,
|
| 249 |
+
"stop": 2999
|
| 250 |
+
},
|
| 251 |
+
"length_limited_response_count": 1,
|
| 252 |
+
"observed_max_completion_tokens": 61440,
|
| 253 |
+
"judge_usage": {
|
| 254 |
+
"judged_record_count": 3000,
|
| 255 |
+
"request_count": 3000,
|
| 256 |
+
"retry_request_count": 0,
|
| 257 |
+
"prompt_tokens": 2265649,
|
| 258 |
+
"cached_prompt_tokens": 20051,
|
| 259 |
+
"cache_write_prompt_tokens": 708179,
|
| 260 |
+
"completion_tokens": 613217,
|
| 261 |
+
"reasoning_tokens": 404654,
|
| 262 |
+
"total_tokens": 2878866
|
| 263 |
+
},
|
| 264 |
+
"judge_api_usage": {
|
| 265 |
+
"request_count": 3000,
|
| 266 |
+
"prompt_tokens": 2265649,
|
| 267 |
+
"cached_prompt_tokens": 20051,
|
| 268 |
+
"cache_write_prompt_tokens": 708179,
|
| 269 |
+
"completion_tokens": 613217,
|
| 270 |
+
"reasoning_tokens": 404654,
|
| 271 |
+
"total_tokens": 2878866
|
| 272 |
+
},
|
| 273 |
+
"benchmark_metrics": {
|
| 274 |
+
"track": "olympiad",
|
| 275 |
+
"n_problems_scored": 100,
|
| 276 |
+
"n_trajectories_scored": 3000,
|
| 277 |
+
"aggregation": "avg@30",
|
| 278 |
+
"reported_accuracy": 0.2803333333333333,
|
| 279 |
+
"mean_rubric_score": null,
|
| 280 |
+
"success_threshold": 0.5,
|
| 281 |
+
"community_profile": null
|
| 282 |
+
},
|
| 283 |
+
"benchmark": "frontierscience-olympiad",
|
| 284 |
+
"model": "allenai/Olmo-3-32B-Think@f2edda15216e738ef2bb73771e11890e152b2112",
|
| 285 |
+
"execution_mode": "non_agent",
|
| 286 |
+
"generation_hash": "0c0880e092358b72c657d804399d3512584c257767b52fc9a2585021e09366f9",
|
| 287 |
+
"scoring_hash": "8aaf11736acc031043aaf52a372c2cd3555a5713a801b76761d0f731b82cee99",
|
| 288 |
+
"selected": 3000,
|
| 289 |
+
"incomplete": 0,
|
| 290 |
+
"generated_now": 3000,
|
| 291 |
+
"scored_now": 3000,
|
| 292 |
+
"cached_completed": 0,
|
| 293 |
+
"stage": "all",
|
| 294 |
+
"worker_count": 1,
|
| 295 |
+
"worker_index": 0,
|
| 296 |
+
"worker_owned_predictions": 3000,
|
| 297 |
+
"generation_concurrency": 256,
|
| 298 |
+
"judge_concurrency": 256,
|
| 299 |
+
"updated_at": "2026-10-30T08:48:47.129999+00:00"
|
| 300 |
+
},
|
| 301 |
+
"frontierscience-research": {
|
| 302 |
+
"total_records": 1800,
|
| 303 |
+
"completed": 1800,
|
| 304 |
+
"query_errors": 0,
|
| 305 |
+
"score_errors": 0,
|
| 306 |
+
"generated_unscored": 0,
|
| 307 |
+
"mean_score": 0.19722499999999998,
|
| 308 |
+
"accuracy": 0.012777777777777779,
|
| 309 |
+
"mean_latency_seconds": 396.48057577082614,
|
| 310 |
+
"finish_reason_counts": {
|
| 311 |
+
"stop": 1800
|
| 312 |
+
},
|
| 313 |
+
"length_limited_response_count": 0,
|
| 314 |
+
"observed_max_completion_tokens": 51807,
|
| 315 |
+
"judge_usage": {
|
| 316 |
+
"judged_record_count": 1800,
|
| 317 |
+
"request_count": 1815,
|
| 318 |
+
"retry_request_count": 15,
|
| 319 |
+
"prompt_tokens": 6376558,
|
| 320 |
+
"cached_prompt_tokens": 49699,
|
| 321 |
+
"cache_write_prompt_tokens": 6321414,
|
| 322 |
+
"completion_tokens": 2575979,
|
| 323 |
+
"reasoning_tokens": 1404249,
|
| 324 |
+
"total_tokens": 8952537
|
| 325 |
+
},
|
| 326 |
+
"judge_api_usage": {
|
| 327 |
+
"request_count": 1815,
|
| 328 |
+
"prompt_tokens": 6376558,
|
| 329 |
+
"cached_prompt_tokens": 49699,
|
| 330 |
+
"cache_write_prompt_tokens": 6321414,
|
| 331 |
+
"completion_tokens": 2575979,
|
| 332 |
+
"reasoning_tokens": 1404249,
|
| 333 |
+
"total_tokens": 8952537
|
| 334 |
+
},
|
| 335 |
+
"benchmark_metrics": {
|
| 336 |
+
"track": "research",
|
| 337 |
+
"n_problems_scored": 60,
|
| 338 |
+
"n_trajectories_scored": 1800,
|
| 339 |
+
"aggregation": "avg@30",
|
| 340 |
+
"reported_accuracy": 0.012777777777777779,
|
| 341 |
+
"mean_rubric_score": 0.197225,
|
| 342 |
+
"success_threshold": 0.7,
|
| 343 |
+
"community_profile": null
|
| 344 |
+
},
|
| 345 |
+
"benchmark": "frontierscience-research",
|
| 346 |
+
"model": "allenai/Olmo-3-32B-Think@f2edda15216e738ef2bb73771e11890e152b2112",
|
| 347 |
+
"execution_mode": "non_agent",
|
| 348 |
+
"generation_hash": "fbf745a42c7809a4bada9a2bce0cd82cb46d45e4fc92bd4357442e77a0bed808",
|
| 349 |
+
"scoring_hash": "76e0e3879fd9bef8fc2c49664760f19a4ff5bca26c1f07144d26ff0b17448171",
|
| 350 |
+
"selected": 1800,
|
| 351 |
+
"incomplete": 0,
|
| 352 |
+
"generated_now": 1800,
|
| 353 |
+
"scored_now": 1800,
|
| 354 |
+
"cached_completed": 0,
|
| 355 |
+
"stage": "all",
|
| 356 |
+
"worker_count": 1,
|
| 357 |
+
"worker_index": 0,
|
| 358 |
+
"worker_owned_predictions": 1800,
|
| 359 |
+
"generation_concurrency": 256,
|
| 360 |
+
"judge_concurrency": 256,
|
| 361 |
+
"updated_at": "2026-10-30T08:19:56.925801+00:00"
|
| 362 |
+
},
|
| 363 |
+
"scicode-verified": {
|
| 364 |
+
"updated_at": 1793353944.742516,
|
| 365 |
+
"expected_problems": 64,
|
| 366 |
+
"expected_steps": 287,
|
| 367 |
+
"generated_steps": 287,
|
| 368 |
+
"scored_steps": 287,
|
| 369 |
+
"passed_steps": 142,
|
| 370 |
+
"completed_problems": 64,
|
| 371 |
+
"passed_problems": 16,
|
| 372 |
+
"subproblem_accuracy": 0.49477351916376305,
|
| 373 |
+
"main_problem_accuracy": 0.25,
|
| 374 |
+
"complete": true,
|
| 375 |
+
"cap_hits": 0,
|
| 376 |
+
"mean_output_tokens": 10676.445993031359,
|
| 377 |
+
"grading_infrastructure_errors": 0,
|
| 378 |
+
"errors": []
|
| 379 |
+
},
|
| 380 |
+
"hle-verified": {
|
| 381 |
+
"gold": {
|
| 382 |
+
"correct": 69,
|
| 383 |
+
"total": 575,
|
| 384 |
+
"accuracy": 0.12,
|
| 385 |
+
"cap_hits": 0
|
| 386 |
+
},
|
| 387 |
+
"revision": {
|
| 388 |
+
"correct": 220,
|
| 389 |
+
"total": 1025,
|
| 390 |
+
"accuracy": 0.2146341463414634,
|
| 391 |
+
"cap_hits": 0
|
| 392 |
+
},
|
| 393 |
+
"correct": 289,
|
| 394 |
+
"total": 1600,
|
| 395 |
+
"overall_accuracy": 0.180625,
|
| 396 |
+
"all_complete": true,
|
| 397 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 398 |
+
"provenance": "hle-verified-provenance.json"
|
| 399 |
+
}
|
| 400 |
+
}
|
Qwen3.5-35B-A3B/README.md
CHANGED
|
@@ -18,3 +18,17 @@ Agentic evaluation used Terminus 2, two independent attempts per task, temperatu
|
|
| 18 |
TB has 176 native scores and two **explicit user-adjudicated zeros**, for FEAL differential cryptanalysis attempt 2 and model-extraction-relu-logits attempt 2. These were originally unscored after agent timeouts. They are not represented as completed native test failures. Original null rewards and unchanged exception evidence are backed up in agentic_recovery_provenance.tar.gz; no model or verifier rerun was performed.
|
| 19 |
|
| 20 |
Prior non-agentic results are unchanged. Non-agentic generation used 98,304 output tokens, temperature 1.0 and top-p 1.0. FrontierScience uses GPT-5.6 Luna, arXiv Math deterministic grading, and SciCode native executable verification.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
TB has 176 native scores and two **explicit user-adjudicated zeros**, for FEAL differential cryptanalysis attempt 2 and model-extraction-relu-logits attempt 2. These were originally unscored after agent timeouts. They are not represented as completed native test failures. Original null rewards and unchanged exception evidence are backed up in agentic_recovery_provenance.tar.gz; no model or verifier rerun was performed.
|
| 19 |
|
| 20 |
Prior non-agentic results are unchanged. Non-agentic generation used 98,304 output tokens, temperature 1.0 and top-p 1.0. FrontierScience uses GPT-5.6 Luna, arXiv Math deterministic grading, and SciCode native executable verification.
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
## HLE-Verified (2026-10-30)
|
| 24 |
+
|
| 25 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **593/1,600 (37.06%)**.
|
| 26 |
+
|
| 27 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 28 |
+
|---|---:|---:|---|
|
| 29 |
+
| Gold | 228/575 | 39.65% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 30 |
+
| Revision | 365/1025 | 35.61% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 31 |
+
|
| 32 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 98,304 tokens; context 131,072 tokens. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 33 |
+
|
| 34 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
Qwen3.5-35B-A3B/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,400 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "qwen35-35b-a3b",
|
| 7 |
+
"folder": "Qwen3.5-35B-A3B",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 228,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.39652173913043476,
|
| 13 |
+
"cap_hits": 41
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 365,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.35609756097560974,
|
| 19 |
+
"cap_hits": 63
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 593,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.370625
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "qwen35-35b-a3b",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "qwen35-35b-a3b",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "qwen",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "qwen3_5_moe_text",
|
| 35 |
+
"hidden_size": 2048,
|
| 36 |
+
"num_hidden_layers": 40,
|
| 37 |
+
"vocab_size": 248320
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 98304,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"top_k": -1,
|
| 46 |
+
"min_p": 0.0,
|
| 47 |
+
"presence_penalty": 0.0,
|
| 48 |
+
"repetition_penalty": 1.0,
|
| 49 |
+
"chat_template_kwargs": {
|
| 50 |
+
"enable_thinking": true
|
| 51 |
+
}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"serving": {
|
| 55 |
+
"context_length": 131072,
|
| 56 |
+
"dtype": "bfloat16",
|
| 57 |
+
"attention_backend": "fa3",
|
| 58 |
+
"mem_fraction_static": 0.85,
|
| 59 |
+
"reasoning_parser": "qwen3",
|
| 60 |
+
"extra_args": [
|
| 61 |
+
"--cuda-graph-max-bs-decode",
|
| 62 |
+
"32",
|
| 63 |
+
"--tool-call-parser",
|
| 64 |
+
"qwen3_coder"
|
| 65 |
+
],
|
| 66 |
+
"tp": 2,
|
| 67 |
+
"replicas_per_node": 4,
|
| 68 |
+
"max_running_requests": 32
|
| 69 |
+
},
|
| 70 |
+
"reasoning": {
|
| 71 |
+
"requests_per_engine": 32,
|
| 72 |
+
"benchmarks": {}
|
| 73 |
+
},
|
| 74 |
+
"agentic_tasks_per_engine": 1
|
| 75 |
+
},
|
| 76 |
+
"node": 1,
|
| 77 |
+
"checkpoint": "/workspace/data/raw_model/Qwen3.5-35B-A3B-59d61f3ce65a",
|
| 78 |
+
"repo": "Qwen/Qwen3.5-35B-A3B",
|
| 79 |
+
"revision": "59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 80 |
+
"checkpoint_config_sha256": "5e4d7f74fec2f360eb9cfbfcd6ec0c4c76e684d3a11caaed259d9fd9bfbc7944",
|
| 81 |
+
"index_sha256": "d8d0b7ca4e61ae107e3e87a3ff21136b3ac7c789e64bb24267227ca804e04205",
|
| 82 |
+
"shard_sizes": {
|
| 83 |
+
"model.safetensors-00001-of-00014.safetensors": 5368709816,
|
| 84 |
+
"model.safetensors-00002-of-00014.safetensors": 5368709808,
|
| 85 |
+
"model.safetensors-00003-of-00014.safetensors": 5368709808,
|
| 86 |
+
"model.safetensors-00004-of-00014.safetensors": 5368709816,
|
| 87 |
+
"model.safetensors-00005-of-00014.safetensors": 5368709816,
|
| 88 |
+
"model.safetensors-00006-of-00014.safetensors": 5368709816,
|
| 89 |
+
"model.safetensors-00007-of-00014.safetensors": 5368709808,
|
| 90 |
+
"model.safetensors-00008-of-00014.safetensors": 5368709808,
|
| 91 |
+
"model.safetensors-00009-of-00014.safetensors": 5255463936,
|
| 92 |
+
"model.safetensors-00010-of-00014.safetensors": 5368710456,
|
| 93 |
+
"model.safetensors-00011-of-00014.safetensors": 5368710464,
|
| 94 |
+
"model.safetensors-00012-of-00014.safetensors": 5368710456,
|
| 95 |
+
"model.safetensors-00013-of-00014.safetensors": 5367839544,
|
| 96 |
+
"model.safetensors-00014-of-00014.safetensors": 2224764664
|
| 97 |
+
},
|
| 98 |
+
"endpoints": [
|
| 99 |
+
{
|
| 100 |
+
"id": "node-1-replica-0",
|
| 101 |
+
"api_base": "http://10.1.136.8:8700/v1"
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"id": "node-1-replica-1",
|
| 105 |
+
"api_base": "http://10.1.136.8:8701/v1"
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"id": "node-1-replica-2",
|
| 109 |
+
"api_base": "http://10.1.136.8:8702/v1"
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"id": "node-1-replica-3",
|
| 113 |
+
"api_base": "http://10.1.136.8:8703/v1"
|
| 114 |
+
}
|
| 115 |
+
],
|
| 116 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-35b-a3b",
|
| 117 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-35b-a3b/configs/hle-verified-gold-revision-96k.yaml",
|
| 118 |
+
"concurrency": 128,
|
| 119 |
+
"judge_concurrency": 64,
|
| 120 |
+
"tracking": {
|
| 121 |
+
"enabled": false,
|
| 122 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 123 |
+
},
|
| 124 |
+
"config_sha256": {
|
| 125 |
+
"model.yaml": "a073001a8ff3ed003b525f772dd452e44788a2c093caa48e0552a6ee3d6cba56",
|
| 126 |
+
"configs/hle-verified-gold-revision-96k.yaml": "8f3608cef1c230c7ccff9a29a1a0bef91d185d9dcc1b9771b114566e1f3efb56",
|
| 127 |
+
"configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4",
|
| 128 |
+
"configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"
|
| 129 |
+
}
|
| 130 |
+
},
|
| 131 |
+
"dataset": {
|
| 132 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 133 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 134 |
+
"counts": {
|
| 135 |
+
"Gold": {
|
| 136 |
+
"total": 668,
|
| 137 |
+
"text": 575,
|
| 138 |
+
"image_excluded": 93
|
| 139 |
+
},
|
| 140 |
+
"Revision": {
|
| 141 |
+
"total": 1143,
|
| 142 |
+
"text": 1025,
|
| 143 |
+
"image_excluded": 118
|
| 144 |
+
}
|
| 145 |
+
},
|
| 146 |
+
"sha256": {
|
| 147 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 148 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 149 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 150 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 151 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 152 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 153 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 154 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 155 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 156 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 157 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 158 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 159 |
+
}
|
| 160 |
+
},
|
| 161 |
+
"token_preflight": {
|
| 162 |
+
"model": "qwen35-35b-a3b",
|
| 163 |
+
"questions": 1600,
|
| 164 |
+
"max_prompt_tokens": 13586,
|
| 165 |
+
"output_cap": 98304,
|
| 166 |
+
"context": 131072,
|
| 167 |
+
"dynamic_output_budget": false,
|
| 168 |
+
"affected_questions": 0,
|
| 169 |
+
"min_effective_output_cap": 98304
|
| 170 |
+
},
|
| 171 |
+
"shared_recovery_provenance": {
|
| 172 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 173 |
+
"publication_reruns": false,
|
| 174 |
+
"all_questions_scored": true,
|
| 175 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 176 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 177 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 178 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 179 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 180 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 181 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 182 |
+
},
|
| 183 |
+
"source_manifest": {
|
| 184 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 185 |
+
"files": {
|
| 186 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 187 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 188 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 189 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 190 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 191 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 192 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 193 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 194 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 195 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 196 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 197 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 198 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 199 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 200 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 201 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 202 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 203 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 204 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 205 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 206 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 207 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 208 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 209 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 210 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 211 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 212 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 213 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 214 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 215 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 216 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 217 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 218 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 219 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 220 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 221 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 222 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 223 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 224 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 225 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 226 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 227 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 228 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 229 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 230 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 231 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 232 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 233 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 234 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 235 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 236 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 237 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 238 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 239 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 240 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 241 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 242 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 243 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 244 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 245 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 246 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 247 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 248 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 249 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 250 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 251 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 252 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 253 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 254 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 255 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 256 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 257 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 258 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 259 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 260 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 261 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 262 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 263 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 264 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 265 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 266 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 267 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 268 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 269 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 270 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 271 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 272 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 273 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 274 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 275 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 276 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 277 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 278 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 279 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 280 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 281 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 282 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 283 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 284 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 285 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 286 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 287 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 288 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 289 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 290 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 291 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 292 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 293 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 294 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 295 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 296 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 297 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 298 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 299 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 300 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 301 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 302 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 303 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 304 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 305 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 306 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 307 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 308 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 309 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 310 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 311 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 312 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 313 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 314 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 315 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 316 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 317 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 318 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 319 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 320 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 321 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 322 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 323 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 324 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 325 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 326 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 327 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 328 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 329 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 330 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 331 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 332 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 333 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 334 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 335 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 336 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 337 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 338 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 339 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 340 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 341 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 342 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 343 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 344 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 345 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 346 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 347 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 348 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 349 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 350 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 351 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 352 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 353 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 354 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 355 |
+
},
|
| 356 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 357 |
+
},
|
| 358 |
+
"artifact_manifest": {
|
| 359 |
+
"Qwen3.5-35B-A3B/hle-verified-gold/configs.json": {
|
| 360 |
+
"size": 97478,
|
| 361 |
+
"sha256": "64ed43cc300dbe1e6c6468936aaa7aaabe0b1381b595afea6ba66efa156a13d6"
|
| 362 |
+
},
|
| 363 |
+
"Qwen3.5-35B-A3B/hle-verified-gold/events.jsonl.gz": {
|
| 364 |
+
"size": 38963380,
|
| 365 |
+
"sha256": "06fb9b5c90ec966dffa5344062a15948bc212e97be536874a5976b71df21aaf0"
|
| 366 |
+
},
|
| 367 |
+
"Qwen3.5-35B-A3B/hle-verified-gold/judge_trajectories.jsonl": {
|
| 368 |
+
"size": 811549,
|
| 369 |
+
"sha256": "32a6043c2f63917e80e52af5e231545b3aed997bf342314730df3207f4f84846"
|
| 370 |
+
},
|
| 371 |
+
"Qwen3.5-35B-A3B/hle-verified-gold/summary.json": {
|
| 372 |
+
"size": 1506,
|
| 373 |
+
"sha256": "b1d67b09d49e4c48dcd194043f9e84e1357b51d13371475723f705214974b223"
|
| 374 |
+
},
|
| 375 |
+
"Qwen3.5-35B-A3B/hle-verified-gold/trajectories.jsonl": {
|
| 376 |
+
"size": 89998977,
|
| 377 |
+
"sha256": "c13192f1aba209208d7d014dbc9722a56f5cfefd369f5cdb78cd53d8469d8340"
|
| 378 |
+
},
|
| 379 |
+
"Qwen3.5-35B-A3B/hle-verified-revision/configs.json": {
|
| 380 |
+
"size": 165910,
|
| 381 |
+
"sha256": "2cf46e0583218f82eabc3556d339e6f037a78089ff93aedbe2525021b646642c"
|
| 382 |
+
},
|
| 383 |
+
"Qwen3.5-35B-A3B/hle-verified-revision/events.jsonl.gz": {
|
| 384 |
+
"size": 73200323,
|
| 385 |
+
"sha256": "0c22b10aa45d9f8318db568f60b10885e0de25597c86d11b493e1a81bb7fd898"
|
| 386 |
+
},
|
| 387 |
+
"Qwen3.5-35B-A3B/hle-verified-revision/judge_trajectories.jsonl": {
|
| 388 |
+
"size": 1463315,
|
| 389 |
+
"sha256": "64c4e07c7f748f38df05005367e8625bd6e7226b235767e8990a33d4f0493f3c"
|
| 390 |
+
},
|
| 391 |
+
"Qwen3.5-35B-A3B/hle-verified-revision/summary.json": {
|
| 392 |
+
"size": 1519,
|
| 393 |
+
"sha256": "97013cb3dbf05820ffc3e035a13c6a994d7f70e4da8fe2e9ed272761c8248748"
|
| 394 |
+
},
|
| 395 |
+
"Qwen3.5-35B-A3B/hle-verified-revision/trajectories.jsonl": {
|
| 396 |
+
"size": 160439763,
|
| 397 |
+
"sha256": "0c77c80238a38e24b1e1edd6df57f58e64bfc1f415cefd08680e8e79470cec4c"
|
| 398 |
+
}
|
| 399 |
+
}
|
| 400 |
+
}
|
Qwen3.5-35B-A3B/summary.json
CHANGED
|
@@ -1 +1,514 @@
|
|
| 1 |
-
{"arxivmath": {"benchmark": "arxivmath", "releases": [{"total_records": 68, "completed": 68, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.39705882352941174, "accuracy": 0.39705882352941174, "mean_latency_seconds": 845.1393557936724, "finish_reason_counts": {"length": 4, "stop": 64}, "length_limited_response_count": 4, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.39705882352941174, "competition_metrics": {"arxiv/december": {"n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.39705882352941174}}}, "benchmark": "matharena-arxivmath-december", "model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754", "execution_mode": "non_agent", "generation_hash": "baa213097ac551107f3a2c39858bee266fb8718ef1188f6aafc74c5dd2a15cd2", "scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661", "selected": 68, "incomplete": 0, "generated_now": 68, "scored_now": 68, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 68, "generation_concurrency": 256, "judge_concurrency": 256, "updated_at": "2026-10-29T17:33:13.338337+00:00"}, {"total_records": 128, "completed": 128, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.3046875, "accuracy": 0.3046875, "mean_latency_seconds": 817.0689857622856, "finish_reason_counts": {"length": 11, "stop": 117}, "length_limited_response_count": 11, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.3046875, "competition_metrics": {"arxiv/february": {"n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.3046875}}}, "benchmark": "matharena-arxivmath-february", "model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754", "execution_mode": "non_agent", "generation_hash": "045a3292e891b2231cc495c89612610390e45fcfea006bbb755ece00473f5bac", "scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5", "selected": 128, "incomplete": 0, "generated_now": 128, "scored_now": 128, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 128, "generation_concurrency": 256, "judge_concurrency": 256, "updated_at": "2026-10-29T18:10:28.869304+00:00"}, {"total_records": 92, "completed": 92, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.5, "accuracy": 0.5, "mean_latency_seconds": 730.6955067747913, "finish_reason_counts": {"length": 7, "stop": 85}, "length_limited_response_count": 7, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.5, "competition_metrics": {"arxiv/january": {"n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.5}}}, "benchmark": "matharena-arxivmath-january", "model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754", "execution_mode": "non_agent", "generation_hash": "7942c6e0fbb604342d159853c5ffe56d95dce774f53bcf8bd632d839d682c503", "scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8", "selected": 92, "incomplete": 0, "generated_now": 92, "scored_now": 92, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 92, "generation_concurrency": 256, "judge_concurrency": 256, "updated_at": "2026-10-29T18:55:44.632311+00:00"}, {"total_records": 124, "completed": 124, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.2903225806451613, "accuracy": 0.2903225806451613, "mean_latency_seconds": 743.6879380084199, "finish_reason_counts": {"length": 2, "stop": 122}, "length_limited_response_count": 2, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.2903225806451613, "competition_metrics": {"arxiv/march": {"n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.2903225806451613}}}, "benchmark": "matharena-arxivmath-march", "model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754", "execution_mode": "non_agent", "generation_hash": "0d2c3ed9d9698f94711f36caff57cdac3412f2b3959deeff32e13f888c79a73d", "scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8", "selected": 124, "incomplete": 0, "generated_now": 124, "scored_now": 124, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 124, "generation_concurrency": 256, "judge_concurrency": 256, "updated_at": "2026-10-29T17:48:05.135968+00:00"}], "total_records": 412, "completed": 412, "accuracy": 0.3592233009708738, "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"}, "frontierscience-olympiad": {"total_records": 3000, "completed": 3000, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.6373333333333333, "accuracy": 0.6373333333333333, "mean_latency_seconds": 506.92777187605293, "finish_reason_counts": {"length": 180, "stop": 2820}, "length_limited_response_count": 180, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 3000, "request_count": 3000, "retry_request_count": 0, "prompt_tokens": 8721756, "cached_prompt_tokens": 8475, "cache_write_prompt_tokens": 8534635, "completion_tokens": 665541, "reasoning_tokens": 416530, "total_tokens": 9387297}, "judge_api_usage": {"request_count": 3000, "prompt_tokens": 8721756, "cached_prompt_tokens": 8475, "cache_write_prompt_tokens": 8534635, "completion_tokens": 665541, "reasoning_tokens": 416530, "total_tokens": 9387297}, "benchmark_metrics": {"track": "olympiad", "n_problems_scored": 100, "n_trajectories_scored": 3000, "aggregation": "avg@30", "reported_accuracy": 0.6373333333333333, "mean_rubric_score": null, "success_threshold": 0.5, "community_profile": null}, "benchmark": "frontierscience-olympiad", "model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754", "execution_mode": "non_agent", "generation_hash": "086b6ba9996efb5d34ac6f23eb95060021b12c0e61f878efea0026922db1aa44", "scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016", "selected": 3000, "incomplete": 0, "generated_now": 3000, "scored_now": 3000, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 3000, "generation_concurrency": 256, "judge_concurrency": 256, "updated_at": "2026-10-29T18:55:48.567365+00:00", "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"}, "frontierscience-research": {"total_records": 1800, "completed": 1800, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.2915011111111111, "accuracy": 0.055, "mean_latency_seconds": 189.94285724576744, "finish_reason_counts": {"length": 1, "stop": 1799}, "length_limited_response_count": 1, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 1800, "request_count": 1824, "retry_request_count": 24, "prompt_tokens": 10622985, "cached_prompt_tokens": 132678, "cache_write_prompt_tokens": 10484174, "completion_tokens": 2920771, "reasoning_tokens": 1717727, "total_tokens": 13543756}, "judge_api_usage": {"request_count": 1824, "prompt_tokens": 10622985, "cached_prompt_tokens": 132678, "cache_write_prompt_tokens": 10484174, "completion_tokens": 2920771, "reasoning_tokens": 1717727, "total_tokens": 13543756}, "benchmark_metrics": {"track": "research", "n_problems_scored": 60, "n_trajectories_scored": 1800, "aggregation": "avg@30", "reported_accuracy": 0.055, "mean_rubric_score": 0.2915011111111111, "success_threshold": 0.7, "community_profile": null}, "benchmark": "frontierscience-research", "model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754", "execution_mode": "non_agent", "generation_hash": "ce22f07a71b105788dc1f84b95e68a2f480b0bbe9da410e399a27a99761bea38", "scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750", "selected": 1800, "incomplete": 0, "generated_now": 1800, "scored_now": 1800, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 1800, "generation_concurrency": 256, "judge_concurrency": 256, "updated_at": "2026-10-29T18:08:19.524662+00:00", "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"}, "scicode-verified": {"updated_at": 1793294847.0649095, "expected_problems": 64, "expected_steps": 287, "generated_steps": 287, "scored_steps": 287, "passed_steps": 183, "completed_problems": 64, "passed_problems": 31, "subproblem_accuracy": 0.6376306620209059, "main_problem_accuracy": 0.484375, "complete": true, "cap_hits": 0, "mean_output_tokens": 8847.940766550522, "grading_infrastructure_errors": 0, "errors": [], "provisional": false, "source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"}, "terminal-bench-2.1": {"tasks": 89, "attempts": 178, "attempts_per_task": 2, "metric": "avg@2", "passed": 61, "avg_at_2": 0.34269662921348315, "unscored": 0, "agent_timeouts": 55, "reward_positive_with_exception": 2, "scoring": "native_verifier_with_explicit_user_adjudications", "scoring_policy": "native-verifier-after-agent-stop-v1", "provisional": false, "grading_complete": true, "native_scored": 176, "adjudicated_zero": 2, "source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/run", "scoring_adjudications": {"feal-differential-cryptanalysis--attempt-002": {"timestamp": "2026-10-30T02:53:28.310866+00:00", "reward": 0.0, "native_reward": null, "original_verifier_result": null, "method": "explicit_user_authorized_zero_for_unscored_attempt", "authorization": "[REDACTED_CREDENTIAL]", "reason": "User-adjudicated zero for missing native verdict after timeout; not a completed native test failure.", "exception_type": "AgentTimeoutError", "backup": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--feal-differential-cryptanalysis--attempt-002", "original_sha256": "0e82941118e176819899059acf8c7991a3316409556f3c9ffdaf6518f2d91006", "generation_or_verifier_rerun": false}, "model-extraction-relu-logits--attempt-002": {"timestamp": "2026-10-30T02:53:28.355050+00:00", "reward": 0.0, "native_reward": null, "original_verifier_result": null, "method": "explicit_user_authorized_zero_for_unscored_attempt", "authorization": "[REDACTED_CREDENTIAL]", "reason": "User-adjudicated zero for missing native verdict after timeout; not a completed native test failure.", "exception_type": "AgentTimeoutError", "backup": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--model-extraction-relu-logits--attempt-002", "original_sha256": "3982db0e631fe569bbc1d3a03af9e22cfe4079bb4e396a0de31555ccb25f2b74", "generation_or_verifier_rerun": false}}, "scoring_note": "Two explicit user-adjudicated zeros, not native test failures. Original null rewards preserved in agentic_recovery_provenance.tar.gz."}, "scicoding-agent": {"tasks": 50, "attempts": 100, "attempts_per_task": 2, "metric": "avg@2", "passed": 36, "avg_at_2": 0.36, "unscored": 0, "agent_timeouts": 1, "reward_positive_with_exception": 0, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1", "provisional": false, "grading_complete": true, "native_scored": 100, "adjudicated_zero": 0, "source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/run"}}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"arxivmath": {
|
| 3 |
+
"benchmark": "arxivmath",
|
| 4 |
+
"releases": [
|
| 5 |
+
{
|
| 6 |
+
"total_records": 68,
|
| 7 |
+
"completed": 68,
|
| 8 |
+
"query_errors": 0,
|
| 9 |
+
"score_errors": 0,
|
| 10 |
+
"generated_unscored": 0,
|
| 11 |
+
"mean_score": 0.39705882352941174,
|
| 12 |
+
"accuracy": 0.39705882352941174,
|
| 13 |
+
"mean_latency_seconds": 845.1393557936724,
|
| 14 |
+
"finish_reason_counts": {
|
| 15 |
+
"length": 4,
|
| 16 |
+
"stop": 64
|
| 17 |
+
},
|
| 18 |
+
"length_limited_response_count": 4,
|
| 19 |
+
"observed_max_completion_tokens": 98304,
|
| 20 |
+
"judge_usage": {
|
| 21 |
+
"judged_record_count": 0,
|
| 22 |
+
"request_count": 0,
|
| 23 |
+
"retry_request_count": 0,
|
| 24 |
+
"prompt_tokens": 0,
|
| 25 |
+
"cached_prompt_tokens": 0,
|
| 26 |
+
"cache_write_prompt_tokens": 0,
|
| 27 |
+
"completion_tokens": 0,
|
| 28 |
+
"reasoning_tokens": 0,
|
| 29 |
+
"total_tokens": 0
|
| 30 |
+
},
|
| 31 |
+
"judge_api_usage": {
|
| 32 |
+
"request_count": 0,
|
| 33 |
+
"prompt_tokens": 0,
|
| 34 |
+
"cached_prompt_tokens": 0,
|
| 35 |
+
"cache_write_prompt_tokens": 0,
|
| 36 |
+
"completion_tokens": 0,
|
| 37 |
+
"reasoning_tokens": 0,
|
| 38 |
+
"total_tokens": 0
|
| 39 |
+
},
|
| 40 |
+
"benchmark_metrics": {
|
| 41 |
+
"competition": "arxivmath",
|
| 42 |
+
"aggregation": "avg@4",
|
| 43 |
+
"n_problems_scored": 17,
|
| 44 |
+
"n_trajectories_scored": 68,
|
| 45 |
+
"reported_accuracy": 0.39705882352941174,
|
| 46 |
+
"competition_metrics": {
|
| 47 |
+
"arxiv/december": {
|
| 48 |
+
"n_problems_scored": 17,
|
| 49 |
+
"n_trajectories_scored": 68,
|
| 50 |
+
"reported_accuracy": 0.39705882352941174
|
| 51 |
+
}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"benchmark": "matharena-arxivmath-december",
|
| 55 |
+
"model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 56 |
+
"execution_mode": "non_agent",
|
| 57 |
+
"generation_hash": "baa213097ac551107f3a2c39858bee266fb8718ef1188f6aafc74c5dd2a15cd2",
|
| 58 |
+
"scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661",
|
| 59 |
+
"selected": 68,
|
| 60 |
+
"incomplete": 0,
|
| 61 |
+
"generated_now": 68,
|
| 62 |
+
"scored_now": 68,
|
| 63 |
+
"cached_completed": 0,
|
| 64 |
+
"stage": "all",
|
| 65 |
+
"worker_count": 1,
|
| 66 |
+
"worker_index": 0,
|
| 67 |
+
"worker_owned_predictions": 68,
|
| 68 |
+
"generation_concurrency": 256,
|
| 69 |
+
"judge_concurrency": 256,
|
| 70 |
+
"updated_at": "2026-10-29T17:33:13.338337+00:00"
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"total_records": 128,
|
| 74 |
+
"completed": 128,
|
| 75 |
+
"query_errors": 0,
|
| 76 |
+
"score_errors": 0,
|
| 77 |
+
"generated_unscored": 0,
|
| 78 |
+
"mean_score": 0.3046875,
|
| 79 |
+
"accuracy": 0.3046875,
|
| 80 |
+
"mean_latency_seconds": 817.0689857622856,
|
| 81 |
+
"finish_reason_counts": {
|
| 82 |
+
"length": 11,
|
| 83 |
+
"stop": 117
|
| 84 |
+
},
|
| 85 |
+
"length_limited_response_count": 11,
|
| 86 |
+
"observed_max_completion_tokens": 98304,
|
| 87 |
+
"judge_usage": {
|
| 88 |
+
"judged_record_count": 0,
|
| 89 |
+
"request_count": 0,
|
| 90 |
+
"retry_request_count": 0,
|
| 91 |
+
"prompt_tokens": 0,
|
| 92 |
+
"cached_prompt_tokens": 0,
|
| 93 |
+
"cache_write_prompt_tokens": 0,
|
| 94 |
+
"completion_tokens": 0,
|
| 95 |
+
"reasoning_tokens": 0,
|
| 96 |
+
"total_tokens": 0
|
| 97 |
+
},
|
| 98 |
+
"judge_api_usage": {
|
| 99 |
+
"request_count": 0,
|
| 100 |
+
"prompt_tokens": 0,
|
| 101 |
+
"cached_prompt_tokens": 0,
|
| 102 |
+
"cache_write_prompt_tokens": 0,
|
| 103 |
+
"completion_tokens": 0,
|
| 104 |
+
"reasoning_tokens": 0,
|
| 105 |
+
"total_tokens": 0
|
| 106 |
+
},
|
| 107 |
+
"benchmark_metrics": {
|
| 108 |
+
"competition": "arxivmath",
|
| 109 |
+
"aggregation": "avg@4",
|
| 110 |
+
"n_problems_scored": 32,
|
| 111 |
+
"n_trajectories_scored": 128,
|
| 112 |
+
"reported_accuracy": 0.3046875,
|
| 113 |
+
"competition_metrics": {
|
| 114 |
+
"arxiv/february": {
|
| 115 |
+
"n_problems_scored": 32,
|
| 116 |
+
"n_trajectories_scored": 128,
|
| 117 |
+
"reported_accuracy": 0.3046875
|
| 118 |
+
}
|
| 119 |
+
}
|
| 120 |
+
},
|
| 121 |
+
"benchmark": "matharena-arxivmath-february",
|
| 122 |
+
"model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 123 |
+
"execution_mode": "non_agent",
|
| 124 |
+
"generation_hash": "045a3292e891b2231cc495c89612610390e45fcfea006bbb755ece00473f5bac",
|
| 125 |
+
"scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5",
|
| 126 |
+
"selected": 128,
|
| 127 |
+
"incomplete": 0,
|
| 128 |
+
"generated_now": 128,
|
| 129 |
+
"scored_now": 128,
|
| 130 |
+
"cached_completed": 0,
|
| 131 |
+
"stage": "all",
|
| 132 |
+
"worker_count": 1,
|
| 133 |
+
"worker_index": 0,
|
| 134 |
+
"worker_owned_predictions": 128,
|
| 135 |
+
"generation_concurrency": 256,
|
| 136 |
+
"judge_concurrency": 256,
|
| 137 |
+
"updated_at": "2026-10-29T18:10:28.869304+00:00"
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"total_records": 92,
|
| 141 |
+
"completed": 92,
|
| 142 |
+
"query_errors": 0,
|
| 143 |
+
"score_errors": 0,
|
| 144 |
+
"generated_unscored": 0,
|
| 145 |
+
"mean_score": 0.5,
|
| 146 |
+
"accuracy": 0.5,
|
| 147 |
+
"mean_latency_seconds": 730.6955067747913,
|
| 148 |
+
"finish_reason_counts": {
|
| 149 |
+
"length": 7,
|
| 150 |
+
"stop": 85
|
| 151 |
+
},
|
| 152 |
+
"length_limited_response_count": 7,
|
| 153 |
+
"observed_max_completion_tokens": 98304,
|
| 154 |
+
"judge_usage": {
|
| 155 |
+
"judged_record_count": 0,
|
| 156 |
+
"request_count": 0,
|
| 157 |
+
"retry_request_count": 0,
|
| 158 |
+
"prompt_tokens": 0,
|
| 159 |
+
"cached_prompt_tokens": 0,
|
| 160 |
+
"cache_write_prompt_tokens": 0,
|
| 161 |
+
"completion_tokens": 0,
|
| 162 |
+
"reasoning_tokens": 0,
|
| 163 |
+
"total_tokens": 0
|
| 164 |
+
},
|
| 165 |
+
"judge_api_usage": {
|
| 166 |
+
"request_count": 0,
|
| 167 |
+
"prompt_tokens": 0,
|
| 168 |
+
"cached_prompt_tokens": 0,
|
| 169 |
+
"cache_write_prompt_tokens": 0,
|
| 170 |
+
"completion_tokens": 0,
|
| 171 |
+
"reasoning_tokens": 0,
|
| 172 |
+
"total_tokens": 0
|
| 173 |
+
},
|
| 174 |
+
"benchmark_metrics": {
|
| 175 |
+
"competition": "arxivmath",
|
| 176 |
+
"aggregation": "avg@4",
|
| 177 |
+
"n_problems_scored": 23,
|
| 178 |
+
"n_trajectories_scored": 92,
|
| 179 |
+
"reported_accuracy": 0.5,
|
| 180 |
+
"competition_metrics": {
|
| 181 |
+
"arxiv/january": {
|
| 182 |
+
"n_problems_scored": 23,
|
| 183 |
+
"n_trajectories_scored": 92,
|
| 184 |
+
"reported_accuracy": 0.5
|
| 185 |
+
}
|
| 186 |
+
}
|
| 187 |
+
},
|
| 188 |
+
"benchmark": "matharena-arxivmath-january",
|
| 189 |
+
"model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 190 |
+
"execution_mode": "non_agent",
|
| 191 |
+
"generation_hash": "7942c6e0fbb604342d159853c5ffe56d95dce774f53bcf8bd632d839d682c503",
|
| 192 |
+
"scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8",
|
| 193 |
+
"selected": 92,
|
| 194 |
+
"incomplete": 0,
|
| 195 |
+
"generated_now": 92,
|
| 196 |
+
"scored_now": 92,
|
| 197 |
+
"cached_completed": 0,
|
| 198 |
+
"stage": "all",
|
| 199 |
+
"worker_count": 1,
|
| 200 |
+
"worker_index": 0,
|
| 201 |
+
"worker_owned_predictions": 92,
|
| 202 |
+
"generation_concurrency": 256,
|
| 203 |
+
"judge_concurrency": 256,
|
| 204 |
+
"updated_at": "2026-10-29T18:55:44.632311+00:00"
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"total_records": 124,
|
| 208 |
+
"completed": 124,
|
| 209 |
+
"query_errors": 0,
|
| 210 |
+
"score_errors": 0,
|
| 211 |
+
"generated_unscored": 0,
|
| 212 |
+
"mean_score": 0.2903225806451613,
|
| 213 |
+
"accuracy": 0.2903225806451613,
|
| 214 |
+
"mean_latency_seconds": 743.6879380084199,
|
| 215 |
+
"finish_reason_counts": {
|
| 216 |
+
"length": 2,
|
| 217 |
+
"stop": 122
|
| 218 |
+
},
|
| 219 |
+
"length_limited_response_count": 2,
|
| 220 |
+
"observed_max_completion_tokens": 98304,
|
| 221 |
+
"judge_usage": {
|
| 222 |
+
"judged_record_count": 0,
|
| 223 |
+
"request_count": 0,
|
| 224 |
+
"retry_request_count": 0,
|
| 225 |
+
"prompt_tokens": 0,
|
| 226 |
+
"cached_prompt_tokens": 0,
|
| 227 |
+
"cache_write_prompt_tokens": 0,
|
| 228 |
+
"completion_tokens": 0,
|
| 229 |
+
"reasoning_tokens": 0,
|
| 230 |
+
"total_tokens": 0
|
| 231 |
+
},
|
| 232 |
+
"judge_api_usage": {
|
| 233 |
+
"request_count": 0,
|
| 234 |
+
"prompt_tokens": 0,
|
| 235 |
+
"cached_prompt_tokens": 0,
|
| 236 |
+
"cache_write_prompt_tokens": 0,
|
| 237 |
+
"completion_tokens": 0,
|
| 238 |
+
"reasoning_tokens": 0,
|
| 239 |
+
"total_tokens": 0
|
| 240 |
+
},
|
| 241 |
+
"benchmark_metrics": {
|
| 242 |
+
"competition": "arxivmath",
|
| 243 |
+
"aggregation": "avg@4",
|
| 244 |
+
"n_problems_scored": 31,
|
| 245 |
+
"n_trajectories_scored": 124,
|
| 246 |
+
"reported_accuracy": 0.2903225806451613,
|
| 247 |
+
"competition_metrics": {
|
| 248 |
+
"arxiv/march": {
|
| 249 |
+
"n_problems_scored": 31,
|
| 250 |
+
"n_trajectories_scored": 124,
|
| 251 |
+
"reported_accuracy": 0.2903225806451613
|
| 252 |
+
}
|
| 253 |
+
}
|
| 254 |
+
},
|
| 255 |
+
"benchmark": "matharena-arxivmath-march",
|
| 256 |
+
"model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 257 |
+
"execution_mode": "non_agent",
|
| 258 |
+
"generation_hash": "0d2c3ed9d9698f94711f36caff57cdac3412f2b3959deeff32e13f888c79a73d",
|
| 259 |
+
"scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8",
|
| 260 |
+
"selected": 124,
|
| 261 |
+
"incomplete": 0,
|
| 262 |
+
"generated_now": 124,
|
| 263 |
+
"scored_now": 124,
|
| 264 |
+
"cached_completed": 0,
|
| 265 |
+
"stage": "all",
|
| 266 |
+
"worker_count": 1,
|
| 267 |
+
"worker_index": 0,
|
| 268 |
+
"worker_owned_predictions": 124,
|
| 269 |
+
"generation_concurrency": 256,
|
| 270 |
+
"judge_concurrency": 256,
|
| 271 |
+
"updated_at": "2026-10-29T17:48:05.135968+00:00"
|
| 272 |
+
}
|
| 273 |
+
],
|
| 274 |
+
"total_records": 412,
|
| 275 |
+
"completed": 412,
|
| 276 |
+
"accuracy": 0.3592233009708738,
|
| 277 |
+
"provisional": false,
|
| 278 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"
|
| 279 |
+
},
|
| 280 |
+
"frontierscience-olympiad": {
|
| 281 |
+
"total_records": 3000,
|
| 282 |
+
"completed": 3000,
|
| 283 |
+
"query_errors": 0,
|
| 284 |
+
"score_errors": 0,
|
| 285 |
+
"generated_unscored": 0,
|
| 286 |
+
"mean_score": 0.6373333333333333,
|
| 287 |
+
"accuracy": 0.6373333333333333,
|
| 288 |
+
"mean_latency_seconds": 506.92777187605293,
|
| 289 |
+
"finish_reason_counts": {
|
| 290 |
+
"length": 180,
|
| 291 |
+
"stop": 2820
|
| 292 |
+
},
|
| 293 |
+
"length_limited_response_count": 180,
|
| 294 |
+
"observed_max_completion_tokens": 98304,
|
| 295 |
+
"judge_usage": {
|
| 296 |
+
"judged_record_count": 3000,
|
| 297 |
+
"request_count": 3000,
|
| 298 |
+
"retry_request_count": 0,
|
| 299 |
+
"prompt_tokens": 8721756,
|
| 300 |
+
"cached_prompt_tokens": 8475,
|
| 301 |
+
"cache_write_prompt_tokens": 8534635,
|
| 302 |
+
"completion_tokens": 665541,
|
| 303 |
+
"reasoning_tokens": 416530,
|
| 304 |
+
"total_tokens": 9387297
|
| 305 |
+
},
|
| 306 |
+
"judge_api_usage": {
|
| 307 |
+
"request_count": 3000,
|
| 308 |
+
"prompt_tokens": 8721756,
|
| 309 |
+
"cached_prompt_tokens": 8475,
|
| 310 |
+
"cache_write_prompt_tokens": 8534635,
|
| 311 |
+
"completion_tokens": 665541,
|
| 312 |
+
"reasoning_tokens": 416530,
|
| 313 |
+
"total_tokens": 9387297
|
| 314 |
+
},
|
| 315 |
+
"benchmark_metrics": {
|
| 316 |
+
"track": "olympiad",
|
| 317 |
+
"n_problems_scored": 100,
|
| 318 |
+
"n_trajectories_scored": 3000,
|
| 319 |
+
"aggregation": "avg@30",
|
| 320 |
+
"reported_accuracy": 0.6373333333333333,
|
| 321 |
+
"mean_rubric_score": null,
|
| 322 |
+
"success_threshold": 0.5,
|
| 323 |
+
"community_profile": null
|
| 324 |
+
},
|
| 325 |
+
"benchmark": "frontierscience-olympiad",
|
| 326 |
+
"model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 327 |
+
"execution_mode": "non_agent",
|
| 328 |
+
"generation_hash": "086b6ba9996efb5d34ac6f23eb95060021b12c0e61f878efea0026922db1aa44",
|
| 329 |
+
"scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016",
|
| 330 |
+
"selected": 3000,
|
| 331 |
+
"incomplete": 0,
|
| 332 |
+
"generated_now": 3000,
|
| 333 |
+
"scored_now": 3000,
|
| 334 |
+
"cached_completed": 0,
|
| 335 |
+
"stage": "all",
|
| 336 |
+
"worker_count": 1,
|
| 337 |
+
"worker_index": 0,
|
| 338 |
+
"worker_owned_predictions": 3000,
|
| 339 |
+
"generation_concurrency": 256,
|
| 340 |
+
"judge_concurrency": 256,
|
| 341 |
+
"updated_at": "2026-10-29T18:55:48.567365+00:00",
|
| 342 |
+
"provisional": false,
|
| 343 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"
|
| 344 |
+
},
|
| 345 |
+
"frontierscience-research": {
|
| 346 |
+
"total_records": 1800,
|
| 347 |
+
"completed": 1800,
|
| 348 |
+
"query_errors": 0,
|
| 349 |
+
"score_errors": 0,
|
| 350 |
+
"generated_unscored": 0,
|
| 351 |
+
"mean_score": 0.2915011111111111,
|
| 352 |
+
"accuracy": 0.055,
|
| 353 |
+
"mean_latency_seconds": 189.94285724576744,
|
| 354 |
+
"finish_reason_counts": {
|
| 355 |
+
"length": 1,
|
| 356 |
+
"stop": 1799
|
| 357 |
+
},
|
| 358 |
+
"length_limited_response_count": 1,
|
| 359 |
+
"observed_max_completion_tokens": 98304,
|
| 360 |
+
"judge_usage": {
|
| 361 |
+
"judged_record_count": 1800,
|
| 362 |
+
"request_count": 1824,
|
| 363 |
+
"retry_request_count": 24,
|
| 364 |
+
"prompt_tokens": 10622985,
|
| 365 |
+
"cached_prompt_tokens": 132678,
|
| 366 |
+
"cache_write_prompt_tokens": 10484174,
|
| 367 |
+
"completion_tokens": 2920771,
|
| 368 |
+
"reasoning_tokens": 1717727,
|
| 369 |
+
"total_tokens": 13543756
|
| 370 |
+
},
|
| 371 |
+
"judge_api_usage": {
|
| 372 |
+
"request_count": 1824,
|
| 373 |
+
"prompt_tokens": 10622985,
|
| 374 |
+
"cached_prompt_tokens": 132678,
|
| 375 |
+
"cache_write_prompt_tokens": 10484174,
|
| 376 |
+
"completion_tokens": 2920771,
|
| 377 |
+
"reasoning_tokens": 1717727,
|
| 378 |
+
"total_tokens": 13543756
|
| 379 |
+
},
|
| 380 |
+
"benchmark_metrics": {
|
| 381 |
+
"track": "research",
|
| 382 |
+
"n_problems_scored": 60,
|
| 383 |
+
"n_trajectories_scored": 1800,
|
| 384 |
+
"aggregation": "avg@30",
|
| 385 |
+
"reported_accuracy": 0.055,
|
| 386 |
+
"mean_rubric_score": 0.2915011111111111,
|
| 387 |
+
"success_threshold": 0.7,
|
| 388 |
+
"community_profile": null
|
| 389 |
+
},
|
| 390 |
+
"benchmark": "frontierscience-research",
|
| 391 |
+
"model": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754",
|
| 392 |
+
"execution_mode": "non_agent",
|
| 393 |
+
"generation_hash": "ce22f07a71b105788dc1f84b95e68a2f480b0bbe9da410e399a27a99761bea38",
|
| 394 |
+
"scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750",
|
| 395 |
+
"selected": 1800,
|
| 396 |
+
"incomplete": 0,
|
| 397 |
+
"generated_now": 1800,
|
| 398 |
+
"scored_now": 1800,
|
| 399 |
+
"cached_completed": 0,
|
| 400 |
+
"stage": "all",
|
| 401 |
+
"worker_count": 1,
|
| 402 |
+
"worker_index": 0,
|
| 403 |
+
"worker_owned_predictions": 1800,
|
| 404 |
+
"generation_concurrency": 256,
|
| 405 |
+
"judge_concurrency": 256,
|
| 406 |
+
"updated_at": "2026-10-29T18:08:19.524662+00:00",
|
| 407 |
+
"provisional": false,
|
| 408 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"
|
| 409 |
+
},
|
| 410 |
+
"scicode-verified": {
|
| 411 |
+
"updated_at": 1793294847.0649095,
|
| 412 |
+
"expected_problems": 64,
|
| 413 |
+
"expected_steps": 287,
|
| 414 |
+
"generated_steps": 287,
|
| 415 |
+
"scored_steps": 287,
|
| 416 |
+
"passed_steps": 183,
|
| 417 |
+
"completed_problems": 64,
|
| 418 |
+
"passed_problems": 31,
|
| 419 |
+
"subproblem_accuracy": 0.6376306620209059,
|
| 420 |
+
"main_problem_accuracy": 0.484375,
|
| 421 |
+
"complete": true,
|
| 422 |
+
"cap_hits": 0,
|
| 423 |
+
"mean_output_tokens": 8847.940766550522,
|
| 424 |
+
"grading_infrastructure_errors": 0,
|
| 425 |
+
"errors": [],
|
| 426 |
+
"provisional": false,
|
| 427 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-nonagentic-20261029/run"
|
| 428 |
+
},
|
| 429 |
+
"terminal-bench-2.1": {
|
| 430 |
+
"tasks": 89,
|
| 431 |
+
"attempts": 178,
|
| 432 |
+
"attempts_per_task": 2,
|
| 433 |
+
"metric": "avg@2",
|
| 434 |
+
"passed": 61,
|
| 435 |
+
"avg_at_2": 0.34269662921348315,
|
| 436 |
+
"unscored": 0,
|
| 437 |
+
"agent_timeouts": 55,
|
| 438 |
+
"reward_positive_with_exception": 2,
|
| 439 |
+
"scoring": "native_verifier_with_explicit_user_adjudications",
|
| 440 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 441 |
+
"provisional": false,
|
| 442 |
+
"grading_complete": true,
|
| 443 |
+
"native_scored": 176,
|
| 444 |
+
"adjudicated_zero": 2,
|
| 445 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/run",
|
| 446 |
+
"scoring_adjudications": {
|
| 447 |
+
"feal-differential-cryptanalysis--attempt-002": {
|
| 448 |
+
"timestamp": "2026-10-30T02:53:28.310866+00:00",
|
| 449 |
+
"reward": 0.0,
|
| 450 |
+
"native_reward": null,
|
| 451 |
+
"original_verifier_result": null,
|
| 452 |
+
"method": "explicit_user_authorized_zero_for_unscored_attempt",
|
| 453 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 454 |
+
"reason": "User-adjudicated zero for missing native verdict after timeout; not a completed native test failure.",
|
| 455 |
+
"exception_type": "AgentTimeoutError",
|
| 456 |
+
"backup": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--feal-differential-cryptanalysis--attempt-002",
|
| 457 |
+
"original_sha256": "0e82941118e176819899059acf8c7991a3316409556f3c9ffdaf6518f2d91006",
|
| 458 |
+
"generation_or_verifier_rerun": false
|
| 459 |
+
},
|
| 460 |
+
"model-extraction-relu-logits--attempt-002": {
|
| 461 |
+
"timestamp": "2026-10-30T02:53:28.355050+00:00",
|
| 462 |
+
"reward": 0.0,
|
| 463 |
+
"native_reward": null,
|
| 464 |
+
"original_verifier_result": null,
|
| 465 |
+
"method": "explicit_user_authorized_zero_for_unscored_attempt",
|
| 466 |
+
"authorization": "[REDACTED_CREDENTIAL]",
|
| 467 |
+
"reason": "User-adjudicated zero for missing native verdict after timeout; not a completed native test failure.",
|
| 468 |
+
"exception_type": "AgentTimeoutError",
|
| 469 |
+
"backup": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/operations/user-zero-adjudication-20261029/terminal-bench-2.1--model-extraction-relu-logits--attempt-002",
|
| 470 |
+
"original_sha256": "3982db0e631fe569bbc1d3a03af9e22cfe4079bb4e396a0de31555ccb25f2b74",
|
| 471 |
+
"generation_or_verifier_rerun": false
|
| 472 |
+
}
|
| 473 |
+
},
|
| 474 |
+
"scoring_note": "Two explicit user-adjudicated zeros, not native test failures. Original null rewards preserved in agentic_recovery_provenance.tar.gz."
|
| 475 |
+
},
|
| 476 |
+
"scicoding-agent": {
|
| 477 |
+
"tasks": 50,
|
| 478 |
+
"attempts": 100,
|
| 479 |
+
"attempts_per_task": 2,
|
| 480 |
+
"metric": "avg@2",
|
| 481 |
+
"passed": 36,
|
| 482 |
+
"avg_at_2": 0.36,
|
| 483 |
+
"unscored": 0,
|
| 484 |
+
"agent_timeouts": 1,
|
| 485 |
+
"reward_positive_with_exception": 0,
|
| 486 |
+
"scoring": "native_verifier",
|
| 487 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 488 |
+
"provisional": false,
|
| 489 |
+
"grading_complete": true,
|
| 490 |
+
"native_scored": 100,
|
| 491 |
+
"adjudicated_zero": 0,
|
| 492 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/qwen35-35b-agentic-20261029/run"
|
| 493 |
+
},
|
| 494 |
+
"hle-verified": {
|
| 495 |
+
"gold": {
|
| 496 |
+
"correct": 228,
|
| 497 |
+
"total": 575,
|
| 498 |
+
"accuracy": 0.39652173913043476,
|
| 499 |
+
"cap_hits": 41
|
| 500 |
+
},
|
| 501 |
+
"revision": {
|
| 502 |
+
"correct": 365,
|
| 503 |
+
"total": 1025,
|
| 504 |
+
"accuracy": 0.35609756097560974,
|
| 505 |
+
"cap_hits": 63
|
| 506 |
+
},
|
| 507 |
+
"correct": 593,
|
| 508 |
+
"total": 1600,
|
| 509 |
+
"overall_accuracy": 0.370625,
|
| 510 |
+
"all_complete": true,
|
| 511 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 512 |
+
"provenance": "hle-verified-provenance.json"
|
| 513 |
+
}
|
| 514 |
+
}
|
Qwen3.5-9B/README.md
CHANGED
|
@@ -22,3 +22,17 @@ See [scicode-verified](scicode-verified/README.md) for the fresh pass@1 evaluati
|
|
| 22 |
- [scicoding-agent](scicoding-agent/README.md): **9/100 (9.00%) avg@2**, 0 unscored.
|
| 23 |
|
| 24 |
Agentic results have separate recovery/adjudication provenance. Existing FS, ArXivMath and SciCode results are unchanged.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
- [scicoding-agent](scicoding-agent/README.md): **9/100 (9.00%) avg@2**, 0 unscored.
|
| 23 |
|
| 24 |
Agentic results have separate recovery/adjudication provenance. Existing FS, ArXivMath and SciCode results are unchanged.
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
## HLE-Verified (2026-10-30)
|
| 28 |
+
|
| 29 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **459/1,600 (28.69%)**.
|
| 30 |
+
|
| 31 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 32 |
+
|---|---:|---:|---|
|
| 33 |
+
| Gold | 170/575 | 29.57% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 34 |
+
| Revision | 289/1025 | 28.20% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 35 |
+
|
| 36 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 98,304 tokens; context 131,072 tokens. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 37 |
+
|
| 38 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
Qwen3.5-9B/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,395 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "qwen35-9b-it",
|
| 7 |
+
"folder": "Qwen3.5-9B",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 170,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.2956521739130435,
|
| 13 |
+
"cap_hits": 72
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 289,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.2819512195121951,
|
| 19 |
+
"cap_hits": 115
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 459,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.286875
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "qwen35-9b-it",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "qwen35-9b",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "qwen",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "qwen3_5_text",
|
| 35 |
+
"hidden_size": 4096,
|
| 36 |
+
"num_hidden_layers": 32,
|
| 37 |
+
"vocab_size": 248320
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 98304,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"top_k": -1,
|
| 46 |
+
"min_p": 0.0,
|
| 47 |
+
"presence_penalty": 0.0,
|
| 48 |
+
"repetition_penalty": 1.0,
|
| 49 |
+
"chat_template_kwargs": {
|
| 50 |
+
"enable_thinking": true
|
| 51 |
+
}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"serving": {
|
| 55 |
+
"context_length": 131072,
|
| 56 |
+
"dtype": "bfloat16",
|
| 57 |
+
"attention_backend": "triton",
|
| 58 |
+
"mem_fraction_static": 0.8,
|
| 59 |
+
"reasoning_parser": "qwen3",
|
| 60 |
+
"model_config_parser": "sci_qwen35_text",
|
| 61 |
+
"external_model_package": "sci_qwen35_text_model",
|
| 62 |
+
"extra_args": [
|
| 63 |
+
"--sampling-backend",
|
| 64 |
+
"flashinfer",
|
| 65 |
+
"--cuda-graph-backend-decode",
|
| 66 |
+
"full",
|
| 67 |
+
"--cuda-graph-backend-prefill",
|
| 68 |
+
"disabled"
|
| 69 |
+
],
|
| 70 |
+
"tp": 2,
|
| 71 |
+
"replicas_per_node": 4,
|
| 72 |
+
"max_running_requests": 32
|
| 73 |
+
},
|
| 74 |
+
"reasoning": {
|
| 75 |
+
"requests_per_engine": 32,
|
| 76 |
+
"benchmarks": {}
|
| 77 |
+
},
|
| 78 |
+
"agentic_tasks_per_engine": 1
|
| 79 |
+
},
|
| 80 |
+
"node": 1,
|
| 81 |
+
"checkpoint": "/workspace/data/raw_model/Qwen3.5-9B-text",
|
| 82 |
+
"repo": "Qwen/Qwen3.5-9B",
|
| 83 |
+
"revision": "c202236235762e1c871ad0ccb60c8ee5ba337b9a",
|
| 84 |
+
"checkpoint_config_sha256": "7ed31d50f8c0e8dc8219aa686a97c08665b8d9c85321e4c2fb766139fb6d01af",
|
| 85 |
+
"index_sha256": "6ef4a22a4e97a96f7d4d55e97d2eeb4bc589720cfefb349583111f21ee78e376",
|
| 86 |
+
"shard_sizes": {
|
| 87 |
+
"model-00001-of-00004.safetensors": 4973901592,
|
| 88 |
+
"model-00002-of-00004.safetensors": 4988362880,
|
| 89 |
+
"model-00003-of-00004.safetensors": 4954200776,
|
| 90 |
+
"model-00004-of-00004.safetensors": 3477773688
|
| 91 |
+
},
|
| 92 |
+
"endpoints": [
|
| 93 |
+
{
|
| 94 |
+
"id": "node-1-replica-0",
|
| 95 |
+
"api_base": "http://10.1.136.8:8700/v1"
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"id": "node-1-replica-1",
|
| 99 |
+
"api_base": "http://10.1.136.8:8701/v1"
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"id": "node-1-replica-2",
|
| 103 |
+
"api_base": "http://10.1.136.8:8702/v1"
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"id": "node-1-replica-3",
|
| 107 |
+
"api_base": "http://10.1.136.8:8703/v1"
|
| 108 |
+
}
|
| 109 |
+
],
|
| 110 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-9b-it",
|
| 111 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-9b-it/configs/hle-verified-gold-revision-96k.yaml",
|
| 112 |
+
"concurrency": 128,
|
| 113 |
+
"judge_concurrency": 64,
|
| 114 |
+
"tracking": {
|
| 115 |
+
"enabled": false,
|
| 116 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 117 |
+
},
|
| 118 |
+
"config_sha256": {
|
| 119 |
+
"model.yaml": "31b150f4d5ef7e169dc07acda909f93270ebbe98c8ba80516e48623ba409d085",
|
| 120 |
+
"configs/hle-verified-gold-revision-96k.yaml": "a417e37e8a3c4a4b3f8a09eedb757b5df167c16ec9a8a9337396900925fc6d1d",
|
| 121 |
+
"configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4",
|
| 122 |
+
"configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"
|
| 123 |
+
},
|
| 124 |
+
"serving_export": "Existing text-only export used in previous IT evaluations; no weight modifications in this sweep"
|
| 125 |
+
},
|
| 126 |
+
"dataset": {
|
| 127 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 128 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 129 |
+
"counts": {
|
| 130 |
+
"Gold": {
|
| 131 |
+
"total": 668,
|
| 132 |
+
"text": 575,
|
| 133 |
+
"image_excluded": 93
|
| 134 |
+
},
|
| 135 |
+
"Revision": {
|
| 136 |
+
"total": 1143,
|
| 137 |
+
"text": 1025,
|
| 138 |
+
"image_excluded": 118
|
| 139 |
+
}
|
| 140 |
+
},
|
| 141 |
+
"sha256": {
|
| 142 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 143 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 144 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 145 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 146 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 147 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 148 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 149 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 150 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 151 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 152 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 153 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 154 |
+
}
|
| 155 |
+
},
|
| 156 |
+
"token_preflight": {
|
| 157 |
+
"model": "qwen35-9b-it",
|
| 158 |
+
"questions": 1600,
|
| 159 |
+
"max_prompt_tokens": 13586,
|
| 160 |
+
"output_cap": 98304,
|
| 161 |
+
"context": 131072,
|
| 162 |
+
"dynamic_output_budget": false,
|
| 163 |
+
"affected_questions": 0,
|
| 164 |
+
"min_effective_output_cap": 98304
|
| 165 |
+
},
|
| 166 |
+
"shared_recovery_provenance": {
|
| 167 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 168 |
+
"publication_reruns": false,
|
| 169 |
+
"all_questions_scored": true,
|
| 170 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 171 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 172 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 173 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 174 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 175 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 176 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 177 |
+
},
|
| 178 |
+
"source_manifest": {
|
| 179 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 180 |
+
"files": {
|
| 181 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 182 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 183 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 184 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 185 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 186 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 187 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 188 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 189 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 190 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 191 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 192 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 193 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 194 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 195 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 196 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 197 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 198 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 199 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 200 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 201 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 202 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 203 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 204 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 205 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 206 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 207 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 208 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 209 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 210 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 211 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 212 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 213 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 214 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 215 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 216 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 217 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 218 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 219 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 220 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 221 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 222 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 223 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 224 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 225 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 226 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 227 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 228 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 229 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 230 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 231 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 232 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 233 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 234 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 235 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 236 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 237 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 238 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 239 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 240 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 241 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 242 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 243 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 244 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 245 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 246 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 247 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 248 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 249 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 250 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 251 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 252 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 253 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 254 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 255 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 256 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 257 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 258 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 259 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 260 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 261 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 262 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 263 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 264 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 265 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 266 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 267 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 268 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 269 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 270 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 271 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 272 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 273 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 274 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 275 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 276 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 277 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 278 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 279 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 280 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 281 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 282 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 283 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 284 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 285 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 286 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 287 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 288 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 289 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 290 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 291 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 292 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 293 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 294 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 295 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 296 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 297 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 298 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 299 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 300 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 301 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 302 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 303 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 304 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 305 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 306 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 307 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 308 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 309 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 310 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 311 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 312 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 313 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 314 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 315 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 316 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 317 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 318 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 319 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 320 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 321 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 322 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 323 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 324 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 325 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 326 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 327 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 328 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 329 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 330 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 331 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 332 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 333 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 334 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 335 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 336 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 337 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 338 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 339 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 340 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 341 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 342 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 343 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 344 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 345 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 346 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 347 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 348 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 349 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 350 |
+
},
|
| 351 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 352 |
+
},
|
| 353 |
+
"artifact_manifest": {
|
| 354 |
+
"Qwen3.5-9B/hle-verified-gold/configs.json": {
|
| 355 |
+
"size": 96465,
|
| 356 |
+
"sha256": "7bb994f1b342c98d66fcd1b1406ce3b2de3084d12b1a0e772cb9a7c3f7ad7b6e"
|
| 357 |
+
},
|
| 358 |
+
"Qwen3.5-9B/hle-verified-gold/events.jsonl.gz": {
|
| 359 |
+
"size": 43633633,
|
| 360 |
+
"sha256": "159496911a57230265fee4f420cc369011efbd4941ece2191c712d32329a5ad3"
|
| 361 |
+
},
|
| 362 |
+
"Qwen3.5-9B/hle-verified-gold/judge_trajectories.jsonl": {
|
| 363 |
+
"size": 806504,
|
| 364 |
+
"sha256": "842b1fec46cefcdf4d0230f2dd7a89b40bd31e45d248b15f2e2bbd2338b795a4"
|
| 365 |
+
},
|
| 366 |
+
"Qwen3.5-9B/hle-verified-gold/summary.json": {
|
| 367 |
+
"size": 1499,
|
| 368 |
+
"sha256": "8db7f858a18f933acb3c3462ff73f75b8304f55d59e010c85cd1b35e2ef07557"
|
| 369 |
+
},
|
| 370 |
+
"Qwen3.5-9B/hle-verified-gold/trajectories.jsonl": {
|
| 371 |
+
"size": 102604390,
|
| 372 |
+
"sha256": "9b5029d1a019217548daa2bcd550007616f1a6e585e2bd08c613ab045fd72ca3"
|
| 373 |
+
},
|
| 374 |
+
"Qwen3.5-9B/hle-verified-revision/configs.json": {
|
| 375 |
+
"size": 164897,
|
| 376 |
+
"sha256": "a6ceca950617a9cc8ca04eee77205df67d4fbfdcbf655df3da5002e4a0efe8ff"
|
| 377 |
+
},
|
| 378 |
+
"Qwen3.5-9B/hle-verified-revision/events.jsonl.gz": {
|
| 379 |
+
"size": 81502302,
|
| 380 |
+
"sha256": "cd44292482845ae3db5a22151cfb17ff800e2b68a456bec5eac8f9e4986069b5"
|
| 381 |
+
},
|
| 382 |
+
"Qwen3.5-9B/hle-verified-revision/judge_trajectories.jsonl": {
|
| 383 |
+
"size": 1458844,
|
| 384 |
+
"sha256": "26eb3d8ecaf7a961bb56442bb2cb44dff5f41649c5f34bbac284bf01d65b59c8"
|
| 385 |
+
},
|
| 386 |
+
"Qwen3.5-9B/hle-verified-revision/summary.json": {
|
| 387 |
+
"size": 1522,
|
| 388 |
+
"sha256": "60d1ed8703ec79e5ae80e64c6373da5e53ddacc48dec0a0decbc78e99cd51479"
|
| 389 |
+
},
|
| 390 |
+
"Qwen3.5-9B/hle-verified-revision/trajectories.jsonl": {
|
| 391 |
+
"size": 184438526,
|
| 392 |
+
"sha256": "394aaab563418001fbcc0c50b1c41a1ee0eabc5f72c34da393847a104699dffe"
|
| 393 |
+
}
|
| 394 |
+
}
|
| 395 |
+
}
|
Qwen3.5-9B/summary.json
CHANGED
|
@@ -1 +1,193 @@
|
|
| 1 |
-
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"frontierscience-olympiad": {
|
| 3 |
+
"responses": 3000,
|
| 4 |
+
"scored": 3000,
|
| 5 |
+
"status_counts": {
|
| 6 |
+
"completed": 3000
|
| 7 |
+
},
|
| 8 |
+
"correct_responses": 1600,
|
| 9 |
+
"accuracy_scored": 0.5333333333333333,
|
| 10 |
+
"normalized_mean_score": 0.5333333333333333,
|
| 11 |
+
"cap_hits": 269,
|
| 12 |
+
"cap_rate": 0.08966666666666667,
|
| 13 |
+
"average_output_tokens": 45270.22066666667,
|
| 14 |
+
"questions": 100,
|
| 15 |
+
"fully_scored_questions": 100,
|
| 16 |
+
"questions_mean_ge_0_7_fully_scored_only": 50,
|
| 17 |
+
"saved_judge_attempts": 3000
|
| 18 |
+
},
|
| 19 |
+
"frontierscience-research": {
|
| 20 |
+
"responses": 1800,
|
| 21 |
+
"scored": 1800,
|
| 22 |
+
"status_counts": {
|
| 23 |
+
"completed": 1800
|
| 24 |
+
},
|
| 25 |
+
"correct_responses": 58,
|
| 26 |
+
"accuracy_scored": 0.03222222222222222,
|
| 27 |
+
"normalized_mean_score": 0.2446551388888889,
|
| 28 |
+
"cap_hits": 13,
|
| 29 |
+
"cap_rate": 0.007222222222222222,
|
| 30 |
+
"average_output_tokens": 17541.786666666667,
|
| 31 |
+
"questions": 60,
|
| 32 |
+
"fully_scored_questions": 60,
|
| 33 |
+
"questions_mean_ge_0_7_fully_scored_only": 1,
|
| 34 |
+
"saved_judge_attempts": 1829
|
| 35 |
+
},
|
| 36 |
+
"arxivmath": {
|
| 37 |
+
"responses": 412,
|
| 38 |
+
"scored": 412,
|
| 39 |
+
"correct_responses": 109,
|
| 40 |
+
"accuracy_scored": 0.2645631067961165,
|
| 41 |
+
"normalized_mean_score": 0.2645631067961165,
|
| 42 |
+
"cap_hits": 42,
|
| 43 |
+
"average_output_tokens": 64449.953883495145,
|
| 44 |
+
"questions": 103,
|
| 45 |
+
"releases": {
|
| 46 |
+
"matharena-arxivmath-december": {
|
| 47 |
+
"responses": 68,
|
| 48 |
+
"scored": 68,
|
| 49 |
+
"status_counts": {
|
| 50 |
+
"completed": 68
|
| 51 |
+
},
|
| 52 |
+
"correct_responses": 24,
|
| 53 |
+
"accuracy_scored": 0.35294117647058826,
|
| 54 |
+
"normalized_mean_score": 0.35294117647058826,
|
| 55 |
+
"cap_hits": 9,
|
| 56 |
+
"cap_rate": 0.1323529411764706,
|
| 57 |
+
"average_output_tokens": 62851.08823529412,
|
| 58 |
+
"questions": 17,
|
| 59 |
+
"fully_scored_questions": 17,
|
| 60 |
+
"questions_mean_ge_0_7_fully_scored_only": 5,
|
| 61 |
+
"saved_judge_attempts": 0
|
| 62 |
+
},
|
| 63 |
+
"matharena-arxivmath-january": {
|
| 64 |
+
"responses": 92,
|
| 65 |
+
"scored": 92,
|
| 66 |
+
"status_counts": {
|
| 67 |
+
"completed": 92
|
| 68 |
+
},
|
| 69 |
+
"correct_responses": 31,
|
| 70 |
+
"accuracy_scored": 0.33695652173913043,
|
| 71 |
+
"normalized_mean_score": 0.33695652173913043,
|
| 72 |
+
"cap_hits": 14,
|
| 73 |
+
"cap_rate": 0.15217391304347827,
|
| 74 |
+
"average_output_tokens": 72861.66304347826,
|
| 75 |
+
"questions": 23,
|
| 76 |
+
"fully_scored_questions": 23,
|
| 77 |
+
"questions_mean_ge_0_7_fully_scored_only": 8,
|
| 78 |
+
"saved_judge_attempts": 0
|
| 79 |
+
},
|
| 80 |
+
"matharena-arxivmath-february": {
|
| 81 |
+
"responses": 128,
|
| 82 |
+
"scored": 128,
|
| 83 |
+
"status_counts": {
|
| 84 |
+
"completed": 128
|
| 85 |
+
},
|
| 86 |
+
"correct_responses": 28,
|
| 87 |
+
"accuracy_scored": 0.21875,
|
| 88 |
+
"normalized_mean_score": 0.21875,
|
| 89 |
+
"cap_hits": 15,
|
| 90 |
+
"cap_rate": 0.1171875,
|
| 91 |
+
"average_output_tokens": 64618.6328125,
|
| 92 |
+
"questions": 32,
|
| 93 |
+
"fully_scored_questions": 32,
|
| 94 |
+
"questions_mean_ge_0_7_fully_scored_only": 5,
|
| 95 |
+
"saved_judge_attempts": 0
|
| 96 |
+
},
|
| 97 |
+
"matharena-arxivmath-march": {
|
| 98 |
+
"responses": 124,
|
| 99 |
+
"scored": 124,
|
| 100 |
+
"status_counts": {
|
| 101 |
+
"completed": 124
|
| 102 |
+
},
|
| 103 |
+
"correct_responses": 26,
|
| 104 |
+
"accuracy_scored": 0.20967741935483872,
|
| 105 |
+
"normalized_mean_score": 0.20967741935483872,
|
| 106 |
+
"cap_hits": 4,
|
| 107 |
+
"cap_rate": 0.03225806451612903,
|
| 108 |
+
"average_output_tokens": 58911.68548387097,
|
| 109 |
+
"questions": 31,
|
| 110 |
+
"fully_scored_questions": 31,
|
| 111 |
+
"questions_mean_ge_0_7_fully_scored_only": 5,
|
| 112 |
+
"saved_judge_attempts": 0
|
| 113 |
+
}
|
| 114 |
+
}
|
| 115 |
+
},
|
| 116 |
+
"scicode-verified": {
|
| 117 |
+
"updated_at": 1792968289.8704367,
|
| 118 |
+
"expected_problems": 64,
|
| 119 |
+
"expected_steps": 287,
|
| 120 |
+
"generated_steps": 287,
|
| 121 |
+
"scored_steps": 287,
|
| 122 |
+
"passed_steps": 114,
|
| 123 |
+
"completed_problems": 64,
|
| 124 |
+
"passed_problems": 9,
|
| 125 |
+
"subproblem_accuracy": 0.397212543554007,
|
| 126 |
+
"main_problem_accuracy": 0.140625,
|
| 127 |
+
"complete": true,
|
| 128 |
+
"cap_hits": 1,
|
| 129 |
+
"mean_output_tokens": 7678.675958188153,
|
| 130 |
+
"grading_infrastructure_errors": 0,
|
| 131 |
+
"errors": []
|
| 132 |
+
},
|
| 133 |
+
"terminal-bench-2.1": {
|
| 134 |
+
"tasks": 89,
|
| 135 |
+
"attempts": 178,
|
| 136 |
+
"attempts_per_task": 2,
|
| 137 |
+
"metric": "avg@2",
|
| 138 |
+
"passed": 38,
|
| 139 |
+
"avg_at_2": 0.21348314606741572,
|
| 140 |
+
"unscored": 1,
|
| 141 |
+
"agent_timeouts": 83,
|
| 142 |
+
"reward_positive_with_exception": 0,
|
| 143 |
+
"scoring": "native_verifier",
|
| 144 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 145 |
+
"grading_complete": false,
|
| 146 |
+
"provisional": true,
|
| 147 |
+
"publication_date": "2026-10-27",
|
| 148 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/agentic-avg2-verifier-v11-20261026/qwen-it",
|
| 149 |
+
"timeout_zero_adjudications": [
|
| 150 |
+
"feal-differential-cryptanalysis--attempt-002",
|
| 151 |
+
"regex-chess--attempt-002",
|
| 152 |
+
"regex-log--attempt-001"
|
| 153 |
+
]
|
| 154 |
+
},
|
| 155 |
+
"scicoding-agent": {
|
| 156 |
+
"tasks": 50,
|
| 157 |
+
"attempts": 100,
|
| 158 |
+
"attempts_per_task": 2,
|
| 159 |
+
"metric": "avg@2",
|
| 160 |
+
"passed": 9,
|
| 161 |
+
"avg_at_2": 0.09,
|
| 162 |
+
"unscored": 0,
|
| 163 |
+
"agent_timeouts": 0,
|
| 164 |
+
"reward_positive_with_exception": 0,
|
| 165 |
+
"scoring": "native_verifier",
|
| 166 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1",
|
| 167 |
+
"grading_complete": true,
|
| 168 |
+
"provisional": false,
|
| 169 |
+
"publication_date": "2026-10-27",
|
| 170 |
+
"source_run": "/workspace/data/sci-llm-training/runtime/agentic-avg2-verifier-v11-20261026/qwen-it",
|
| 171 |
+
"timeout_zero_adjudications": []
|
| 172 |
+
},
|
| 173 |
+
"hle-verified": {
|
| 174 |
+
"gold": {
|
| 175 |
+
"correct": 170,
|
| 176 |
+
"total": 575,
|
| 177 |
+
"accuracy": 0.2956521739130435,
|
| 178 |
+
"cap_hits": 72
|
| 179 |
+
},
|
| 180 |
+
"revision": {
|
| 181 |
+
"correct": 289,
|
| 182 |
+
"total": 1025,
|
| 183 |
+
"accuracy": 0.2819512195121951,
|
| 184 |
+
"cap_hits": 115
|
| 185 |
+
},
|
| 186 |
+
"correct": 459,
|
| 187 |
+
"total": 1600,
|
| 188 |
+
"overall_accuracy": 0.286875,
|
| 189 |
+
"all_complete": true,
|
| 190 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 191 |
+
"provenance": "hle-verified-provenance.json"
|
| 192 |
+
}
|
| 193 |
+
}
|
README.md
CHANGED
|
@@ -41,6 +41,18 @@ Results are organized by model and benchmark. See each model README for protocol
|
|
| 41 |
See each model folder for complete benchmark summaries, frozen settings, full available trajectories and grading provenance.
|
| 42 |
|
| 43 |
|
| 44 |
-
## HLE-Verified
|
| 45 |
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
See each model folder for complete benchmark summaries, frozen settings, full available trajectories and grading provenance.
|
| 42 |
|
| 43 |
|
| 44 |
+
## HLE-Verified (2026-10-30)
|
| 45 |
|
| 46 |
+
HLE results are stored directly in each model folder under `hle-verified-gold/` and `hle-verified-revision/`. Each model README and summary includes HLE; `hle-verified-provenance.json` preserves that model’s configuration, data/source hashes and recovery evidence. There is no separate HLE results folder.
|
| 47 |
+
|
| 48 |
+
| Model | Gold | Revision | Overall (1,600) |
|
| 49 |
+
|---|---:|---:|---:|
|
| 50 |
+
| [Qwen3.5-35B-A3B](Qwen3.5-35B-A3B/README.md#hle-verified-2026-10-30) | 39.65% | 35.61% | 37.06% |
|
| 51 |
+
| [sft-agentic-stage2](sft-agentic-stage2/README.md#hle-verified-2026-10-30) | 28.70% | 32.68% | 31.25% |
|
| 52 |
+
| [gpt-oss-20b](gpt-oss-20b/README.md#hle-verified-2026-10-30) | 28.17% | 32.39% | 30.88% |
|
| 53 |
+
| [Gemma-4-12B-it](Gemma-4-12B-it/README.md#hle-verified-2026-10-30) | 27.65% | 29.95% | 29.12% |
|
| 54 |
+
| [Qwen3.5-9B](Qwen3.5-9B/README.md#hle-verified-2026-10-30) | 29.57% | 28.20% | 28.69% |
|
| 55 |
+
| [NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16](NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/README.md#hle-verified-2026-10-30) | 23.30% | 28.49% | 26.62% |
|
| 56 |
+
| [Olmo-3-32B-Think](Olmo-3-32B-Think/README.md#hle-verified-2026-10-30) | 12.00% | 21.46% | 18.06% |
|
| 57 |
+
|
| 58 |
+
All 11,200 answers are scored. Gold has 575 text questions and Revision 1,025 per model. Overall is question-weighted accuracy. OLMo uses its native 64K context / up to 60K output; other models use 96K output. See the model-level protocol caveats.
|
gpt-oss-20b/README.md
CHANGED
|
@@ -24,3 +24,17 @@ Research mean rubric and trajectory success are separate metrics. Agentic avg@2
|
|
| 24 |
| scicode-verified | Main-problem accuracy | 48.44% | 64/64 problems |
|
| 25 |
| terminal-bench-2.1 | avg@2 | 9.55% | 176/178 scored |
|
| 26 |
| scicoding-agent | avg@2 | 22.00% | 100/100 scored |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
| scicode-verified | Main-problem accuracy | 48.44% | 64/64 problems |
|
| 25 |
| terminal-bench-2.1 | avg@2 | 9.55% | 176/178 scored |
|
| 26 |
| scicoding-agent | avg@2 | 22.00% | 100/100 scored |
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
## HLE-Verified (2026-10-30)
|
| 30 |
+
|
| 31 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **494/1,600 (30.88%)**.
|
| 32 |
+
|
| 33 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 34 |
+
|---|---:|---:|---|
|
| 35 |
+
| Gold | 162/575 | 28.17% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 36 |
+
| Revision | 332/1025 | 32.39% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 37 |
+
|
| 38 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 98,304 tokens; context 131,072 tokens. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 39 |
+
|
| 40 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
gpt-oss-20b/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,406 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "gpt-oss-20b",
|
| 7 |
+
"folder": "gpt-oss-20b",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 162,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.2817391304347826,
|
| 13 |
+
"cap_hits": 72
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 332,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.32390243902439025,
|
| 19 |
+
"cap_hits": 150
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 494,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.30875
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "gpt-oss-20b",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "gpt-oss-20b",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "gpt-oss",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "gpt_oss",
|
| 35 |
+
"hidden_size": 2880,
|
| 36 |
+
"num_hidden_layers": 24,
|
| 37 |
+
"vocab_size": 201088
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 98304,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"top_k": -1,
|
| 46 |
+
"min_p": 0.0,
|
| 47 |
+
"presence_penalty": 0.0,
|
| 48 |
+
"repetition_penalty": 1.0,
|
| 49 |
+
"reasoning_effort": "high"
|
| 50 |
+
}
|
| 51 |
+
},
|
| 52 |
+
"serving": {
|
| 53 |
+
"context_length": 131072,
|
| 54 |
+
"dtype": "bfloat16",
|
| 55 |
+
"attention_backend": "fa3",
|
| 56 |
+
"mem_fraction_static": 0.85,
|
| 57 |
+
"reasoning_parser": "gpt-oss",
|
| 58 |
+
"extra_args": [
|
| 59 |
+
"--cuda-graph-max-bs-decode",
|
| 60 |
+
"64",
|
| 61 |
+
"--tool-call-parser",
|
| 62 |
+
"gpt-oss",
|
| 63 |
+
"--moe-runner-backend",
|
| 64 |
+
"triton_kernel",
|
| 65 |
+
"--disable-radix-cache"
|
| 66 |
+
],
|
| 67 |
+
"tp": 1,
|
| 68 |
+
"replicas_per_node": 8,
|
| 69 |
+
"max_running_requests": 64
|
| 70 |
+
},
|
| 71 |
+
"reasoning": {
|
| 72 |
+
"requests_per_engine": 64,
|
| 73 |
+
"benchmarks": {}
|
| 74 |
+
},
|
| 75 |
+
"agentic_tasks_per_engine": 1
|
| 76 |
+
},
|
| 77 |
+
"node": 0,
|
| 78 |
+
"checkpoint": "/workspace/data/raw_model/gpt-oss-20b-6cee5e81ee83",
|
| 79 |
+
"repo": "openai/gpt-oss-20b",
|
| 80 |
+
"revision": "6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 81 |
+
"checkpoint_config_sha256": "3a2a26ded679375b7928ddeca59764df7cea83220c1961035f6d6e232659e9ce",
|
| 82 |
+
"index_sha256": "0e085b977c4c9942f85938828e8c989ed7d5cdabf852e4da6a67c116cd502cd1",
|
| 83 |
+
"shard_sizes": {
|
| 84 |
+
"model-00000-of-00002.safetensors": 4792272488,
|
| 85 |
+
"model-00001-of-00002.safetensors": 4798702184,
|
| 86 |
+
"model-00002-of-00002.safetensors": 4170342232
|
| 87 |
+
},
|
| 88 |
+
"endpoints": [
|
| 89 |
+
{
|
| 90 |
+
"id": "node-0-replica-0",
|
| 91 |
+
"api_base": "http://10.2.5.84:8700/v1"
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"id": "node-0-replica-1",
|
| 95 |
+
"api_base": "http://10.2.5.84:8701/v1"
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"id": "node-0-replica-2",
|
| 99 |
+
"api_base": "http://10.2.5.84:8702/v1"
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"id": "node-0-replica-3",
|
| 103 |
+
"api_base": "http://10.2.5.84:8703/v1"
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"id": "node-0-replica-4",
|
| 107 |
+
"api_base": "http://10.2.5.84:8704/v1"
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"id": "node-0-replica-5",
|
| 111 |
+
"api_base": "http://10.2.5.84:8705/v1"
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"id": "node-0-replica-6",
|
| 115 |
+
"api_base": "http://10.2.5.84:8706/v1"
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"id": "node-0-replica-7",
|
| 119 |
+
"api_base": "http://10.2.5.84:8707/v1"
|
| 120 |
+
}
|
| 121 |
+
],
|
| 122 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gpt-oss-20b",
|
| 123 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gpt-oss-20b/configs/hle-verified-gold-revision-96k.yaml",
|
| 124 |
+
"concurrency": 512,
|
| 125 |
+
"judge_concurrency": 64,
|
| 126 |
+
"tracking": {
|
| 127 |
+
"enabled": false,
|
| 128 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 129 |
+
},
|
| 130 |
+
"config_sha256": {
|
| 131 |
+
"model.yaml": "8a8bd3d39ee530a401e8004ba6dea0bf05164689556521205b18fe1dffb6b6f8",
|
| 132 |
+
"configs/hle-verified-gold-revision-96k.yaml": "05098321700593bc74123855f7b4d970f0c9718c72d378827e16b8882de1072f",
|
| 133 |
+
"configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4",
|
| 134 |
+
"configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"
|
| 135 |
+
}
|
| 136 |
+
},
|
| 137 |
+
"dataset": {
|
| 138 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 139 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 140 |
+
"counts": {
|
| 141 |
+
"Gold": {
|
| 142 |
+
"total": 668,
|
| 143 |
+
"text": 575,
|
| 144 |
+
"image_excluded": 93
|
| 145 |
+
},
|
| 146 |
+
"Revision": {
|
| 147 |
+
"total": 1143,
|
| 148 |
+
"text": 1025,
|
| 149 |
+
"image_excluded": 118
|
| 150 |
+
}
|
| 151 |
+
},
|
| 152 |
+
"sha256": {
|
| 153 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 154 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 155 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 156 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 157 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 158 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 159 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 160 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 161 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 162 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 163 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 164 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 165 |
+
}
|
| 166 |
+
},
|
| 167 |
+
"token_preflight": {
|
| 168 |
+
"model": "gpt-oss-20b",
|
| 169 |
+
"questions": 1600,
|
| 170 |
+
"max_prompt_tokens": 12308,
|
| 171 |
+
"output_cap": 98304,
|
| 172 |
+
"context": 131072,
|
| 173 |
+
"dynamic_output_budget": false,
|
| 174 |
+
"affected_questions": 0,
|
| 175 |
+
"min_effective_output_cap": 98304
|
| 176 |
+
},
|
| 177 |
+
"shared_recovery_provenance": {
|
| 178 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 179 |
+
"publication_reruns": false,
|
| 180 |
+
"all_questions_scored": true,
|
| 181 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 182 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 183 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 184 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 185 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 186 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 187 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 188 |
+
},
|
| 189 |
+
"source_manifest": {
|
| 190 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 191 |
+
"files": {
|
| 192 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 193 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 194 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 195 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 196 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 197 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 198 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 199 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 200 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 201 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 202 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 203 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 204 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 205 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 206 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 207 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 208 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 209 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 210 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 211 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 212 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 213 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 214 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 215 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 216 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 217 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 218 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 219 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 220 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 221 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 222 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 223 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 224 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 225 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 226 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 227 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 228 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 229 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 230 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 231 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 232 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 233 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 234 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 235 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 236 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 237 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 238 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 239 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 240 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 241 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 242 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 243 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 244 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 245 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 246 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 247 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 248 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 249 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 250 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 251 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 252 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 253 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 254 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 255 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 256 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 257 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 258 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 259 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 260 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 261 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 262 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 263 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 264 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 265 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 266 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 267 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 268 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 269 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 270 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 271 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 272 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 273 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 274 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 275 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 276 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 277 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 278 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 279 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 280 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 281 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 282 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 283 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 284 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 285 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 286 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 287 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 288 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 289 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 290 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 291 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 292 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 293 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 294 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 295 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 296 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 297 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 298 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 299 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 300 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 301 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 302 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 303 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 304 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 305 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 306 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 307 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 308 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 309 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 310 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 311 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 312 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 313 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 314 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 315 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 316 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 317 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 318 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 319 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 320 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 321 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 322 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 323 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 324 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 325 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 326 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 327 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 328 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 329 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 330 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 331 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 332 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 333 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 334 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 335 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 336 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 337 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 338 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 339 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 340 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 341 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 342 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 343 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 344 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 345 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 346 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 347 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 348 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 349 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 350 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 351 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 352 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 353 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 354 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 355 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 356 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 357 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 358 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 359 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 360 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 361 |
+
},
|
| 362 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 363 |
+
},
|
| 364 |
+
"artifact_manifest": {
|
| 365 |
+
"gpt-oss-20b/hle-verified-gold/configs.json": {
|
| 366 |
+
"size": 96732,
|
| 367 |
+
"sha256": "aba3464b730503bbf359c8e7ccb02197c6a3c2acc74ad1080d8de16588d95970"
|
| 368 |
+
},
|
| 369 |
+
"gpt-oss-20b/hle-verified-gold/events.jsonl.gz": {
|
| 370 |
+
"size": 49832341,
|
| 371 |
+
"sha256": "187167c4c0cf1fdb45836d03b1f74ddfdc2442cfc7f275fe610c0a2e6c552ebb"
|
| 372 |
+
},
|
| 373 |
+
"gpt-oss-20b/hle-verified-gold/judge_trajectories.jsonl": {
|
| 374 |
+
"size": 807469,
|
| 375 |
+
"sha256": "70ed778048e88437e3a6a99261df5320349f4ce0ae68c2ee86a2cfbf1b5cd28f"
|
| 376 |
+
},
|
| 377 |
+
"gpt-oss-20b/hle-verified-gold/summary.json": {
|
| 378 |
+
"size": 1492,
|
| 379 |
+
"sha256": "1328061cbb354b82e82b47c95fd8a30564fbb65835ced09e9861dcdb0c067cb9"
|
| 380 |
+
},
|
| 381 |
+
"gpt-oss-20b/hle-verified-gold/trajectories.jsonl": {
|
| 382 |
+
"size": 89318447,
|
| 383 |
+
"sha256": "498119901fa4e4c1bbc22231744abe3a20db7211cc9e6f8664a82ce01cbe4320"
|
| 384 |
+
},
|
| 385 |
+
"gpt-oss-20b/hle-verified-revision/configs.json": {
|
| 386 |
+
"size": 165164,
|
| 387 |
+
"sha256": "599ef040aed792ba77a02878391c4bc26ba24105defb488ff7d061b9d4443ee7"
|
| 388 |
+
},
|
| 389 |
+
"gpt-oss-20b/hle-verified-revision/events.jsonl.gz": {
|
| 390 |
+
"size": 90826955,
|
| 391 |
+
"sha256": "812fb9e7d831676b8e9d1aa0a36166edc97ce7cd4f31e207313dff4ec7deaed6"
|
| 392 |
+
},
|
| 393 |
+
"gpt-oss-20b/hle-verified-revision/judge_trajectories.jsonl": {
|
| 394 |
+
"size": 1455964,
|
| 395 |
+
"sha256": "86e0dac921215fcff2913d381d06b903084c8d380070142013e1590a3866d846"
|
| 396 |
+
},
|
| 397 |
+
"gpt-oss-20b/hle-verified-revision/summary.json": {
|
| 398 |
+
"size": 1516,
|
| 399 |
+
"sha256": "a94c187e6ae8bdb9e49125767b2d847899d7e97505af4e6d503cd052d02e0ab5"
|
| 400 |
+
},
|
| 401 |
+
"gpt-oss-20b/hle-verified-revision/trajectories.jsonl": {
|
| 402 |
+
"size": 161160950,
|
| 403 |
+
"sha256": "e24efccf23a011a6caa879a88650726f8abd29dd2592a12494052493f786cd1e"
|
| 404 |
+
}
|
| 405 |
+
}
|
| 406 |
+
}
|
gpt-oss-20b/summary.json
CHANGED
|
@@ -1 +1,467 @@
|
|
| 1 |
-
{"arxivmath": {"benchmark": "arxivmath", "releases": [{"total_records": 68, "completed": 68, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.20588235294117646, "accuracy": 0.20588235294117646, "mean_latency_seconds": 1510.148789786398, "finish_reason_counts": {"length": 32, "stop": 36}, "length_limited_response_count": 32, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.20588235294117646, "competition_metrics": {"arxiv/december": {"n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.20588235294117646}}}, "benchmark": "matharena-arxivmath-december", "model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee", "execution_mode": "non_agent", "generation_hash": "e85d62c2830fb58605e1076026c76e867d11a7c4055b3366d22e5759f026c944", "scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661", "selected": 68, "incomplete": 0, "generated_now": 68, "scored_now": 68, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 68, "generation_concurrency": 1024, "judge_concurrency": 256, "updated_at": "2026-10-30T06:34:15.411748+00:00"}, {"total_records": 128, "completed": 128, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.171875, "accuracy": 0.171875, "mean_latency_seconds": 1599.4384132545292, "finish_reason_counts": {"length": 56, "stop": 72}, "length_limited_response_count": 56, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.171875, "competition_metrics": {"arxiv/february": {"n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.171875}}}, "benchmark": "matharena-arxivmath-february", "model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee", "execution_mode": "non_agent", "generation_hash": "1edb8159c0bb3665051eb8ab6ef37969ff50d968e36c1e30976f28ac375ac9e9", "scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5", "selected": 128, "incomplete": 0, "generated_now": 128, "scored_now": 128, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 128, "generation_concurrency": 1024, "judge_concurrency": 256, "updated_at": "2026-10-30T06:41:28.674317+00:00"}, {"total_records": 92, "completed": 92, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.2717391304347826, "accuracy": 0.2717391304347826, "mean_latency_seconds": 1284.2728273393473, "finish_reason_counts": {"length": 35, "stop": 57}, "length_limited_response_count": 35, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.2717391304347826, "competition_metrics": {"arxiv/january": {"n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.2717391304347826}}}, "benchmark": "matharena-arxivmath-january", "model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee", "execution_mode": "non_agent", "generation_hash": "3850df94c6726f23eb4dec3a80faad0a684b3b8de847c52fb0e5f39ee6e1fb4e", "scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8", "selected": 92, "incomplete": 0, "generated_now": 92, "scored_now": 92, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 92, "generation_concurrency": 1024, "judge_concurrency": 256, "updated_at": "2026-10-30T06:16:56.156107+00:00"}, {"total_records": 124, "completed": 124, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.21774193548387097, "accuracy": 0.21774193548387097, "mean_latency_seconds": 998.8992833425948, "finish_reason_counts": {"length": 25, "stop": 99}, "length_limited_response_count": 25, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.21774193548387097, "competition_metrics": {"arxiv/march": {"n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.21774193548387097}}}, "benchmark": "matharena-arxivmath-march", "model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee", "execution_mode": "non_agent", "generation_hash": "273290cabd1298769c7bca237f225bec6f3f044fd479e35af01434c07561a57f", "scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8", "selected": 124, "incomplete": 0, "generated_now": 124, "scored_now": 124, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 124, "generation_concurrency": 1024, "judge_concurrency": 256, "updated_at": "2026-10-30T06:16:58.988534+00:00"}], "total_records": 412, "completed": 412, "accuracy": 0.21359223300970873}, "frontierscience-olympiad": {"total_records": 3000, "completed": 3000, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.5386666666666666, "accuracy": 0.5386666666666666, "mean_latency_seconds": 841.7236672996808, "finish_reason_counts": {"length": 147, "stop": 2853}, "length_limited_response_count": 147, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 3000, "request_count": 3000, "retry_request_count": 0, "prompt_tokens": 2268631, "cached_prompt_tokens": 22526, "cache_write_prompt_tokens": 609141, "completion_tokens": 686180, "reasoning_tokens": 478227, "total_tokens": 2954811}, "judge_api_usage": {"request_count": 3000, "prompt_tokens": 2268631, "cached_prompt_tokens": 22526, "cache_write_prompt_tokens": 609141, "completion_tokens": 686180, "reasoning_tokens": 478227, "total_tokens": 2954811}, "benchmark_metrics": {"track": "olympiad", "n_problems_scored": 100, "n_trajectories_scored": 3000, "aggregation": "avg@30", "reported_accuracy": 0.5386666666666666, "mean_rubric_score": null, "success_threshold": 0.5, "community_profile": null}, "benchmark": "frontierscience-olympiad", "model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee", "execution_mode": "non_agent", "generation_hash": "c2ae5873170f9ecc765909ab207406edaae2683668e52e3a6f3a71646bc87d18", "scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016", "selected": 3000, "incomplete": 0, "generated_now": 3000, "scored_now": 3000, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 3000, "generation_concurrency": 1024, "judge_concurrency": 256, "updated_at": "2026-10-30T07:12:45.955152+00:00"}, "frontierscience-research": {"total_records": 1800, "completed": 1799, "query_errors": 0, "score_errors": 1, "generated_unscored": 0, "mean_score": 0.32770525291828795, "accuracy": 0.07226236798221233, "mean_latency_seconds": 569.9840808942698, "finish_reason_counts": {"length": 18, "stop": 1781}, "length_limited_response_count": 18, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 1799, "request_count": 1821, "retry_request_count": 22, "prompt_tokens": 9816612, "cached_prompt_tokens": 120976, "cache_write_prompt_tokens": 9676591, "completion_tokens": 3183991, "reasoning_tokens": 1921665, "total_tokens": 13000603}, "judge_api_usage": {"request_count": 1822, "prompt_tokens": 9834515, "cached_prompt_tokens": 120976, "cache_write_prompt_tokens": 9694491, "completion_tokens": 3192183, "reasoning_tokens": 1929857, "total_tokens": 13026698}, "benchmark_metrics": {"track": "research", "n_problems_scored": 60, "n_trajectories_scored": 1799, "aggregation": "average_trials", "reported_accuracy": 0.07222222222222222, "mean_rubric_score": 0.3275955603448276, "success_threshold": 0.7, "community_profile": null}, "benchmark": "frontierscience-research", "model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee", "execution_mode": "non_agent", "generation_hash": "b86f7a8a1172f905444f6d2df6a245fbc642f440479171932b5e425e54e1353e", "scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750", "selected": 1800, "incomplete": 1, "generated_now": 1800, "scored_now": 1799, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 1800, "generation_concurrency": 1024, "judge_concurrency": 256, "updated_at": "2026-10-30T06:46:49.734658+00:00"}, "scicode-verified": {"updated_at": 1793344949.0352376, "expected_problems": 64, "expected_steps": 287, "generated_steps": 287, "scored_steps": 287, "passed_steps": 197, "completed_problems": 64, "passed_problems": 31, "subproblem_accuracy": 0.686411149825784, "main_problem_accuracy": 0.484375, "complete": true, "cap_hits": 1, "mean_output_tokens": 8765.648083623693, "grading_infrastructure_errors": 0, "errors": []}, "scicoding-agent": {"tasks": 50, "attempts": 100, "attempts_per_task": 2, "metric": "avg@2", "passed": 22, "avg_at_2": 0.22, "unscored": 0, "agent_timeouts": 0, "reward_positive_with_exception": 0, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1"}, "terminal-bench-2.1": {"tasks": 89, "attempts": 178, "attempts_per_task": 2, "metric": "avg@2", "passed": 17, "avg_at_2": 0.09550561797752809, "unscored": 2, "agent_timeouts": 12, "reward_positive_with_exception": 1, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1"}}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"arxivmath": {
|
| 3 |
+
"benchmark": "arxivmath",
|
| 4 |
+
"releases": [
|
| 5 |
+
{
|
| 6 |
+
"total_records": 68,
|
| 7 |
+
"completed": 68,
|
| 8 |
+
"query_errors": 0,
|
| 9 |
+
"score_errors": 0,
|
| 10 |
+
"generated_unscored": 0,
|
| 11 |
+
"mean_score": 0.20588235294117646,
|
| 12 |
+
"accuracy": 0.20588235294117646,
|
| 13 |
+
"mean_latency_seconds": 1510.148789786398,
|
| 14 |
+
"finish_reason_counts": {
|
| 15 |
+
"length": 32,
|
| 16 |
+
"stop": 36
|
| 17 |
+
},
|
| 18 |
+
"length_limited_response_count": 32,
|
| 19 |
+
"observed_max_completion_tokens": 98304,
|
| 20 |
+
"judge_usage": {
|
| 21 |
+
"judged_record_count": 0,
|
| 22 |
+
"request_count": 0,
|
| 23 |
+
"retry_request_count": 0,
|
| 24 |
+
"prompt_tokens": 0,
|
| 25 |
+
"cached_prompt_tokens": 0,
|
| 26 |
+
"cache_write_prompt_tokens": 0,
|
| 27 |
+
"completion_tokens": 0,
|
| 28 |
+
"reasoning_tokens": 0,
|
| 29 |
+
"total_tokens": 0
|
| 30 |
+
},
|
| 31 |
+
"judge_api_usage": {
|
| 32 |
+
"request_count": 0,
|
| 33 |
+
"prompt_tokens": 0,
|
| 34 |
+
"cached_prompt_tokens": 0,
|
| 35 |
+
"cache_write_prompt_tokens": 0,
|
| 36 |
+
"completion_tokens": 0,
|
| 37 |
+
"reasoning_tokens": 0,
|
| 38 |
+
"total_tokens": 0
|
| 39 |
+
},
|
| 40 |
+
"benchmark_metrics": {
|
| 41 |
+
"competition": "arxivmath",
|
| 42 |
+
"aggregation": "avg@4",
|
| 43 |
+
"n_problems_scored": 17,
|
| 44 |
+
"n_trajectories_scored": 68,
|
| 45 |
+
"reported_accuracy": 0.20588235294117646,
|
| 46 |
+
"competition_metrics": {
|
| 47 |
+
"arxiv/december": {
|
| 48 |
+
"n_problems_scored": 17,
|
| 49 |
+
"n_trajectories_scored": 68,
|
| 50 |
+
"reported_accuracy": 0.20588235294117646
|
| 51 |
+
}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"benchmark": "matharena-arxivmath-december",
|
| 55 |
+
"model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 56 |
+
"execution_mode": "non_agent",
|
| 57 |
+
"generation_hash": "e85d62c2830fb58605e1076026c76e867d11a7c4055b3366d22e5759f026c944",
|
| 58 |
+
"scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661",
|
| 59 |
+
"selected": 68,
|
| 60 |
+
"incomplete": 0,
|
| 61 |
+
"generated_now": 68,
|
| 62 |
+
"scored_now": 68,
|
| 63 |
+
"cached_completed": 0,
|
| 64 |
+
"stage": "all",
|
| 65 |
+
"worker_count": 1,
|
| 66 |
+
"worker_index": 0,
|
| 67 |
+
"worker_owned_predictions": 68,
|
| 68 |
+
"generation_concurrency": 1024,
|
| 69 |
+
"judge_concurrency": 256,
|
| 70 |
+
"updated_at": "2026-10-30T06:34:15.411748+00:00"
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"total_records": 128,
|
| 74 |
+
"completed": 128,
|
| 75 |
+
"query_errors": 0,
|
| 76 |
+
"score_errors": 0,
|
| 77 |
+
"generated_unscored": 0,
|
| 78 |
+
"mean_score": 0.171875,
|
| 79 |
+
"accuracy": 0.171875,
|
| 80 |
+
"mean_latency_seconds": 1599.4384132545292,
|
| 81 |
+
"finish_reason_counts": {
|
| 82 |
+
"length": 56,
|
| 83 |
+
"stop": 72
|
| 84 |
+
},
|
| 85 |
+
"length_limited_response_count": 56,
|
| 86 |
+
"observed_max_completion_tokens": 98304,
|
| 87 |
+
"judge_usage": {
|
| 88 |
+
"judged_record_count": 0,
|
| 89 |
+
"request_count": 0,
|
| 90 |
+
"retry_request_count": 0,
|
| 91 |
+
"prompt_tokens": 0,
|
| 92 |
+
"cached_prompt_tokens": 0,
|
| 93 |
+
"cache_write_prompt_tokens": 0,
|
| 94 |
+
"completion_tokens": 0,
|
| 95 |
+
"reasoning_tokens": 0,
|
| 96 |
+
"total_tokens": 0
|
| 97 |
+
},
|
| 98 |
+
"judge_api_usage": {
|
| 99 |
+
"request_count": 0,
|
| 100 |
+
"prompt_tokens": 0,
|
| 101 |
+
"cached_prompt_tokens": 0,
|
| 102 |
+
"cache_write_prompt_tokens": 0,
|
| 103 |
+
"completion_tokens": 0,
|
| 104 |
+
"reasoning_tokens": 0,
|
| 105 |
+
"total_tokens": 0
|
| 106 |
+
},
|
| 107 |
+
"benchmark_metrics": {
|
| 108 |
+
"competition": "arxivmath",
|
| 109 |
+
"aggregation": "avg@4",
|
| 110 |
+
"n_problems_scored": 32,
|
| 111 |
+
"n_trajectories_scored": 128,
|
| 112 |
+
"reported_accuracy": 0.171875,
|
| 113 |
+
"competition_metrics": {
|
| 114 |
+
"arxiv/february": {
|
| 115 |
+
"n_problems_scored": 32,
|
| 116 |
+
"n_trajectories_scored": 128,
|
| 117 |
+
"reported_accuracy": 0.171875
|
| 118 |
+
}
|
| 119 |
+
}
|
| 120 |
+
},
|
| 121 |
+
"benchmark": "matharena-arxivmath-february",
|
| 122 |
+
"model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 123 |
+
"execution_mode": "non_agent",
|
| 124 |
+
"generation_hash": "1edb8159c0bb3665051eb8ab6ef37969ff50d968e36c1e30976f28ac375ac9e9",
|
| 125 |
+
"scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5",
|
| 126 |
+
"selected": 128,
|
| 127 |
+
"incomplete": 0,
|
| 128 |
+
"generated_now": 128,
|
| 129 |
+
"scored_now": 128,
|
| 130 |
+
"cached_completed": 0,
|
| 131 |
+
"stage": "all",
|
| 132 |
+
"worker_count": 1,
|
| 133 |
+
"worker_index": 0,
|
| 134 |
+
"worker_owned_predictions": 128,
|
| 135 |
+
"generation_concurrency": 1024,
|
| 136 |
+
"judge_concurrency": 256,
|
| 137 |
+
"updated_at": "2026-10-30T06:41:28.674317+00:00"
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"total_records": 92,
|
| 141 |
+
"completed": 92,
|
| 142 |
+
"query_errors": 0,
|
| 143 |
+
"score_errors": 0,
|
| 144 |
+
"generated_unscored": 0,
|
| 145 |
+
"mean_score": 0.2717391304347826,
|
| 146 |
+
"accuracy": 0.2717391304347826,
|
| 147 |
+
"mean_latency_seconds": 1284.2728273393473,
|
| 148 |
+
"finish_reason_counts": {
|
| 149 |
+
"length": 35,
|
| 150 |
+
"stop": 57
|
| 151 |
+
},
|
| 152 |
+
"length_limited_response_count": 35,
|
| 153 |
+
"observed_max_completion_tokens": 98304,
|
| 154 |
+
"judge_usage": {
|
| 155 |
+
"judged_record_count": 0,
|
| 156 |
+
"request_count": 0,
|
| 157 |
+
"retry_request_count": 0,
|
| 158 |
+
"prompt_tokens": 0,
|
| 159 |
+
"cached_prompt_tokens": 0,
|
| 160 |
+
"cache_write_prompt_tokens": 0,
|
| 161 |
+
"completion_tokens": 0,
|
| 162 |
+
"reasoning_tokens": 0,
|
| 163 |
+
"total_tokens": 0
|
| 164 |
+
},
|
| 165 |
+
"judge_api_usage": {
|
| 166 |
+
"request_count": 0,
|
| 167 |
+
"prompt_tokens": 0,
|
| 168 |
+
"cached_prompt_tokens": 0,
|
| 169 |
+
"cache_write_prompt_tokens": 0,
|
| 170 |
+
"completion_tokens": 0,
|
| 171 |
+
"reasoning_tokens": 0,
|
| 172 |
+
"total_tokens": 0
|
| 173 |
+
},
|
| 174 |
+
"benchmark_metrics": {
|
| 175 |
+
"competition": "arxivmath",
|
| 176 |
+
"aggregation": "avg@4",
|
| 177 |
+
"n_problems_scored": 23,
|
| 178 |
+
"n_trajectories_scored": 92,
|
| 179 |
+
"reported_accuracy": 0.2717391304347826,
|
| 180 |
+
"competition_metrics": {
|
| 181 |
+
"arxiv/january": {
|
| 182 |
+
"n_problems_scored": 23,
|
| 183 |
+
"n_trajectories_scored": 92,
|
| 184 |
+
"reported_accuracy": 0.2717391304347826
|
| 185 |
+
}
|
| 186 |
+
}
|
| 187 |
+
},
|
| 188 |
+
"benchmark": "matharena-arxivmath-january",
|
| 189 |
+
"model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 190 |
+
"execution_mode": "non_agent",
|
| 191 |
+
"generation_hash": "3850df94c6726f23eb4dec3a80faad0a684b3b8de847c52fb0e5f39ee6e1fb4e",
|
| 192 |
+
"scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8",
|
| 193 |
+
"selected": 92,
|
| 194 |
+
"incomplete": 0,
|
| 195 |
+
"generated_now": 92,
|
| 196 |
+
"scored_now": 92,
|
| 197 |
+
"cached_completed": 0,
|
| 198 |
+
"stage": "all",
|
| 199 |
+
"worker_count": 1,
|
| 200 |
+
"worker_index": 0,
|
| 201 |
+
"worker_owned_predictions": 92,
|
| 202 |
+
"generation_concurrency": 1024,
|
| 203 |
+
"judge_concurrency": 256,
|
| 204 |
+
"updated_at": "2026-10-30T06:16:56.156107+00:00"
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"total_records": 124,
|
| 208 |
+
"completed": 124,
|
| 209 |
+
"query_errors": 0,
|
| 210 |
+
"score_errors": 0,
|
| 211 |
+
"generated_unscored": 0,
|
| 212 |
+
"mean_score": 0.21774193548387097,
|
| 213 |
+
"accuracy": 0.21774193548387097,
|
| 214 |
+
"mean_latency_seconds": 998.8992833425948,
|
| 215 |
+
"finish_reason_counts": {
|
| 216 |
+
"length": 25,
|
| 217 |
+
"stop": 99
|
| 218 |
+
},
|
| 219 |
+
"length_limited_response_count": 25,
|
| 220 |
+
"observed_max_completion_tokens": 98304,
|
| 221 |
+
"judge_usage": {
|
| 222 |
+
"judged_record_count": 0,
|
| 223 |
+
"request_count": 0,
|
| 224 |
+
"retry_request_count": 0,
|
| 225 |
+
"prompt_tokens": 0,
|
| 226 |
+
"cached_prompt_tokens": 0,
|
| 227 |
+
"cache_write_prompt_tokens": 0,
|
| 228 |
+
"completion_tokens": 0,
|
| 229 |
+
"reasoning_tokens": 0,
|
| 230 |
+
"total_tokens": 0
|
| 231 |
+
},
|
| 232 |
+
"judge_api_usage": {
|
| 233 |
+
"request_count": 0,
|
| 234 |
+
"prompt_tokens": 0,
|
| 235 |
+
"cached_prompt_tokens": 0,
|
| 236 |
+
"cache_write_prompt_tokens": 0,
|
| 237 |
+
"completion_tokens": 0,
|
| 238 |
+
"reasoning_tokens": 0,
|
| 239 |
+
"total_tokens": 0
|
| 240 |
+
},
|
| 241 |
+
"benchmark_metrics": {
|
| 242 |
+
"competition": "arxivmath",
|
| 243 |
+
"aggregation": "avg@4",
|
| 244 |
+
"n_problems_scored": 31,
|
| 245 |
+
"n_trajectories_scored": 124,
|
| 246 |
+
"reported_accuracy": 0.21774193548387097,
|
| 247 |
+
"competition_metrics": {
|
| 248 |
+
"arxiv/march": {
|
| 249 |
+
"n_problems_scored": 31,
|
| 250 |
+
"n_trajectories_scored": 124,
|
| 251 |
+
"reported_accuracy": 0.21774193548387097
|
| 252 |
+
}
|
| 253 |
+
}
|
| 254 |
+
},
|
| 255 |
+
"benchmark": "matharena-arxivmath-march",
|
| 256 |
+
"model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 257 |
+
"execution_mode": "non_agent",
|
| 258 |
+
"generation_hash": "273290cabd1298769c7bca237f225bec6f3f044fd479e35af01434c07561a57f",
|
| 259 |
+
"scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8",
|
| 260 |
+
"selected": 124,
|
| 261 |
+
"incomplete": 0,
|
| 262 |
+
"generated_now": 124,
|
| 263 |
+
"scored_now": 124,
|
| 264 |
+
"cached_completed": 0,
|
| 265 |
+
"stage": "all",
|
| 266 |
+
"worker_count": 1,
|
| 267 |
+
"worker_index": 0,
|
| 268 |
+
"worker_owned_predictions": 124,
|
| 269 |
+
"generation_concurrency": 1024,
|
| 270 |
+
"judge_concurrency": 256,
|
| 271 |
+
"updated_at": "2026-10-30T06:16:58.988534+00:00"
|
| 272 |
+
}
|
| 273 |
+
],
|
| 274 |
+
"total_records": 412,
|
| 275 |
+
"completed": 412,
|
| 276 |
+
"accuracy": 0.21359223300970873
|
| 277 |
+
},
|
| 278 |
+
"frontierscience-olympiad": {
|
| 279 |
+
"total_records": 3000,
|
| 280 |
+
"completed": 3000,
|
| 281 |
+
"query_errors": 0,
|
| 282 |
+
"score_errors": 0,
|
| 283 |
+
"generated_unscored": 0,
|
| 284 |
+
"mean_score": 0.5386666666666666,
|
| 285 |
+
"accuracy": 0.5386666666666666,
|
| 286 |
+
"mean_latency_seconds": 841.7236672996808,
|
| 287 |
+
"finish_reason_counts": {
|
| 288 |
+
"length": 147,
|
| 289 |
+
"stop": 2853
|
| 290 |
+
},
|
| 291 |
+
"length_limited_response_count": 147,
|
| 292 |
+
"observed_max_completion_tokens": 98304,
|
| 293 |
+
"judge_usage": {
|
| 294 |
+
"judged_record_count": 3000,
|
| 295 |
+
"request_count": 3000,
|
| 296 |
+
"retry_request_count": 0,
|
| 297 |
+
"prompt_tokens": 2268631,
|
| 298 |
+
"cached_prompt_tokens": 22526,
|
| 299 |
+
"cache_write_prompt_tokens": 609141,
|
| 300 |
+
"completion_tokens": 686180,
|
| 301 |
+
"reasoning_tokens": 478227,
|
| 302 |
+
"total_tokens": 2954811
|
| 303 |
+
},
|
| 304 |
+
"judge_api_usage": {
|
| 305 |
+
"request_count": 3000,
|
| 306 |
+
"prompt_tokens": 2268631,
|
| 307 |
+
"cached_prompt_tokens": 22526,
|
| 308 |
+
"cache_write_prompt_tokens": 609141,
|
| 309 |
+
"completion_tokens": 686180,
|
| 310 |
+
"reasoning_tokens": 478227,
|
| 311 |
+
"total_tokens": 2954811
|
| 312 |
+
},
|
| 313 |
+
"benchmark_metrics": {
|
| 314 |
+
"track": "olympiad",
|
| 315 |
+
"n_problems_scored": 100,
|
| 316 |
+
"n_trajectories_scored": 3000,
|
| 317 |
+
"aggregation": "avg@30",
|
| 318 |
+
"reported_accuracy": 0.5386666666666666,
|
| 319 |
+
"mean_rubric_score": null,
|
| 320 |
+
"success_threshold": 0.5,
|
| 321 |
+
"community_profile": null
|
| 322 |
+
},
|
| 323 |
+
"benchmark": "frontierscience-olympiad",
|
| 324 |
+
"model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 325 |
+
"execution_mode": "non_agent",
|
| 326 |
+
"generation_hash": "c2ae5873170f9ecc765909ab207406edaae2683668e52e3a6f3a71646bc87d18",
|
| 327 |
+
"scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016",
|
| 328 |
+
"selected": 3000,
|
| 329 |
+
"incomplete": 0,
|
| 330 |
+
"generated_now": 3000,
|
| 331 |
+
"scored_now": 3000,
|
| 332 |
+
"cached_completed": 0,
|
| 333 |
+
"stage": "all",
|
| 334 |
+
"worker_count": 1,
|
| 335 |
+
"worker_index": 0,
|
| 336 |
+
"worker_owned_predictions": 3000,
|
| 337 |
+
"generation_concurrency": 1024,
|
| 338 |
+
"judge_concurrency": 256,
|
| 339 |
+
"updated_at": "2026-10-30T07:12:45.955152+00:00"
|
| 340 |
+
},
|
| 341 |
+
"frontierscience-research": {
|
| 342 |
+
"total_records": 1800,
|
| 343 |
+
"completed": 1799,
|
| 344 |
+
"query_errors": 0,
|
| 345 |
+
"score_errors": 1,
|
| 346 |
+
"generated_unscored": 0,
|
| 347 |
+
"mean_score": 0.32770525291828795,
|
| 348 |
+
"accuracy": 0.07226236798221233,
|
| 349 |
+
"mean_latency_seconds": 569.9840808942698,
|
| 350 |
+
"finish_reason_counts": {
|
| 351 |
+
"length": 18,
|
| 352 |
+
"stop": 1781
|
| 353 |
+
},
|
| 354 |
+
"length_limited_response_count": 18,
|
| 355 |
+
"observed_max_completion_tokens": 98304,
|
| 356 |
+
"judge_usage": {
|
| 357 |
+
"judged_record_count": 1799,
|
| 358 |
+
"request_count": 1821,
|
| 359 |
+
"retry_request_count": 22,
|
| 360 |
+
"prompt_tokens": 9816612,
|
| 361 |
+
"cached_prompt_tokens": 120976,
|
| 362 |
+
"cache_write_prompt_tokens": 9676591,
|
| 363 |
+
"completion_tokens": 3183991,
|
| 364 |
+
"reasoning_tokens": 1921665,
|
| 365 |
+
"total_tokens": 13000603
|
| 366 |
+
},
|
| 367 |
+
"judge_api_usage": {
|
| 368 |
+
"request_count": 1822,
|
| 369 |
+
"prompt_tokens": 9834515,
|
| 370 |
+
"cached_prompt_tokens": 120976,
|
| 371 |
+
"cache_write_prompt_tokens": 9694491,
|
| 372 |
+
"completion_tokens": 3192183,
|
| 373 |
+
"reasoning_tokens": 1929857,
|
| 374 |
+
"total_tokens": 13026698
|
| 375 |
+
},
|
| 376 |
+
"benchmark_metrics": {
|
| 377 |
+
"track": "research",
|
| 378 |
+
"n_problems_scored": 60,
|
| 379 |
+
"n_trajectories_scored": 1799,
|
| 380 |
+
"aggregation": "average_trials",
|
| 381 |
+
"reported_accuracy": 0.07222222222222222,
|
| 382 |
+
"mean_rubric_score": 0.3275955603448276,
|
| 383 |
+
"success_threshold": 0.7,
|
| 384 |
+
"community_profile": null
|
| 385 |
+
},
|
| 386 |
+
"benchmark": "frontierscience-research",
|
| 387 |
+
"model": "openai/gpt-oss-20b@6cee5e81ee83917806bbde320786a8fb61efebee",
|
| 388 |
+
"execution_mode": "non_agent",
|
| 389 |
+
"generation_hash": "b86f7a8a1172f905444f6d2df6a245fbc642f440479171932b5e425e54e1353e",
|
| 390 |
+
"scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750",
|
| 391 |
+
"selected": 1800,
|
| 392 |
+
"incomplete": 1,
|
| 393 |
+
"generated_now": 1800,
|
| 394 |
+
"scored_now": 1799,
|
| 395 |
+
"cached_completed": 0,
|
| 396 |
+
"stage": "all",
|
| 397 |
+
"worker_count": 1,
|
| 398 |
+
"worker_index": 0,
|
| 399 |
+
"worker_owned_predictions": 1800,
|
| 400 |
+
"generation_concurrency": 1024,
|
| 401 |
+
"judge_concurrency": 256,
|
| 402 |
+
"updated_at": "2026-10-30T06:46:49.734658+00:00"
|
| 403 |
+
},
|
| 404 |
+
"scicode-verified": {
|
| 405 |
+
"updated_at": 1793344949.0352376,
|
| 406 |
+
"expected_problems": 64,
|
| 407 |
+
"expected_steps": 287,
|
| 408 |
+
"generated_steps": 287,
|
| 409 |
+
"scored_steps": 287,
|
| 410 |
+
"passed_steps": 197,
|
| 411 |
+
"completed_problems": 64,
|
| 412 |
+
"passed_problems": 31,
|
| 413 |
+
"subproblem_accuracy": 0.686411149825784,
|
| 414 |
+
"main_problem_accuracy": 0.484375,
|
| 415 |
+
"complete": true,
|
| 416 |
+
"cap_hits": 1,
|
| 417 |
+
"mean_output_tokens": 8765.648083623693,
|
| 418 |
+
"grading_infrastructure_errors": 0,
|
| 419 |
+
"errors": []
|
| 420 |
+
},
|
| 421 |
+
"scicoding-agent": {
|
| 422 |
+
"tasks": 50,
|
| 423 |
+
"attempts": 100,
|
| 424 |
+
"attempts_per_task": 2,
|
| 425 |
+
"metric": "avg@2",
|
| 426 |
+
"passed": 22,
|
| 427 |
+
"avg_at_2": 0.22,
|
| 428 |
+
"unscored": 0,
|
| 429 |
+
"agent_timeouts": 0,
|
| 430 |
+
"reward_positive_with_exception": 0,
|
| 431 |
+
"scoring": "native_verifier",
|
| 432 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1"
|
| 433 |
+
},
|
| 434 |
+
"terminal-bench-2.1": {
|
| 435 |
+
"tasks": 89,
|
| 436 |
+
"attempts": 178,
|
| 437 |
+
"attempts_per_task": 2,
|
| 438 |
+
"metric": "avg@2",
|
| 439 |
+
"passed": 17,
|
| 440 |
+
"avg_at_2": 0.09550561797752809,
|
| 441 |
+
"unscored": 2,
|
| 442 |
+
"agent_timeouts": 12,
|
| 443 |
+
"reward_positive_with_exception": 1,
|
| 444 |
+
"scoring": "native_verifier",
|
| 445 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1"
|
| 446 |
+
},
|
| 447 |
+
"hle-verified": {
|
| 448 |
+
"gold": {
|
| 449 |
+
"correct": 162,
|
| 450 |
+
"total": 575,
|
| 451 |
+
"accuracy": 0.2817391304347826,
|
| 452 |
+
"cap_hits": 72
|
| 453 |
+
},
|
| 454 |
+
"revision": {
|
| 455 |
+
"correct": 332,
|
| 456 |
+
"total": 1025,
|
| 457 |
+
"accuracy": 0.32390243902439025,
|
| 458 |
+
"cap_hits": 150
|
| 459 |
+
},
|
| 460 |
+
"correct": 494,
|
| 461 |
+
"total": 1600,
|
| 462 |
+
"overall_accuracy": 0.30875,
|
| 463 |
+
"all_complete": true,
|
| 464 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 465 |
+
"provenance": "hle-verified-provenance.json"
|
| 466 |
+
}
|
| 467 |
+
}
|
hle-verified-20261030/README.md
DELETED
|
@@ -1,29 +0,0 @@
|
|
| 1 |
-
# HLE-Verified: seven-model text-only comparison — 2026-10-30
|
| 2 |
-
|
| 3 |
-
All seven models completed 575 Gold + 1,025 Revision questions: **11,200 scored answers**, no remaining generation/judge errors or unscored answers. This is a custom HLE-Verified text-only evaluation, not the original full HLE leaderboard protocol. Gold and Revision are also reported separately; overall is question-weighted, not the unweighted mean of the two subset accuracies.
|
| 4 |
-
|
| 5 |
-
## Results
|
| 6 |
-
|
| 7 |
-
| Model | Gold (575) | Revision (1,025) | Correct / 1,600 | Overall |
|
| 8 |
-
|---|---:|---:|---:|---:|
|
| 9 |
-
| Qwen3.5-35B-A3B | 39.65% | 35.61% | 593/1,600 | 37.06% |
|
| 10 |
-
| sft-agentic-stage2 | 28.70% | 32.68% | 500/1,600 | 31.25% |
|
| 11 |
-
| gpt-oss-20b | 28.17% | 32.39% | 494/1,600 | 30.88% |
|
| 12 |
-
| Gemma-4-12B-it | 27.65% | 29.95% | 466/1,600 | 29.12% |
|
| 13 |
-
| Qwen3.5-9B | 29.57% | 28.20% | 459/1,600 | 28.69% |
|
| 14 |
-
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 | 23.30% | 28.49% | 426/1,600 | 26.62% |
|
| 15 |
-
| Olmo-3-32B-Think | 12.00% | 21.46% | 289/1,600 | 18.06% |
|
| 16 |
-
|
| 17 |
-
## Protocol
|
| 18 |
-
|
| 19 |
-
Dataset: `skylenage-ai/HLE-Verified`, revision `0bc83643672d4f68a5f89998617a639d85e7318b`. Gold excludes 93 image questions; Revision excludes 118. Uncertain is excluded. One response per question, temperature 1, top-p 1, seed 373363, model-specific reasoning settings preserved in configs. Luna judges final answers using the HLE binary comparison template, with an 8,192-token judge budget.
|
| 20 |
-
|
| 21 |
-
Candidate output limit is 98,304 tokens, except OLMo-3-32B-Think: up to 61,440 within its native 65,536-token total context, dynamically reduced for long prompts (no prompt truncation). This budget difference must be considered when comparing models.
|
| 22 |
-
|
| 23 |
-
Our model is v3 agentic SFT step 1144 from `JLiangHe/sci-ckpt-final/sft-agentic/iter_0001144`, revision `06129b3983eab7e8fe01e5c30577e76250605c66`. It is not the v2 RL checkpoint.
|
| 24 |
-
|
| 25 |
-
## Artifact layout
|
| 26 |
-
|
| 27 |
-
Each existing model folder now has `hle-verified-gold/` and `hle-verified-revision/`, containing `summary.json`, `trajectories.jsonl`, `judge_trajectories.jsonl`, `events.jsonl.gz`, and `configs.json`. Trajectories retain saved reasoning and final answer content; events retain the complete saved append-only history. Judge files preserve native Luna score details and raw judge responses. Credentials are redacted by the maintained exporter. Publication performs no generation or regrading.
|
| 28 |
-
|
| 29 |
-
This comparison folder contains machine-readable results, frozen source/config manifests and recovery provenance. Existing results for other benchmarks are untouched.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
hle-verified-20261030/manifest.json
DELETED
|
@@ -1 +0,0 @@
|
|
| 1 |
-
{"Gemma-4-12B-it/hle-verified-gold/configs.json": {"size": 95876, "sha256": "25a009e6c6226da7d37495e9a923c6519228bd06c0abef92f8185fffd54d97c8"}, "Gemma-4-12B-it/hle-verified-gold/events.jsonl.gz": {"size": 12394882, "sha256": "372211276b94cd99f2d455bda1aee8ca8f96f9461143b850d3f582bf05c47e74"}, "Gemma-4-12B-it/hle-verified-gold/judge_trajectories.jsonl": {"size": 808737, "sha256": "8deb1da2f9285ac2e39b151e074785844f16f1da4bc64788032ad9e932fda6d7"}, "Gemma-4-12B-it/hle-verified-gold/summary.json": {"size": 1495, "sha256": "6e5cffccacef2f8eb03cbcf45da90f97468cefb6153e8fdede1d217a32341e0e"}, "Gemma-4-12B-it/hle-verified-gold/trajectories.jsonl": {"size": 30459259, "sha256": "3b943c4da59d980a98c8e027c7b4a5c46893825ba6e1c6facfb0f0dcb5dfc2aa"}, "Gemma-4-12B-it/hle-verified-revision/configs.json": {"size": 164308, "sha256": "35c646cf84d89f08772cd52ea897a63040166e7e8355965f007f3af05a5bc483"}, "Gemma-4-12B-it/hle-verified-revision/events.jsonl.gz": {"size": 24343432, "sha256": "f25a4183fe986f31821b984244964c866449fb5280d91ffb9d0673684a6eaab0"}, "Gemma-4-12B-it/hle-verified-revision/judge_trajectories.jsonl": {"size": 1461032, "sha256": "2bd13649258614bf6b9ae852dedec6de9affb24a70d68b0f748b1fb5003e5844"}, "Gemma-4-12B-it/hle-verified-revision/summary.json": {"size": 1516, "sha256": "77a100d4185ae9069742c37fd316297355723476fd44ced24b535dd6ad8f1960"}, "Gemma-4-12B-it/hle-verified-revision/trajectories.jsonl": {"size": 57776284, "sha256": "943f338e784a0e4677818a0148748b65e64c059cf7ca35b6730abd1171f7c840"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/configs.json": {"size": 96540, "sha256": "63416cb9660523e05f57d41dba3d99dac1a41bdfcb901896daf3074a3ba6af9c"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/events.jsonl.gz": {"size": 43697680, "sha256": "8b4f3ddd8cdd092dd94e45ecbf2bd058dfc778025dfacc5564174d3f22bf3d6d"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/judge_trajectories.jsonl": {"size": 807987, "sha256": "194d663838cd73be2ca90824ddbf75386937fa67b0b98bc07b399f54a3f6560d"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/summary.json": {"size": 1504, "sha256": "69aa51f1a177ae8489ec03df54b4b82b0940cb4927c2217aaad75fd1fe4ff3a9"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-gold/trajectories.jsonl": {"size": 86865668, "sha256": "a03274a1b9b7c94f4250161e93343c41747022bdd35c9e14bd660d0574b273fd"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/configs.json": {"size": 164972, "sha256": "9110a79193b19841f399f51c14b824c2e4c15a2dfc745a8ca08805b4ce69d037"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/events.jsonl.gz": {"size": 81111047, "sha256": "83707e1fd3752c6828b5f53d57761141bc6c05f942a577bf3cbe20a18bcaf26c"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/judge_trajectories.jsonl": {"size": 1460099, "sha256": "d3bc211cc10b3443ed1baafbb45336c8bb961e176b31c31998864240ad47f6f8"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/summary.json": {"size": 1519, "sha256": "737e73ccadc0e159ef4caae8a78565cf668f74ff41889f30c10719014acab918"}, "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/hle-verified-revision/trajectories.jsonl": {"size": 158052659, "sha256": "a1a76fdad16f0182616ca18263ddb1d4e6c4f3d47b5466ffeca42e87749932ba"}, "Olmo-3-32B-Think/hle-verified-gold/configs.json": {"size": 96624, "sha256": "83cf37950d455c9c52a1a69e4f0e772602083e0257d6f04b7d6f456698391bd0"}, "Olmo-3-32B-Think/hle-verified-gold/events.jsonl.gz": {"size": 20232761, "sha256": "7e32e5885184e09eab0c7f96a15dc3fa3ff5ea9a0da174ed279f7cd0f3a1dee6"}, "Olmo-3-32B-Think/hle-verified-gold/judge_trajectories.jsonl": {"size": 802265, "sha256": "b04d68535d47e5c4746fb2320355f3daeb86c5eb078063bfb1334ace5b2542e9"}, "Olmo-3-32B-Think/hle-verified-gold/summary.json": {"size": 1440, "sha256": "c4d000570412503eaa8c8e3e7cca4e285209e73a1a3916d03768c917f3a25dd6"}, "Olmo-3-32B-Think/hle-verified-gold/trajectories.jsonl": {"size": 37719102, "sha256": "21e99af8b23fc0814a7b79aacb5ab4c8ff6b19f051e28ae09ca1f6846a6a4937"}, "Olmo-3-32B-Think/hle-verified-revision/configs.json": {"size": 165056, "sha256": "db5e7b7d094dbffde70da87a3bec9e21dfe83a2791898de9bb6b11723ed24e9e"}, "Olmo-3-32B-Think/hle-verified-revision/events.jsonl.gz": {"size": 37941813, "sha256": "b7b3aaca63e47ad81212f580ebcdd253d26c022fff10fdb0ef6dd2c0ffc523c8"}, "Olmo-3-32B-Think/hle-verified-revision/judge_trajectories.jsonl": {"size": 1446837, "sha256": "1fe962a321092f2a6a162be42ac90905b68f2dccea08f3a0e53ddab220d193ee"}, "Olmo-3-32B-Think/hle-verified-revision/summary.json": {"size": 1497, "sha256": "46fb3a59e8f35fb903dfb0ee2d572838fdea1f7f08692f0f69f08440dfa234e7"}, "Olmo-3-32B-Think/hle-verified-revision/trajectories.jsonl": {"size": 68910479, "sha256": "57b1f5dbdfd8e8c35db47bedcb63d93d36075530d2bcf336cfb9ee203e98416f"}, "Qwen3.5-35B-A3B/hle-verified-gold/configs.json": {"size": 97478, "sha256": "64ed43cc300dbe1e6c6468936aaa7aaabe0b1381b595afea6ba66efa156a13d6"}, "Qwen3.5-35B-A3B/hle-verified-gold/events.jsonl.gz": {"size": 38963380, "sha256": "06fb9b5c90ec966dffa5344062a15948bc212e97be536874a5976b71df21aaf0"}, "Qwen3.5-35B-A3B/hle-verified-gold/judge_trajectories.jsonl": {"size": 811549, "sha256": "32a6043c2f63917e80e52af5e231545b3aed997bf342314730df3207f4f84846"}, "Qwen3.5-35B-A3B/hle-verified-gold/summary.json": {"size": 1506, "sha256": "b1d67b09d49e4c48dcd194043f9e84e1357b51d13371475723f705214974b223"}, "Qwen3.5-35B-A3B/hle-verified-gold/trajectories.jsonl": {"size": 89998977, "sha256": "c13192f1aba209208d7d014dbc9722a56f5cfefd369f5cdb78cd53d8469d8340"}, "Qwen3.5-35B-A3B/hle-verified-revision/configs.json": {"size": 165910, "sha256": "2cf46e0583218f82eabc3556d339e6f037a78089ff93aedbe2525021b646642c"}, "Qwen3.5-35B-A3B/hle-verified-revision/events.jsonl.gz": {"size": 73200323, "sha256": "0c22b10aa45d9f8318db568f60b10885e0de25597c86d11b493e1a81bb7fd898"}, "Qwen3.5-35B-A3B/hle-verified-revision/judge_trajectories.jsonl": {"size": 1463315, "sha256": "64c4e07c7f748f38df05005367e8625bd6e7226b235767e8990a33d4f0493f3c"}, "Qwen3.5-35B-A3B/hle-verified-revision/summary.json": {"size": 1519, "sha256": "97013cb3dbf05820ffc3e035a13c6a994d7f70e4da8fe2e9ed272761c8248748"}, "Qwen3.5-35B-A3B/hle-verified-revision/trajectories.jsonl": {"size": 160439763, "sha256": "0c77c80238a38e24b1e1edd6df57f58e64bfc1f415cefd08680e8e79470cec4c"}, "Qwen3.5-9B/hle-verified-gold/configs.json": {"size": 96465, "sha256": "7bb994f1b342c98d66fcd1b1406ce3b2de3084d12b1a0e772cb9a7c3f7ad7b6e"}, "Qwen3.5-9B/hle-verified-gold/events.jsonl.gz": {"size": 43633633, "sha256": "159496911a57230265fee4f420cc369011efbd4941ece2191c712d32329a5ad3"}, "Qwen3.5-9B/hle-verified-gold/judge_trajectories.jsonl": {"size": 806504, "sha256": "842b1fec46cefcdf4d0230f2dd7a89b40bd31e45d248b15f2e2bbd2338b795a4"}, "Qwen3.5-9B/hle-verified-gold/summary.json": {"size": 1499, "sha256": "8db7f858a18f933acb3c3462ff73f75b8304f55d59e010c85cd1b35e2ef07557"}, "Qwen3.5-9B/hle-verified-gold/trajectories.jsonl": {"size": 102604390, "sha256": "9b5029d1a019217548daa2bcd550007616f1a6e585e2bd08c613ab045fd72ca3"}, "Qwen3.5-9B/hle-verified-revision/configs.json": {"size": 164897, "sha256": "a6ceca950617a9cc8ca04eee77205df67d4fbfdcbf655df3da5002e4a0efe8ff"}, "Qwen3.5-9B/hle-verified-revision/events.jsonl.gz": {"size": 81502302, "sha256": "cd44292482845ae3db5a22151cfb17ff800e2b68a456bec5eac8f9e4986069b5"}, "Qwen3.5-9B/hle-verified-revision/judge_trajectories.jsonl": {"size": 1458844, "sha256": "26eb3d8ecaf7a961bb56442bb2cb44dff5f41649c5f34bbac284bf01d65b59c8"}, "Qwen3.5-9B/hle-verified-revision/summary.json": {"size": 1522, "sha256": "60d1ed8703ec79e5ae80e64c6373da5e53ddacc48dec0a0decbc78e99cd51479"}, "Qwen3.5-9B/hle-verified-revision/trajectories.jsonl": {"size": 184438526, "sha256": "394aaab563418001fbcc0c50b1c41a1ee0eabc5f72c34da393847a104699dffe"}, "gpt-oss-20b/hle-verified-gold/configs.json": {"size": 96732, "sha256": "aba3464b730503bbf359c8e7ccb02197c6a3c2acc74ad1080d8de16588d95970"}, "gpt-oss-20b/hle-verified-gold/events.jsonl.gz": {"size": 49832341, "sha256": "187167c4c0cf1fdb45836d03b1f74ddfdc2442cfc7f275fe610c0a2e6c552ebb"}, "gpt-oss-20b/hle-verified-gold/judge_trajectories.jsonl": {"size": 807469, "sha256": "70ed778048e88437e3a6a99261df5320349f4ce0ae68c2ee86a2cfbf1b5cd28f"}, "gpt-oss-20b/hle-verified-gold/summary.json": {"size": 1492, "sha256": "1328061cbb354b82e82b47c95fd8a30564fbb65835ced09e9861dcdb0c067cb9"}, "gpt-oss-20b/hle-verified-gold/trajectories.jsonl": {"size": 89318447, "sha256": "498119901fa4e4c1bbc22231744abe3a20db7211cc9e6f8664a82ce01cbe4320"}, "gpt-oss-20b/hle-verified-revision/configs.json": {"size": 165164, "sha256": "599ef040aed792ba77a02878391c4bc26ba24105defb488ff7d061b9d4443ee7"}, "gpt-oss-20b/hle-verified-revision/events.jsonl.gz": {"size": 90826955, "sha256": "812fb9e7d831676b8e9d1aa0a36166edc97ce7cd4f31e207313dff4ec7deaed6"}, "gpt-oss-20b/hle-verified-revision/judge_trajectories.jsonl": {"size": 1455964, "sha256": "86e0dac921215fcff2913d381d06b903084c8d380070142013e1590a3866d846"}, "gpt-oss-20b/hle-verified-revision/summary.json": {"size": 1516, "sha256": "a94c187e6ae8bdb9e49125767b2d847899d7e97505af4e6d503cd052d02e0ab5"}, "gpt-oss-20b/hle-verified-revision/trajectories.jsonl": {"size": 161160950, "sha256": "e24efccf23a011a6caa879a88650726f8abd29dd2592a12494052493f786cd1e"}, "hle-verified-20261030/README.md": {"size": 2510, "sha256": "4751ed303f9967c679c0e0cb874482c64167c64d707eb30cea21589a63d73544"}, "hle-verified-20261030/provenance.json": {"size": 1030, "sha256": "d1779be6b0ea5874dd37922ad8687c300e852f467dfb0cb4f0f775c6979bb9d2"}, "hle-verified-20261030/source-manifest.json": {"size": 21925, "sha256": "29c276e5e3707eea72193647af96f2aedddbe1d4d9d255ce7bf848077787c43a"}, "hle-verified-20261030/summary.json": {"size": 2328, "sha256": "972222f1fc602f688207fa4832520aa19ea1f44130c725e28f32bbced4dc58b8"}, "hle-verified-20261030/sweep.json": {"size": 21101, "sha256": "1023b5ee5973c1a6c37d27554e8e1c075ce01e12720e88776f31bb37bfe0c32a"}, "hle-verified-20261030/token-preflight.json": {"size": 1461, "sha256": "4be53ba616b8a96ebc0ce96199e9ea16397d32932ce01d9d8f498bb651b6ad01"}, "sft-agentic-stage2/hle-verified-gold/configs.json": {"size": 97117, "sha256": "e47530b3d95ee8e5058347d6a7c496f650a6075c8cbf390894cb4a7b04b56403"}, "sft-agentic-stage2/hle-verified-gold/events.jsonl.gz": {"size": 40849329, "sha256": "6354006b0987e42aa8623cf384aceb56fe4ded2453242a1f7d0df0775c0e7774"}, "sft-agentic-stage2/hle-verified-gold/judge_trajectories.jsonl": {"size": 808092, "sha256": "1966a9542e50ad4611569df699749f49fe6c7ab38348fe9a4878101dcd06935e"}, "sft-agentic-stage2/hle-verified-gold/summary.json": {"size": 1511, "sha256": "12864d1f6cb71017dfd770bedeb26332721afc6f3762e9d5681907506f24af93"}, "sft-agentic-stage2/hle-verified-gold/trajectories.jsonl": {"size": 67049194, "sha256": "f0db38689f2018b6192d85e48a8b669a7ea017e703f0d33f4f2de1b18476fc05"}, "sft-agentic-stage2/hle-verified-revision/configs.json": {"size": 165549, "sha256": "fb902ff8e06c3fc9fbdcaa5d9c3e0fad16fc64a806970868fb29c2bb433ca9ac"}, "sft-agentic-stage2/hle-verified-revision/events.jsonl.gz": {"size": 75369864, "sha256": "d8a9d38adaed854200ec0c7a7cb534740deca65e8db71df68ac151d68f60d622"}, "sft-agentic-stage2/hle-verified-revision/judge_trajectories.jsonl": {"size": 1459606, "sha256": "8277bf3b4053d472660fde2105a30fdcb5e7acabfb7f58385800a42e05dc6dfd"}, "sft-agentic-stage2/hle-verified-revision/summary.json": {"size": 1518, "sha256": "07d966e568f850d54878e50b80158a1d4f26ca4392b23c7fa84de07cb8d1f737"}, "sft-agentic-stage2/hle-verified-revision/trajectories.jsonl": {"size": 122766620, "sha256": "cb3df4e9d3f79da9ad894fe892273d9d66dd1ada40200f7b61946f010d866465"}}
|
|
|
|
|
|
hle-verified-20261030/provenance.json
DELETED
|
@@ -1 +0,0 @@
|
|
| 1 |
-
{"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030", "publication_reruns": false, "all_questions_scored": true, "subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded", "dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b", "qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.", "gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.", "scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.", "judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.", "mlflow": "Explicit user-approved no-MLflow run"}
|
|
|
|
|
|
hle-verified-20261030/source-manifest.json
DELETED
|
@@ -1 +0,0 @@
|
|
| 1 |
-
{"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344", "files": {"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729", "eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f", "eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc", "eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0", "eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce", "eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2", "eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577", "eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f", "eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820", "eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7", "eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca", "eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57", "eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e", "eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692", "eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5", "eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c", "eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452", "eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e", "eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59", "eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582", "eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855", "eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb", "eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f", "eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef", "eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263", "eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0", "eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5", "eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8", "eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436", "eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca", "eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858", "eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309", "eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b", "eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a", "eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e", "eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76", "eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0", "eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", "eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235", "eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3", "eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266", "eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1", "eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617", "eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92", "eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c", "eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965", "eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18", "eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63", "eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a", "eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe", "eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41", "eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b", "eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780", "eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c", "eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75", "eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816", "eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c", "eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d", "eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361", "eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3", "eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7", "eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41", "eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc", "eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff", "eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76", "eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9", "eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f", "eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a", "eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43", "eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e", "eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8", "eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d", "eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9", "eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed", "eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc", "eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b", "eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe", "eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0", "eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2", "eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4", "eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a", "eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a", "eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f", "eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd", "eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5", "eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5", "eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03", "eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356", "eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211", "eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15", "eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66", "eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768", "eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa", "eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f", "eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157", "eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24", "eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9", "eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60", "eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9", "eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba", "eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb", "eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465", "eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276", "eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000", "eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4", "eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc", "eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6", "eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663", "eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362", "eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85", "eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15", "eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf", "eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c", "eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1", "eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5", "eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190", "eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd", "eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3", "eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d", "eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92", "eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf", "eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e", "eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b", "eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45", "eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50", "eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c", "eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906", "eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4", "eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099", "eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb", "eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c", "eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d", "eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510", "eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9", "eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955", "eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd", "eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f", "eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624", "eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5", "eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da", "eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454", "eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059", "eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece", "eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc", "eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368", "eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c", "eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0", "eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f", "eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc", "eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767", "eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a", "eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a", "eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e", "eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1", "eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3", "eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5", "eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709", "eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea", "eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939", "eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917", "eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb", "eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f", "eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c", "eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c", "eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a", "eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf", "eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b", "eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"}, "scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."}
|
|
|
|
|
|
hle-verified-20261030/summary.json
DELETED
|
@@ -1 +0,0 @@
|
|
| 1 |
-
{"models": [{"model": "qwen35-35b-a3b", "folder": "Qwen3.5-35B-A3B", "subsets": {"gold": {"correct": 228, "total": 575, "accuracy": 0.39652173913043476, "cap_hits": 41}, "revision": {"correct": 365, "total": 1025, "accuracy": 0.35609756097560974, "cap_hits": 63}}, "correct": 593, "total": 1600, "overall_accuracy": 0.370625}, {"model": "v3-agentic-sft-1144", "folder": "sft-agentic-stage2", "subsets": {"gold": {"correct": 165, "total": 575, "accuracy": 0.28695652173913044, "cap_hits": 27}, "revision": {"correct": 335, "total": 1025, "accuracy": 0.32682926829268294, "cap_hits": 56}}, "correct": 500, "total": 1600, "overall_accuracy": 0.3125}, {"model": "gpt-oss-20b", "folder": "gpt-oss-20b", "subsets": {"gold": {"correct": 162, "total": 575, "accuracy": 0.2817391304347826, "cap_hits": 72}, "revision": {"correct": 332, "total": 1025, "accuracy": 0.32390243902439025, "cap_hits": 150}}, "correct": 494, "total": 1600, "overall_accuracy": 0.30875}, {"model": "gemma4-12b-it", "folder": "Gemma-4-12B-it", "subsets": {"gold": {"correct": 159, "total": 575, "accuracy": 0.27652173913043476, "cap_hits": 3}, "revision": {"correct": 307, "total": 1025, "accuracy": 0.2995121951219512, "cap_hits": 10}}, "correct": 466, "total": 1600, "overall_accuracy": 0.29125}, {"model": "qwen35-9b-it", "folder": "Qwen3.5-9B", "subsets": {"gold": {"correct": 170, "total": 575, "accuracy": 0.2956521739130435, "cap_hits": 72}, "revision": {"correct": 289, "total": 1025, "accuracy": 0.2819512195121951, "cap_hits": 115}}, "correct": 459, "total": 1600, "overall_accuracy": 0.286875}, {"model": "nemotron35-lightning", "folder": "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16", "subsets": {"gold": {"correct": 134, "total": 575, "accuracy": 0.23304347826086957, "cap_hits": 57}, "revision": {"correct": 292, "total": 1025, "accuracy": 0.2848780487804878, "cap_hits": 95}}, "correct": 426, "total": 1600, "overall_accuracy": 0.26625}, {"model": "olmo3-32b-think", "folder": "Olmo-3-32B-Think", "subsets": {"gold": {"correct": 69, "total": 575, "accuracy": 0.12, "cap_hits": 0}, "revision": {"correct": 220, "total": 1025, "accuracy": 0.2146341463414634, "cap_hits": 0}}, "correct": 289, "total": 1600, "overall_accuracy": 0.180625}], "total_scored": 11200, "overall_definition": "(Gold correct + Revision correct) / 1600", "all_complete": true}
|
|
|
|
|
|
hle-verified-20261030/sweep.json
DELETED
|
@@ -1 +0,0 @@
|
|
| 1 |
-
{"models": [{"name": "gpt-oss-20b", "profile": {"name": "gpt-oss-20b", "gpu": "h100", "model_family": "gpt-oss", "checkpoint": {"model_type": "gpt_oss", "hidden_size": 2880, "num_hidden_layers": 24, "vocab_size": 201088}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 98304, "seed": 373363, "extra": {"top_k": -1, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0, "reasoning_effort": "high"}}, "serving": {"context_length": 131072, "dtype": "bfloat16", "attention_backend": "fa3", "mem_fraction_static": 0.85, "reasoning_parser": "gpt-oss", "extra_args": ["--cuda-graph-max-bs-decode", "64", "--tool-call-parser", "gpt-oss", "--moe-runner-backend", "triton_kernel", "--disable-radix-cache"], "tp": 1, "replicas_per_node": 8, "max_running_requests": 64}, "reasoning": {"requests_per_engine": 64, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 0, "checkpoint": "/workspace/data/raw_model/gpt-oss-20b-6cee5e81ee83", "repo": "openai/gpt-oss-20b", "revision": "6cee5e81ee83917806bbde320786a8fb61efebee", "checkpoint_config_sha256": "3a2a26ded679375b7928ddeca59764df7cea83220c1961035f6d6e232659e9ce", "index_sha256": "0e085b977c4c9942f85938828e8c989ed7d5cdabf852e4da6a67c116cd502cd1", "shard_sizes": {"model-00000-of-00002.safetensors": 4792272488, "model-00001-of-00002.safetensors": 4798702184, "model-00002-of-00002.safetensors": 4170342232}, "endpoints": [{"id": "node-0-replica-0", "api_base": "http://10.2.5.84:8700/v1"}, {"id": "node-0-replica-1", "api_base": "http://10.2.5.84:8701/v1"}, {"id": "node-0-replica-2", "api_base": "http://10.2.5.84:8702/v1"}, {"id": "node-0-replica-3", "api_base": "http://10.2.5.84:8703/v1"}, {"id": "node-0-replica-4", "api_base": "http://10.2.5.84:8704/v1"}, {"id": "node-0-replica-5", "api_base": "http://10.2.5.84:8705/v1"}, {"id": "node-0-replica-6", "api_base": "http://10.2.5.84:8706/v1"}, {"id": "node-0-replica-7", "api_base": "http://10.2.5.84:8707/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gpt-oss-20b", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gpt-oss-20b/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 512, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "8a8bd3d39ee530a401e8004ba6dea0bf05164689556521205b18fe1dffb6b6f8", "configs/hle-verified-gold-revision-96k.yaml": "05098321700593bc74123855f7b4d970f0c9718c72d378827e16b8882de1072f", "configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4", "configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"}}, {"name": "nemotron35-lightning", "profile": {"name": "nemotron35-lightning-30b-a3b", "gpu": "h100", "model_family": "nemotron", "checkpoint": {"model_type": "nemotron_h", "hidden_size": 2688, "num_hidden_layers": 52, "vocab_size": 131072}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 98304, "seed": 373363, "extra": {"chat_template_kwargs": {"enable_thinking": true}}}, "serving": {"context_length": 131072, "dtype": "bfloat16", "attention_backend": "fa3", "mem_fraction_static": 0.85, "reasoning_parser": "nemotron_3", "extra_args": ["--mamba-ssm-dtype", "float16", "--cuda-graph-max-bs-decode", "32", "--tool-call-parser", "qwen3_coder"], "tp": 2, "replicas_per_node": 4, "max_running_requests": 32}, "reasoning": {"requests_per_engine": 32, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 0, "checkpoint": "/workspace/data/raw_model/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16-a9904d24bcc1", "repo": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16", "revision": "a9904d24bcc1d289a1950fa9d2b978c47cf903b9", "checkpoint_config_sha256": "a3827a0f5e311547b40943dc081e3ff2f8a277466e8c1a3df2291e8db8a7617c", "index_sha256": "67f21da80ce245a3e24967f54eef3fa10e67a63eba16ef19a2d0569f06103f50", "shard_sizes": {"model-00001-of-00014.safetensors": 4991210024, "model-00002-of-00014.safetensors": 4992601016, "model-00003-of-00014.safetensors": 4992601368, "model-00004-of-00014.safetensors": 4995692800, "model-00005-of-00014.safetensors": 4980553832, "model-00006-of-00014.safetensors": 4989423536, "model-00007-of-00014.safetensors": 4992601488, "model-00008-of-00014.safetensors": 4992601520, "model-00009-of-00014.safetensors": 4995692800, "model-00010-of-00014.safetensors": 4992601520, "model-00011-of-00014.safetensors": 4995692800, "model-00012-of-00014.safetensors": 4995692816, "model-00013-of-00014.safetensors": 3249723504, "model-00014-of-00014.safetensors": 2670685240}, "endpoints": [{"id": "node-0-replica-0", "api_base": "http://10.2.5.84:8700/v1"}, {"id": "node-0-replica-1", "api_base": "http://10.2.5.84:8701/v1"}, {"id": "node-0-replica-2", "api_base": "http://10.2.5.84:8702/v1"}, {"id": "node-0-replica-3", "api_base": "http://10.2.5.84:8703/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/nemotron35-lightning", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/nemotron35-lightning/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 128, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "2021a2aa4f08953902ccd9e33ff9ab2ad754f301c3e66e4735df32591127b0ed", "configs/hle-verified-gold-revision-96k.yaml": "eaa36f2db209bce2f9de06724792d4d03f0ad0139c88b50ccd0e9a780126117c", "configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4", "configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"}}, {"name": "qwen35-9b-it", "profile": {"name": "qwen35-9b", "gpu": "h100", "model_family": "qwen", "checkpoint": {"model_type": "qwen3_5_text", "hidden_size": 4096, "num_hidden_layers": 32, "vocab_size": 248320}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 98304, "seed": 373363, "extra": {"top_k": -1, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0, "chat_template_kwargs": {"enable_thinking": true}}}, "serving": {"context_length": 131072, "dtype": "bfloat16", "attention_backend": "triton", "mem_fraction_static": 0.8, "reasoning_parser": "qwen3", "model_config_parser": "sci_qwen35_text", "external_model_package": "sci_qwen35_text_model", "extra_args": ["--sampling-backend", "flashinfer", "--cuda-graph-backend-decode", "full", "--cuda-graph-backend-prefill", "disabled"], "tp": 2, "replicas_per_node": 4, "max_running_requests": 32}, "reasoning": {"requests_per_engine": 32, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 1, "checkpoint": "/workspace/data/raw_model/Qwen3.5-9B-text", "repo": "Qwen/Qwen3.5-9B", "revision": "c202236235762e1c871ad0ccb60c8ee5ba337b9a", "checkpoint_config_sha256": "7ed31d50f8c0e8dc8219aa686a97c08665b8d9c85321e4c2fb766139fb6d01af", "index_sha256": "6ef4a22a4e97a96f7d4d55e97d2eeb4bc589720cfefb349583111f21ee78e376", "shard_sizes": {"model-00001-of-00004.safetensors": 4973901592, "model-00002-of-00004.safetensors": 4988362880, "model-00003-of-00004.safetensors": 4954200776, "model-00004-of-00004.safetensors": 3477773688}, "endpoints": [{"id": "node-1-replica-0", "api_base": "http://10.1.136.8:8700/v1"}, {"id": "node-1-replica-1", "api_base": "http://10.1.136.8:8701/v1"}, {"id": "node-1-replica-2", "api_base": "http://10.1.136.8:8702/v1"}, {"id": "node-1-replica-3", "api_base": "http://10.1.136.8:8703/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-9b-it", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-9b-it/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 128, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "31b150f4d5ef7e169dc07acda909f93270ebbe98c8ba80516e48623ba409d085", "configs/hle-verified-gold-revision-96k.yaml": "a417e37e8a3c4a4b3f8a09eedb757b5df167c16ec9a8a9337396900925fc6d1d", "configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4", "configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"}, "serving_export": "Existing text-only export used in previous IT evaluations; no weight modifications in this sweep"}, {"name": "qwen35-35b-a3b", "profile": {"name": "qwen35-35b-a3b", "gpu": "h100", "model_family": "qwen", "checkpoint": {"model_type": "qwen3_5_moe_text", "hidden_size": 2048, "num_hidden_layers": 40, "vocab_size": 248320}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 98304, "seed": 373363, "extra": {"top_k": -1, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0, "chat_template_kwargs": {"enable_thinking": true}}}, "serving": {"context_length": 131072, "dtype": "bfloat16", "attention_backend": "fa3", "mem_fraction_static": 0.85, "reasoning_parser": "qwen3", "extra_args": ["--cuda-graph-max-bs-decode", "32", "--tool-call-parser", "qwen3_coder"], "tp": 2, "replicas_per_node": 4, "max_running_requests": 32}, "reasoning": {"requests_per_engine": 32, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 1, "checkpoint": "/workspace/data/raw_model/Qwen3.5-35B-A3B-59d61f3ce65a", "repo": "Qwen/Qwen3.5-35B-A3B", "revision": "59d61f3ce65a6d9863b86d2e96597125219dc754", "checkpoint_config_sha256": "5e4d7f74fec2f360eb9cfbfcd6ec0c4c76e684d3a11caaed259d9fd9bfbc7944", "index_sha256": "d8d0b7ca4e61ae107e3e87a3ff21136b3ac7c789e64bb24267227ca804e04205", "shard_sizes": {"model.safetensors-00001-of-00014.safetensors": 5368709816, "model.safetensors-00002-of-00014.safetensors": 5368709808, "model.safetensors-00003-of-00014.safetensors": 5368709808, "model.safetensors-00004-of-00014.safetensors": 5368709816, "model.safetensors-00005-of-00014.safetensors": 5368709816, "model.safetensors-00006-of-00014.safetensors": 5368709816, "model.safetensors-00007-of-00014.safetensors": 5368709808, "model.safetensors-00008-of-00014.safetensors": 5368709808, "model.safetensors-00009-of-00014.safetensors": 5255463936, "model.safetensors-00010-of-00014.safetensors": 5368710456, "model.safetensors-00011-of-00014.safetensors": 5368710464, "model.safetensors-00012-of-00014.safetensors": 5368710456, "model.safetensors-00013-of-00014.safetensors": 5367839544, "model.safetensors-00014-of-00014.safetensors": 2224764664}, "endpoints": [{"id": "node-1-replica-0", "api_base": "http://10.1.136.8:8700/v1"}, {"id": "node-1-replica-1", "api_base": "http://10.1.136.8:8701/v1"}, {"id": "node-1-replica-2", "api_base": "http://10.1.136.8:8702/v1"}, {"id": "node-1-replica-3", "api_base": "http://10.1.136.8:8703/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-35b-a3b", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/qwen35-35b-a3b/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 128, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "a073001a8ff3ed003b525f772dd452e44788a2c093caa48e0552a6ee3d6cba56", "configs/hle-verified-gold-revision-96k.yaml": "8f3608cef1c230c7ccff9a29a1a0bef91d185d9dcc1b9771b114566e1f3efb56", "configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4", "configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"}}, {"name": "gemma4-12b-it", "profile": {"name": "gemma4-12b", "gpu": "h100", "model_family": "gemma", "checkpoint": {"model_type": "gemma4_unified_text", "hidden_size": 3840, "num_hidden_layers": 48, "vocab_size": 262144}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 98304, "seed": 373363, "extra": {"top_k": -1, "presence_penalty": 0.0, "chat_template_kwargs": {"enable_thinking": true}}}, "serving": {"context_length": 131072, "dtype": "bfloat16", "attention_backend": "triton", "mem_fraction_static": 0.85, "reasoning_parser": "gemma4", "external_model_package": "gemma4_compat", "extra_args": ["--chunked-prefill-size", "4096"], "tp": 2, "replicas_per_node": 4, "max_running_requests": 32}, "reasoning": {"requests_per_engine": 32, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 2, "checkpoint": "/workspace/data/raw_model/gemma-4-12B-it-707f0a3b8a3c", "repo": "google/gemma-4-12B-it", "revision": "707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7", "checkpoint_config_sha256": "478c46e8d2c52d5c2d85bf67e3b3e8c90e7c9d91086cee27e3c267907e936bd9", "index_sha256": null, "shard_sizes": {"model.safetensors": 23919549408}, "endpoints": [{"id": "node-2-replica-0", "api_base": "http://10.4.70.214:8700/v1"}, {"id": "node-2-replica-1", "api_base": "http://10.4.70.214:8701/v1"}, {"id": "node-2-replica-2", "api_base": "http://10.4.70.214:8702/v1"}, {"id": "node-2-replica-3", "api_base": "http://10.4.70.214:8703/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gemma4-12b-it", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/gemma4-12b-it/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 128, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "ef319aecd370202f03b6bf612bfb2507a92d17be361615b0abe65b7ba9222beb", "configs/hle-verified-gold-revision-96k.yaml": "42ffa0f22d18b4e04a4321d5904208c87583ffccca5de393b690ccc29893dab3", "configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4", "configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"}}, {"name": "v3-agentic-sft-1144", "profile": {"name": "qwen35-9b", "gpu": "h100", "model_family": "qwen", "checkpoint": {"model_type": "qwen3_5_text", "hidden_size": 4096, "num_hidden_layers": 32, "vocab_size": 248320}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 98304, "seed": 373363, "extra": {"top_k": -1, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0, "chat_template_kwargs": {"enable_thinking": true}}}, "serving": {"context_length": 131072, "dtype": "bfloat16", "attention_backend": "triton", "mem_fraction_static": 0.8, "reasoning_parser": "qwen3", "model_config_parser": "sci_qwen35_text", "external_model_package": "sci_qwen35_text_model", "extra_args": ["--sampling-backend", "flashinfer", "--cuda-graph-backend-decode", "full", "--cuda-graph-backend-prefill", "disabled"], "tp": 2, "replicas_per_node": 4, "max_running_requests": 32}, "reasoning": {"requests_per_engine": 32, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 2, "checkpoint": "/workspace/data/raw_model/sci-ckpt-final/06129b3983eab7e8fe01e5c30577e76250605c66/sft-agentic/iter_0001144", "repo": "JLiangHe/sci-ckpt-final/sft-agentic/iter_0001144", "revision": "06129b3983eab7e8fe01e5c30577e76250605c66", "checkpoint_config_sha256": "7ed31d50f8c0e8dc8219aa686a97c08665b8d9c85321e4c2fb766139fb6d01af", "index_sha256": "a2507a6b56459640fc02cefe71a21ce253574d5aab3f023f6336ce980f5c955d", "shard_sizes": {"model-00001-of-00004.safetensors": 5276436016, "model-00002-of-00004.safetensors": 5335160776, "model-00003-of-00004.safetensors": 4932508344, "model-00004-of-00004.safetensors": 2363550592}, "endpoints": [{"id": "node-2-replica-0", "api_base": "http://10.4.70.214:8700/v1"}, {"id": "node-2-replica-1", "api_base": "http://10.4.70.214:8701/v1"}, {"id": "node-2-replica-2", "api_base": "http://10.4.70.214:8702/v1"}, {"id": "node-2-replica-3", "api_base": "http://10.4.70.214:8703/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/v3-agentic-sft-1144", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/v3-agentic-sft-1144/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 128, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "722bc05f950a385bb19f1bb2585b696020c74d3f1330554382405232b555bf5b", "configs/hle-verified-gold-revision-96k.yaml": "0f71a6185cd62ac1537c2623305b92cce4c5ebb1facce8b83fbe548bd8f53a72", "configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4", "configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"}}, {"name": "olmo3-32b-think", "profile": {"name": "olmo3-32b-think", "gpu": "h100", "model_family": "olmo", "checkpoint": {"model_type": "olmo3", "hidden_size": 5120, "num_hidden_layers": 64, "vocab_size": 100278}, "generation": {"temperature": 1.0, "top_p": 1.0, "max_tokens": 61440, "seed": 373363, "extra": {"top_k": -1, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0, "_max_total_tokens": 65536}}, "serving": {"context_length": 65536, "dtype": "bfloat16", "attention_backend": "fa3", "mem_fraction_static": 0.85, "reasoning_parser": "deepseek-r1", "extra_args": ["--cuda-graph-max-bs-decode", "32"], "tp": 2, "replicas_per_node": 4, "max_running_requests": 32}, "reasoning": {"requests_per_engine": 32, "benchmarks": {}}, "agentic_tasks_per_engine": 1}, "node": 3, "checkpoint": "/workspace/data/raw_model/Olmo-3-32B-Think-f2edda15216e", "repo": "allenai/Olmo-3-32B-Think", "revision": "f2edda15216e738ef2bb73771e11890e152b2112", "checkpoint_config_sha256": "d47a120cb2669cd46482d5c2da9488903b67759ca167cd85c1127b2e05755f95", "index_sha256": "3f639e0b434e75af1bc14942b8ad0e4ab100923f2ca071c6985208d50293a269", "shard_sizes": {"model-00001-of-00014.safetensors": 4990600552, "model-00002-of-00014.safetensors": 4938975944, "model-00003-of-00014.safetensors": 4876048688, "model-00004-of-00014.safetensors": 4876048704, "model-00005-of-00014.safetensors": 4876048704, "model-00006-of-00014.safetensors": 4876048704, "model-00007-of-00014.safetensors": 4876048704, "model-00008-of-00014.safetensors": 4876048704, "model-00009-of-00014.safetensors": 4876048704, "model-00010-of-00014.safetensors": 4876048704, "model-00011-of-00014.safetensors": 4876048704, "model-00012-of-00014.safetensors": 4876048704, "model-00013-of-00014.safetensors": 4750216928, "model-00014-of-00014.safetensors": 1026846848}, "endpoints": [{"id": "node-3-replica-0", "api_base": "http://10.0.154.99:8700/v1"}, {"id": "node-3-replica-1", "api_base": "http://10.0.154.99:8701/v1"}, {"id": "node-3-replica-2", "api_base": "http://10.0.154.99:8702/v1"}, {"id": "node-3-replica-3", "api_base": "http://10.0.154.99:8703/v1"}], "root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/olmo3-32b-think", "suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/olmo3-32b-think/configs/hle-verified-gold-revision-96k.yaml", "concurrency": 128, "judge_concurrency": 64, "tracking": {"enabled": false, "reason": "Explicit user approval for this HLE sweep"}, "config_sha256": {"model.yaml": "8ce9f0fbf7af64b5c1367f60a31ecceb4573cffaaaabadb341fac66e4f65ef7b", "configs/hle-verified-gold-revision-96k.yaml": "1cb6a3f9c82920a2b19d4b3eb8da385fb30a494981bbbe357d0d1417b34c3275", "configs/hle-verified-gold.yaml": "d9d0b773707980d1259e5b4500a454e3da651a47132e1bbb52c28661f804221d", "configs/hle-verified-revision.yaml": "b3f43ef4b948a0a719e1914755df24ad4164c26e312c86603b7efaab1bac36da"}}], "data": {"dataset": "skylenage-ai/HLE-Verified", "revision": "0bc83643672d4f68a5f89998617a639d85e7318b", "counts": {"Gold": {"total": 668, "text": 575, "image_excluded": 93}, "Revision": {"total": 1143, "text": 1025, "image_excluded": 118}}, "sha256": {"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b", "data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f", "data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa", "data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737", "data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2", "data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485", "data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70", "data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8", "data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8", "data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d", "gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d", "revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"}}, "excluded": ["v2 RL step120", "Luna", "DS Flash", "GPT-OSS120B"], "resources": "four existing eight-H100 pods; CPU coordinator 0"}
|
|
|
|
|
|
hle-verified-20261030/token-preflight.json
DELETED
|
@@ -1 +0,0 @@
|
|
| 1 |
-
[{"model": "gpt-oss-20b", "questions": 1600, "max_prompt_tokens": 12308, "output_cap": 98304, "context": 131072, "dynamic_output_budget": false, "affected_questions": 0, "min_effective_output_cap": 98304}, {"model": "nemotron35-lightning", "questions": 1600, "max_prompt_tokens": 13776, "output_cap": 98304, "context": 131072, "dynamic_output_budget": false, "affected_questions": 0, "min_effective_output_cap": 98304}, {"model": "qwen35-9b-it", "questions": 1600, "max_prompt_tokens": 13586, "output_cap": 98304, "context": 131072, "dynamic_output_budget": false, "affected_questions": 0, "min_effective_output_cap": 98304}, {"model": "qwen35-35b-a3b", "questions": 1600, "max_prompt_tokens": 13586, "output_cap": 98304, "context": 131072, "dynamic_output_budget": false, "affected_questions": 0, "min_effective_output_cap": 98304}, {"model": "gemma4-12b-it", "questions": 1600, "max_prompt_tokens": 10151, "output_cap": 98304, "context": 131072, "dynamic_output_budget": false, "affected_questions": 0, "min_effective_output_cap": 98304}, {"model": "v3-agentic-sft-1144", "questions": 1600, "max_prompt_tokens": 13586, "output_cap": 98304, "context": 131072, "dynamic_output_budget": false, "affected_questions": 0, "min_effective_output_cap": 98304}, {"model": "olmo3-32b-think", "questions": 1600, "max_prompt_tokens": 13591, "output_cap": 61440, "context": 65536, "dynamic_output_budget": true, "affected_questions": 5, "min_effective_output_cap": 51945}]
|
|
|
|
|
|
sft-agentic-stage2/README.md
CHANGED
|
@@ -22,3 +22,17 @@ Each benchmark folder contains `trajectories.jsonl`, `judge_trajectories.jsonl`,
|
|
| 22 |
No generation or judging was performed for this upload. Existing completed results were packaged directly on AWS FSx. Credential values are excluded/redacted. `manifest.json` records sizes and SHA-256 hashes. `provenance.json` contains exact frozen evaluation configuration and source identities.
|
| 23 |
|
| 24 |
Non-agentic output cap: 98,304 tokens; temperature1, top-p1, top-k disabled, presence penalty0, thinking enabled. B200 inference TP1. Agentic evaluations use temperature1 and two attempts per task. TB retains native task timeouts; SciCoding token budgets and both benchmarks' complete settings are recorded in the frozen plan.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
No generation or judging was performed for this upload. Existing completed results were packaged directly on AWS FSx. Credential values are excluded/redacted. `manifest.json` records sizes and SHA-256 hashes. `provenance.json` contains exact frozen evaluation configuration and source identities.
|
| 23 |
|
| 24 |
Non-agentic output cap: 98,304 tokens; temperature1, top-p1, top-k disabled, presence penalty0, thinking enabled. B200 inference TP1. Agentic evaluations use temperature1 and two attempts per task. TB retains native task timeouts; SciCoding token budgets and both benchmarks' complete settings are recorded in the frozen plan.
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
## HLE-Verified (2026-10-30)
|
| 28 |
+
|
| 29 |
+
All **1,600 text questions scored**, with no remaining errors. Overall accuracy: **500/1,600 (31.25%)**.
|
| 30 |
+
|
| 31 |
+
| Subset | Correct / questions | Accuracy | Artifacts |
|
| 32 |
+
|---|---:|---:|---|
|
| 33 |
+
| Gold | 165/575 | 28.70% | [Scores, trajectories and judgments](hle-verified-gold/) |
|
| 34 |
+
| Revision | 335/1025 | 32.68% | [Scores, trajectories and judgments](hle-verified-revision/) |
|
| 35 |
+
|
| 36 |
+
Pinned `skylenage-ai/HLE-Verified` revision `0bc83643672d4f68a5f89998617a639d85e7318b`; Gold and Revision text-only, images and Uncertain excluded. One response per question, temperature 1, top-p 1, seed 373363. Output cap 98,304 tokens; context 131,072 tokens. GPT-5.6 Luna judges final answers using the HLE reference-comparison grader with an 8,192-token judge budget. This is not an official full-HLE leaderboard reproduction. Overall accuracy weights all questions equally, rather than averaging the two subset percentages.
|
| 37 |
+
|
| 38 |
+
[Frozen settings, source hashes, artifact hashes and recovery provenance](hle-verified-provenance.json). Each subset folder retains full saved reasoning/final answers, judgments, summaries, events and configs.
|
sft-agentic-stage2/hle-verified-provenance.json
ADDED
|
@@ -0,0 +1,394 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"original_publication_revision": "4954a184445b672cc8050619bce956558f8af8f6",
|
| 4 |
+
"reorganization": "Metadata only; no generation, regrading or trajectory changes.",
|
| 5 |
+
"result": {
|
| 6 |
+
"model": "v3-agentic-sft-1144",
|
| 7 |
+
"folder": "sft-agentic-stage2",
|
| 8 |
+
"subsets": {
|
| 9 |
+
"gold": {
|
| 10 |
+
"correct": 165,
|
| 11 |
+
"total": 575,
|
| 12 |
+
"accuracy": 0.28695652173913044,
|
| 13 |
+
"cap_hits": 27
|
| 14 |
+
},
|
| 15 |
+
"revision": {
|
| 16 |
+
"correct": 335,
|
| 17 |
+
"total": 1025,
|
| 18 |
+
"accuracy": 0.32682926829268294,
|
| 19 |
+
"cap_hits": 56
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"correct": 500,
|
| 23 |
+
"total": 1600,
|
| 24 |
+
"overall_accuracy": 0.3125
|
| 25 |
+
},
|
| 26 |
+
"overall_definition": "(Gold correct + Revision correct) / 1600",
|
| 27 |
+
"model_plan": {
|
| 28 |
+
"name": "v3-agentic-sft-1144",
|
| 29 |
+
"profile": {
|
| 30 |
+
"name": "qwen35-9b",
|
| 31 |
+
"gpu": "h100",
|
| 32 |
+
"model_family": "qwen",
|
| 33 |
+
"checkpoint": {
|
| 34 |
+
"model_type": "qwen3_5_text",
|
| 35 |
+
"hidden_size": 4096,
|
| 36 |
+
"num_hidden_layers": 32,
|
| 37 |
+
"vocab_size": 248320
|
| 38 |
+
},
|
| 39 |
+
"generation": {
|
| 40 |
+
"temperature": 1.0,
|
| 41 |
+
"top_p": 1.0,
|
| 42 |
+
"max_tokens": 98304,
|
| 43 |
+
"seed": 373363,
|
| 44 |
+
"extra": {
|
| 45 |
+
"top_k": -1,
|
| 46 |
+
"min_p": 0.0,
|
| 47 |
+
"presence_penalty": 0.0,
|
| 48 |
+
"repetition_penalty": 1.0,
|
| 49 |
+
"chat_template_kwargs": {
|
| 50 |
+
"enable_thinking": true
|
| 51 |
+
}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"serving": {
|
| 55 |
+
"context_length": 131072,
|
| 56 |
+
"dtype": "bfloat16",
|
| 57 |
+
"attention_backend": "triton",
|
| 58 |
+
"mem_fraction_static": 0.8,
|
| 59 |
+
"reasoning_parser": "qwen3",
|
| 60 |
+
"model_config_parser": "sci_qwen35_text",
|
| 61 |
+
"external_model_package": "sci_qwen35_text_model",
|
| 62 |
+
"extra_args": [
|
| 63 |
+
"--sampling-backend",
|
| 64 |
+
"flashinfer",
|
| 65 |
+
"--cuda-graph-backend-decode",
|
| 66 |
+
"full",
|
| 67 |
+
"--cuda-graph-backend-prefill",
|
| 68 |
+
"disabled"
|
| 69 |
+
],
|
| 70 |
+
"tp": 2,
|
| 71 |
+
"replicas_per_node": 4,
|
| 72 |
+
"max_running_requests": 32
|
| 73 |
+
},
|
| 74 |
+
"reasoning": {
|
| 75 |
+
"requests_per_engine": 32,
|
| 76 |
+
"benchmarks": {}
|
| 77 |
+
},
|
| 78 |
+
"agentic_tasks_per_engine": 1
|
| 79 |
+
},
|
| 80 |
+
"node": 2,
|
| 81 |
+
"checkpoint": "/workspace/data/raw_model/sci-ckpt-final/06129b3983eab7e8fe01e5c30577e76250605c66/sft-agentic/iter_0001144",
|
| 82 |
+
"repo": "JLiangHe/sci-ckpt-final/sft-agentic/iter_0001144",
|
| 83 |
+
"revision": "06129b3983eab7e8fe01e5c30577e76250605c66",
|
| 84 |
+
"checkpoint_config_sha256": "7ed31d50f8c0e8dc8219aa686a97c08665b8d9c85321e4c2fb766139fb6d01af",
|
| 85 |
+
"index_sha256": "a2507a6b56459640fc02cefe71a21ce253574d5aab3f023f6336ce980f5c955d",
|
| 86 |
+
"shard_sizes": {
|
| 87 |
+
"model-00001-of-00004.safetensors": 5276436016,
|
| 88 |
+
"model-00002-of-00004.safetensors": 5335160776,
|
| 89 |
+
"model-00003-of-00004.safetensors": 4932508344,
|
| 90 |
+
"model-00004-of-00004.safetensors": 2363550592
|
| 91 |
+
},
|
| 92 |
+
"endpoints": [
|
| 93 |
+
{
|
| 94 |
+
"id": "node-2-replica-0",
|
| 95 |
+
"api_base": "http://10.4.70.214:8700/v1"
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"id": "node-2-replica-1",
|
| 99 |
+
"api_base": "http://10.4.70.214:8701/v1"
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"id": "node-2-replica-2",
|
| 103 |
+
"api_base": "http://10.4.70.214:8702/v1"
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"id": "node-2-replica-3",
|
| 107 |
+
"api_base": "http://10.4.70.214:8703/v1"
|
| 108 |
+
}
|
| 109 |
+
],
|
| 110 |
+
"root": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/v3-agentic-sft-1144",
|
| 111 |
+
"suite": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030/models/v3-agentic-sft-1144/configs/hle-verified-gold-revision-96k.yaml",
|
| 112 |
+
"concurrency": 128,
|
| 113 |
+
"judge_concurrency": 64,
|
| 114 |
+
"tracking": {
|
| 115 |
+
"enabled": false,
|
| 116 |
+
"reason": "Explicit user approval for this HLE sweep"
|
| 117 |
+
},
|
| 118 |
+
"config_sha256": {
|
| 119 |
+
"model.yaml": "722bc05f950a385bb19f1bb2585b696020c74d3f1330554382405232b555bf5b",
|
| 120 |
+
"configs/hle-verified-gold-revision-96k.yaml": "0f71a6185cd62ac1537c2623305b92cce4c5ebb1facce8b83fbe548bd8f53a72",
|
| 121 |
+
"configs/hle-verified-gold.yaml": "7b8e690edba1b05295faca19fcc24308fbb450f3c86b2b48abb0d11afd3b10f4",
|
| 122 |
+
"configs/hle-verified-revision.yaml": "201f06bf26074d9ab2c8c14f3fc4717599a8bec992af909739124d4f36082bfc"
|
| 123 |
+
}
|
| 124 |
+
},
|
| 125 |
+
"dataset": {
|
| 126 |
+
"dataset": "skylenage-ai/HLE-Verified",
|
| 127 |
+
"revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 128 |
+
"counts": {
|
| 129 |
+
"Gold": {
|
| 130 |
+
"total": 668,
|
| 131 |
+
"text": 575,
|
| 132 |
+
"image_excluded": 93
|
| 133 |
+
},
|
| 134 |
+
"Revision": {
|
| 135 |
+
"total": 1143,
|
| 136 |
+
"text": 1025,
|
| 137 |
+
"image_excluded": 118
|
| 138 |
+
}
|
| 139 |
+
},
|
| 140 |
+
"sha256": {
|
| 141 |
+
"data/Gold_subset.part01.parquet": "0f9347730c0b9a7b690931bfe38f748d2b142be8b4b3318e16d23844b18af98b",
|
| 142 |
+
"data/Gold_subset.part02.parquet": "9661a90148056d7d39e0fef058159e72532119aa8ee30a26f06dfa27b61d015f",
|
| 143 |
+
"data/Gold_subset.part03.parquet": "ace18833057f09e9071be632239b255a1cd23a3a602dddd524b88e573bf427aa",
|
| 144 |
+
"data/Gold_subset.part04.parquet": "f09e14efe051a8a5af54e5f14f0bb2231ddb4b68cbba8886afb2f0a2995e3737",
|
| 145 |
+
"data/Gold_subset.part05.parquet": "23211a403e2c013b01dd8634ffa982b9f6dfcb6a4c126a3223c2a18913030ee2",
|
| 146 |
+
"data/Revision_subset.part01.parquet": "2a0765c0cc3f5b93a42083ea18ab4f8f23fbd59196e01114ed1c4a6b1a756485",
|
| 147 |
+
"data/Revision_subset.part02.parquet": "a7cf12a8eb817aa5ac16a52069f3f8685857d309acd5fb8944401414ce7eed70",
|
| 148 |
+
"data/Revision_subset.part03.parquet": "ed1016a6f67aba3dfbadd9cc42f22c049bb7e3ab7d1dad1dfd33b2556fa0d6f8",
|
| 149 |
+
"data/Revision_subset.part04.parquet": "e33e244b55c8b7da7a9e2b6c1e4cf0195b645fc633d25c76a0a19e0990996ea8",
|
| 150 |
+
"data/Revision_subset.part05.parquet": "acfd27b451d93fb6359c9e915efb3928ed24a28ddc8159e13f523e96de6d2d1d",
|
| 151 |
+
"gold-text.jsonl": "731e0bb6c7cc9d382e853c0c399d75127733e635fd3d51d8425feb9adc065a6d",
|
| 152 |
+
"revision-text.jsonl": "8d9aa587fab3bb711928102622653cbeba86c1949c014ada66c198a65b9d65d4"
|
| 153 |
+
}
|
| 154 |
+
},
|
| 155 |
+
"token_preflight": {
|
| 156 |
+
"model": "v3-agentic-sft-1144",
|
| 157 |
+
"questions": 1600,
|
| 158 |
+
"max_prompt_tokens": 13586,
|
| 159 |
+
"output_cap": 98304,
|
| 160 |
+
"context": 131072,
|
| 161 |
+
"dynamic_output_budget": false,
|
| 162 |
+
"affected_questions": 0,
|
| 163 |
+
"min_effective_output_cap": 98304
|
| 164 |
+
},
|
| 165 |
+
"shared_recovery_provenance": {
|
| 166 |
+
"native_runtime": "/workspace/data/sci-llm-training/runtime/hle-seven-models-20261030",
|
| 167 |
+
"publication_reruns": false,
|
| 168 |
+
"all_questions_scored": true,
|
| 169 |
+
"subset_selection": "Gold and Revision text-only; Uncertain and image questions excluded",
|
| 170 |
+
"dataset_revision": "0bc83643672d4f68a5f89998617a639d85e7318b",
|
| 171 |
+
"qwen_export": "Before generation, raw vs text export vocabulary and all 1600 rendered prompt token sequences were verified identical.",
|
| 172 |
+
"gemma_recovery": "One engine exited after evaluation completion. Ownership-aware handover resumed SFT; all Gemma result hashes verified unchanged.",
|
| 173 |
+
"scaleout": "Qwen35B and SFT expanded from one to two pods each. Frozen generation/scoring hashes unchanged; saved records retained byte-identically. Unsaved in-flight generations may have restarted; this was infrastructure recovery, not best-of selection.",
|
| 174 |
+
"judge_artifacts": "Saved native score details/raw Luna responses are retained. No missing native request content is fabricated.",
|
| 175 |
+
"mlflow": "Explicit user-approved no-MLflow run"
|
| 176 |
+
},
|
| 177 |
+
"source_manifest": {
|
| 178 |
+
"git_sha": "e2810931c3520b2fd283c94bb9fb4167aa752344",
|
| 179 |
+
"files": {
|
| 180 |
+
"eval-framework/README.md": "53c46ec67ad9be8d9b821577dac53f41eda18664104b807d740cbca1c0b1f729",
|
| 181 |
+
"eval-framework/benchmarks/README.md": "745a321667f864cc7f2d15d6a7a1d5678116eca48bf7de4082822217f72c341f",
|
| 182 |
+
"eval-framework/benchmarks/arxivmath/README.md": "20e7bfd04794c235fbe86624c6ba1ff685055bce2bb479fceeb8a86d5439d8cc",
|
| 183 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-0226.yaml": "81f517f96bddb7011e423eff80677c4f40bca08a61efaeca0b1ae3cae3bd4bc0",
|
| 184 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-december.yaml": "1cceafd31ea63bc0a98a90c2aafe8543eac71a9d7f9b893542df85da2b361bce",
|
| 185 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-february.yaml": "15054966cdbc0832c7ed66bd3e57ef3de007ab6f0e0306d3aad220024acdbad2",
|
| 186 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-january.yaml": "6d28d616d81b303c6c62d89579352b46c309ef24fe6c3b31f91308fbb1079577",
|
| 187 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-march.yaml": "284a3f7c69986bfbebbcedabe84a45f653598cf9d4484b5011517d79b3646b5f",
|
| 188 |
+
"eval-framework/benchmarks/arxivmath/configs/matharena-arxivmath-official.yaml": "ee748365823a2eb6410211717a173e0cd5b0ada4bd1b71cf14328219864de820",
|
| 189 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-64k-avg10.yaml": "3dec288838a60a47fd2847dc59376a01854f086be1da1ef0d8d91382fffa13c7",
|
| 190 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-0226-96k-avg10.yaml": "a04846fe0e115ed42b3ca1e4cf70903369c63c732b595b520c02756ba56593ca",
|
| 191 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-64k-avg4.yaml": "1318f298c7db93a28bd6ba5a86788b3cf163d7a4bfa9822e206f3a444d270f57",
|
| 192 |
+
"eval-framework/benchmarks/arxivmath/configs/qwen35-arxivmath-official-96k-avg4.yaml": "07328616dbdf2890b9ac0b845b1abd6f630b79b1363b34ec334456a00f9b0c8e",
|
| 193 |
+
"eval-framework/benchmarks/arxivmath/scripts/run_arxivmath.sh": "52ca53730d0da1944186b94a9d50b24dd775192fd048bfda0e558e9d21227692",
|
| 194 |
+
"eval-framework/benchmarks/frontier-science/README.md": "41c6c273b405048365b33c97ac8c0c9d48df1a54714661a4f521a5e168aeeac5",
|
| 195 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-64k.yaml": "9a5bb0e367abf03672f4ebc37ac788f11a05841f5b20a30568ce788660bad48c",
|
| 196 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-olympiad-thinking-trajectories-96k.yaml": "a3b01461445b252b4f478ab8d6be06b74c0292abc07b28b6b6599ff100544452",
|
| 197 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-64k.yaml": "b1d06b3a0b9e8b7ba48defa260018af9b985c9c00a029ac98823f615848a2d8e",
|
| 198 |
+
"eval-framework/benchmarks/frontier-science/configs/frontierscience-research-thinking-trajectories-96k.yaml": "036e3b18f4ae70a01ca7c14708edad2bf81c89376ceafb7daca1d910c57cfc59",
|
| 199 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-dsflash-frontierscience-64k-r30.yaml": "374e33fb3fde6f275a85c3b5a32e80b0b9832bb7b45029eee191acd9b3af3582",
|
| 200 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-96k-r30.yaml": "876cf7692ce8c6629f162f47602dea0b724882e37990c107a2a295aa01dfc855",
|
| 201 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-olympiad-96k-r30.yaml": "78137bb0ea1a2474af51e881f3beaa1a8dc27eb8c0ec2eb4fb3f139e84c556cb",
|
| 202 |
+
"eval-framework/benchmarks/frontier-science/configs/qwen35-frontierscience-research-96k-r30.yaml": "15f118314a4e5615173030114a47956860736ea322d4f73a2be819cfa133310f",
|
| 203 |
+
"eval-framework/benchmarks/hle-verified/README.md": "b8479f43d8360015737a40fa64d6599d008e9746bb740fd00b34768da9308fef",
|
| 204 |
+
"eval-framework/benchmarks/hle-verified/scripts/prepare_suite.py": "2b0a9338a0f227856ff3ec7832a124931eff4b0c2db76fd28c55cd97f44c6263",
|
| 205 |
+
"eval-framework/benchmarks/scicode-verified/.gitignore": "b31f667b3aa816369e85a811bbcb5a56c8cc1fe811d9b204b7cbf041e4f64cd0",
|
| 206 |
+
"eval-framework/benchmarks/scicode-verified/Dockerfile": "1c72b688ea139757ee0e22a7680602097725110c6bc3088aea86383a04f218d5",
|
| 207 |
+
"eval-framework/benchmarks/scicode-verified/README.md": "d349063417f2dbb569974bf8bac5f20df32c9403286c5a6fa9061b4f117ec0b8",
|
| 208 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2024.txt": "96b6a6189023658e0fa2f3b8e41750d350bf17ae91246054e84c89bab4536436",
|
| 209 |
+
"eval-framework/benchmarks/scicode-verified/requirements-2025.txt": "914e3ab53523e0323d6216a8ce34c01877bf71be3dd8bfd2498e1d16bea186ca",
|
| 210 |
+
"eval-framework/benchmarks/scicode-verified/scripts/install_helpers.py": "5a34e6e0f27a6e38ad342c975c0cc74e788b6f31833386b8c1b31312c5e5d858",
|
| 211 |
+
"eval-framework/benchmarks/scicode-verified/scripts/regrade.py": "5b363c3fd8dca09f6b5142e72499ff02009493cbf2556f7ea2909648cbe84309",
|
| 212 |
+
"eval-framework/benchmarks/scicode-verified/scripts/render_pod.py": "0a557dd8b71a2149b30a45ca5c85680ea2b4205e32c20b4c72df0031b08b3a3b",
|
| 213 |
+
"eval-framework/benchmarks/scicode-verified/scripts/run.py": "2c1020fbf9782594bdc160bd75f2c35277e5f730f7162dea0338ea0f59a9430a",
|
| 214 |
+
"eval-framework/benchmarks/scicode-verified/scripts/smoke_grader.py": "1e163e2d2c34700e4e909b433695afd03b093c9916aeeb49b54269f5eeb7558e",
|
| 215 |
+
"eval-framework/benchmarks/scicode-verified/tests/test_runner.py": "fb0a236253bc26a8746688c79021d061d787e6a79779e3c9dd735226918d9a76",
|
| 216 |
+
"eval-framework/benchmarks/scicode-verified/vendor/README.md": "ae36c9370feb39922c0c09e97893ccf84afd4579f85ceb87d10763ea8d8221e0",
|
| 217 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
|
| 218 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
|
| 219 |
+
"eval-framework/benchmarks/scicode-verified/vendor/compare/cmp.py": "8b3e790cc29dc87ef1f0f1b9bfd68c877728c78c40b541a69d10015221e9e235",
|
| 220 |
+
"eval-framework/benchmarks/scicoding-agent/README.md": "d4b80a2823458bbc0f62afb4f48f8c3ed50cde093b914a777fae66a1fde6cce3",
|
| 221 |
+
"eval-framework/benchmarks/scicoding-agent/harness/__init__.py": "4ccaffbb46e9585d7fa8a5cd810082b855a7c66a01e7574e5e31c9a04b8da266",
|
| 222 |
+
"eval-framework/benchmarks/scicoding-agent/harness/constants.py": "718bfe6e812c6f47cb6190f5476403e048095390232c2368ffcf3142c2a4f4c1",
|
| 223 |
+
"eval-framework/benchmarks/scicoding-agent/harness/facet_runtime.py": "b1c5c5af8ed9a165194b74191a5a9d76a740d5dee33abfb3de2f75622e992617",
|
| 224 |
+
"eval-framework/benchmarks/scicoding-agent/harness/provenance.json": "756b0587fdf37ae8233cdcc390f08ee9c27e193c3152162e11ff11e37f33dd92",
|
| 225 |
+
"eval-framework/benchmarks/scicoding-agent/harness/reasoning_budget.py": "7309e3c11b3e271d0023e379ee3edab0f7d713edc621c37543fde5575f67418c",
|
| 226 |
+
"eval-framework/benchmarks/scicoding-agent/harness/terminus_distillation.py": "37e61ea9fce55a24b0d5f2d5eb04555546dcf0e5d410ab4c61015aa38eaa9965",
|
| 227 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/prepare.py": "9cc53bc0c6e4b090336e5bd5124641fcbe2e7488910b2e4642aff063373c0e18",
|
| 228 |
+
"eval-framework/benchmarks/scicoding-agent/scripts/run_trial.py": "ebb89ad022b3ecf172774cbe66b8ae62746c9e2f54121e25eb5701f9a4922c63",
|
| 229 |
+
"eval-framework/benchmarks/terminal-bench-2.1/README.md": "ef5382e2095afb42756fb560f6f19f73f7d4120fa27420ba25dc7c6c6785094a",
|
| 230 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shard.py": "1ea3f7b2267c2cfc6077c17afc32f22999c10eeecec38eabcea0832bab0d29fe",
|
| 231 |
+
"eval-framework/benchmarks/terminal-bench-2.1/scripts/run_shared_queue.py": "0ae05be22ecdf31ddd316ac52ebcf575cf6c34f016ccc7295d2e3ea4062d5a41",
|
| 232 |
+
"eval-framework/benchmarks/terminal-bench-science/README.md": "ed6000fc0a58c5ddef71f2dc6f998387b6cd4a1fb8f2517c5407b6608026743b",
|
| 233 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/qwen35-9b-it-v0.1.json": "9be4f8f2eb0f94b849f54c9263a0b7e8bc66765d9a0c78633e1ddc2dd930e780",
|
| 234 |
+
"eval-framework/benchmarks/terminal-bench-science/configs/task-ids-v0.1.txt": "642cfe18685b8b7a2358a0c7b46c7527306109c297f2ed79c627ccbe518b979c",
|
| 235 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/harbor_runner_pod.py": "c86667a49088853d4d262426353bfbd6a6aca0df8cc9bc94c4711d58c2c89a75",
|
| 236 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/prepare_shared_queue.py": "16c15adafb57a1b0b7e904e4a743ea3ef16eae8026fb5c4c27e9c24872ad8816",
|
| 237 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/request_worker.py": "e95313feaf39a4703882da1bd8ee20acf42e9b68af72802914f61febfd57909c",
|
| 238 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/submit.py": "781ff39ee1c6c044f49a5a872286d7b281b644103baced1e8047d8021efc917d",
|
| 239 |
+
"eval-framework/benchmarks/terminal-bench-science/scripts/terminal_bench_science.sh": "5d025e8ffcd5a397bbbe73eece4e6776515fbc9a94b280ab3b168501251db361",
|
| 240 |
+
"eval-framework/benchmarks/terminal-bench-science/tests/test_integration.py": "9418422f3841d6ab9f72739acea6b39289f98589a4e8dab799a5d73839cd88a3",
|
| 241 |
+
"eval-framework/configs/AGENTIC_EVAL.md": "389c4e6fd66e5764b093a8e9fa05ed2be0a5c13342a15a2e069d27218a86d1d7",
|
| 242 |
+
"eval-framework/configs/CHECKPOINT_EVAL.md": "a14998923dee0c5a9b958db8a05e26540c9b77e9eb9249b9bd01ba0719831d41",
|
| 243 |
+
"eval-framework/configs/VERIFIER_RECOVERY.md": "003b923fb978273159c13a5078a83c3f6cb079f17cec8325ed0e142567e197bc",
|
| 244 |
+
"eval-framework/configs/agentic-eval.json": "a7347054ae76e7124643faa3f77166497b146490b41bdb3af198328426cf24ff",
|
| 245 |
+
"eval-framework/configs/agentic-sampling.json": "b055372cb817f779ec9616f64563bbcb462ac0e63fb9c70303cc06c1975cca76",
|
| 246 |
+
"eval-framework/configs/checkpoint-eval.json": "0c4f4c702e3efe61e9a7ecc7675d86dd05aa61ff553be6b2f9997ef9d22e48c9",
|
| 247 |
+
"eval-framework/eval.sh": "8ebb50e285b83281f5f246a7d76ec1c1703450279a4a80b190023024718cea6f",
|
| 248 |
+
"eval-framework/judges/gpt-5.6-luna-frontierscience-judge.yaml": "3b4d1703428a4b1383a05e0d73fad2a9f21aa87e6f09ac4af5929538aeb0912a",
|
| 249 |
+
"eval-framework/models/deepseek-flash0731-high-96k.yaml": "9a3f9cf9d275e4a47340c9ad50150a4965b31a4afd293edaea455187e716ac43",
|
| 250 |
+
"eval-framework/models/qwen35-mixed-model-32x96k.yaml": "c263c7428786bb65d6712061fc3c3302a00bd03e65882d16251ce58b9803218e",
|
| 251 |
+
"eval-framework/models/qwen35-mixed-model-64x96k.yaml": "2e8542dc811039d06722c44cfb0398040746f4b1b967e7617ebf3f319a004ac8",
|
| 252 |
+
"eval-framework/profiles/README.md": "85ea539887794749b8a316fdeae5eb2a24da729625858c13b592e812409be43d",
|
| 253 |
+
"eval-framework/profiles/deepseek-flash0731.yaml": "7bc7f399e462cb697952ece3f49854f6a69955dc5c7effc766b48f2a07c0a0c9",
|
| 254 |
+
"eval-framework/profiles/gemma4-12b.yaml": "54bfb0d7399770538131b0d878738b40f92bd60e91a42acc3f737447d4aa2bed",
|
| 255 |
+
"eval-framework/profiles/gpt-oss-120b.yaml": "0c6e21a6e705f3b27fc45f32ce72540c08cd8097d8c0ab9f3202dbc31a256dfc",
|
| 256 |
+
"eval-framework/profiles/gpt-oss-20b.yaml": "40ec7148c8affe5f18d73433c58ce4d739dc9a175148fa8f007b8672f0f6e79b",
|
| 257 |
+
"eval-framework/profiles/nemotron35-lightning-30b-a3b.yaml": "d125c5efeba8becb19cbac8a3696d7401587b12ae60e0f8a937eaace160626fe",
|
| 258 |
+
"eval-framework/profiles/olmo3-32b-think.yaml": "4e39de22d90bbd49a2e9349412fb65f8dc43c0e388ed62dee9601a87450868a0",
|
| 259 |
+
"eval-framework/profiles/qwen35-35b-a3b.yaml": "e80cc18da70e71a02418dfedc596d98f48c81ed862236dfd73a815ea88cf9bd2",
|
| 260 |
+
"eval-framework/profiles/qwen35-9b.yaml": "cfc1e4f9120106dad3fe43d0d1a7ce55472279103ac8e39d338b2eb86da339e4",
|
| 261 |
+
"eval-framework/scripts/agentic_harmony.py": "c5778bdeeb90c280b004552a09adf65dc6856b2597127a16f902cfc37e74438a",
|
| 262 |
+
"eval-framework/scripts/agentic_plan.py": "5b15625933e97fb2c9552b7d7933ab931b1a76a8c05f832cf5822481eba8777a",
|
| 263 |
+
"eval-framework/scripts/agentic_queue.py": "a10c337b6c869837e3df6b720072801251c70cb9901de0af02cad4ebb7fd9b6f",
|
| 264 |
+
"eval-framework/scripts/agentic_recovery.py": "5902ff269956a74a1334dd9a9a9375284d06c97c09d416df1b4e88540340c6bd",
|
| 265 |
+
"eval-framework/scripts/agentic_sampling.py": "437f51d4ad3109d1736f7631bf4aa8bfd55b7bae1e85300077238cd7cf73efe5",
|
| 266 |
+
"eval-framework/scripts/checkpoint_eval.py": "cfceb8e614aeeb4836d1c99e49fb5e6484883099904afbc38b692d9497da62b5",
|
| 267 |
+
"eval-framework/scripts/checkpoint_io.py": "8ded23c709c003d74c30ff1ec194245aa23d21452115641de1985f4ac3dc6b03",
|
| 268 |
+
"eval-framework/scripts/endpoint_health.py": "31629053e17b571f33dabfde694363d1d7b38ea54519653be41a14ea50f59356",
|
| 269 |
+
"eval-framework/scripts/eval_profile.py": "5a99d56637a8210cedd97c006793e6e3365f5515778ee744338045ee7749d211",
|
| 270 |
+
"eval-framework/scripts/export_results.py": "3aec4e729f9b4b7d63fd9c183c07226a5e3350a7e5991a7e0c58738f3d4c0d15",
|
| 271 |
+
"eval-framework/scripts/prepare_agentic_harbor.py": "30590b04d8189c64bb2d7a494d482a04757cbffefd6bbab8271b6f157cdecf66",
|
| 272 |
+
"eval-framework/scripts/rescore_agentic_snapshot.py": "cdaabca41a1f74f525e33b19d1abc7737b9edbc9b644fc96f609bc3f91cc6768",
|
| 273 |
+
"eval-framework/scripts/run_agentic_harbor.py": "2779eacbb427a623b29454fea987f54ff7b58df427b0c63cc4232f36b2ba23fa",
|
| 274 |
+
"eval-framework/scripts/run_suite.sh": "c55623478b13f11667090c3cd37b3b7a560edfaa218fd38f416876dbe06d242f",
|
| 275 |
+
"eval-framework/scripts/scicode_suite.py": "796339c4bd89e5be7562552f0b7f213109398f97aadd226c5eaa4492526d1157",
|
| 276 |
+
"eval-framework/scripts/serve_sglang_node.sh": "d599fa4a2100be1f802209f2150bcd2529b01672bd249c9f7dfaa527dc944d24",
|
| 277 |
+
"eval-framework/scripts/start_checkpoint_eval.py": "0f7814c206b805aaf679caa49d2b68eaee258319ad8cbe1d6775cb89624001d9",
|
| 278 |
+
"eval-framework/scripts/verifier_health.py": "046d069aa0ed0ebf238b410eab98a9e50074254aedd7370041b4191594893f60",
|
| 279 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/__init__.py": "632612ae71e30ebfe66eeaaaf201aa687d1ab20cc9a36b8b337d6f95e505a0a9",
|
| 280 |
+
"eval-framework/serving/sglang_external_models/gemma4_compat/model.py": "c73f576392dda11d1b9adb76233282f65dbc620d66c9f5e7e5d1aaf0037abeba",
|
| 281 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/__init__.py": "e70dc084d24abfa4b9578f47e0e224d8e91771b449c9479449938831055503eb",
|
| 282 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/qwen35_text.py": "a7232c4a5b097909a1ce2939eb2d5e143961d76e2abd02501720dbe033d07465",
|
| 283 |
+
"eval-framework/serving/sglang_external_models/sci_qwen35_text_model/sglang_config.py": "ba306790d7eba6b770d239027226e456bb55526cbba4025902bb8314bbaa7276",
|
| 284 |
+
"eval-framework/serving/sglang_external_models/sitecustomize.py": "f3a3bcaefe09508be16d8a08a552625ca3065a68c77ec613ce101dfd56cb6000",
|
| 285 |
+
"eval-framework/tests/RECOVERY_RELEASE_20261026.md": "575798deb30fc34edab6e02eb83e90bea94657c16488baacc69f30bcceda60d4",
|
| 286 |
+
"eval-framework/tests/test_agentic_harmony.py": "4434a7018c9c0b798ad9f93ec45d9f92d1de161e44caff1311504d1cc56635dc",
|
| 287 |
+
"eval-framework/tests/test_agentic_queue_profiles.py": "7a8c94bb4c19a16d2e5cd7373f443d678d033c178f81e343d8ae55ce55b4a3c6",
|
| 288 |
+
"eval-framework/tests/test_agentic_recovery.py": "5cd8b5d38cce2c1e0a0d25c5fdbf696cc00f1e54cbb92a1569e29c1252f3b663",
|
| 289 |
+
"eval-framework/tests/test_arxivmath_selection.py": "0e3634c234f37de1f9468f3cdf16f5d3809b124a76e707ce30854a2e612d5362",
|
| 290 |
+
"eval-framework/tests/test_checkpoint_groups.py": "5b111988f9ad347fba573eac26000e87b2634087440d7b176bb2cf9d87b31c85",
|
| 291 |
+
"eval-framework/tests/test_endpoint_health.py": "3d665ff629a99db383aff66756303b3a7dfdfab3ea2d9ceb371abed557528c15",
|
| 292 |
+
"eval-framework/tests/test_eval_profiles.py": "3762f96c33e9b80e7853c83d58a2b49baacbaad6c09ec83c631e31313a376dcf",
|
| 293 |
+
"eval-framework/tests/test_export_results.py": "766bb3fa7ce580b8839c2e9248589f2aede7589b8eca3c3265b948e5926e3f3c",
|
| 294 |
+
"eval-framework/tests/test_frontierscience_official.py": "10f71356dcb913c14bd32e872d04dd8f1d41d45fddf5d96f340b261296fc57b1",
|
| 295 |
+
"eval-framework/tests/test_hle_verified.py": "427621e09c71e294a78ed45be51a5be4314f882840ec5f4531c4214741b829d5",
|
| 296 |
+
"eval-framework/tests/test_judge_contract.py": "ccb087f205cd0a96668ade0ed6782d92058adf51ba3d88e381acad5eabc67190",
|
| 297 |
+
"eval-framework/tests/test_layout.py": "e1925b7d173101ecb722ed637ee711e5ed2517c03ba7bc4148dba686944c55dd",
|
| 298 |
+
"eval-framework/tests/test_one_line_eval.py": "144987804029ef32a46bdac1561585a0a22e54c123a44592f155dd99b41f5fa3",
|
| 299 |
+
"eval-framework/tests/test_openai_backend.py": "53b8f6d658069b72d9188c280788201c775c2c8b2aefa8f42a9e9bd53149517d",
|
| 300 |
+
"eval-framework/tests/test_prompt_stage_outcome.py": "26b75edc9185582361e7537aecd5159c95c15acc6c0b00d56db658b0a8d10f92",
|
| 301 |
+
"eval-framework/tests/test_rescore_agentic_snapshot.py": "ee882c65cde964d13e3a4f0f1284b260b5004ec96e60ece4d41c921d3491b5cf",
|
| 302 |
+
"eval-framework/tests/test_scicode_suite.py": "569037b6892327429a5efe09fcfd7c65ad7e1cc5bea3d706876ff363c486470e",
|
| 303 |
+
"eval-framework/tests/test_scicoding_length_recovery.py": "d460ec5f5489d6f92905237584885407edd4f5ed63a6798ab8aa34d8c18ddb4b",
|
| 304 |
+
"eval-framework/tests/test_scicoding_output_recovery.py": "01cbaed0ef6d4f535d8a767e708409cb61906257acdee12c87f9cafca83a6f45",
|
| 305 |
+
"eval-framework/tests/test_scicoding_sandbox_identity.py": "31fb2036b98e19fbeea03e79246128ad0f8d02385aebe9d43d8e0099c1132e50",
|
| 306 |
+
"eval-framework/tests/test_scicoding_tokenizer_provenance.py": "3eb776c10883333259dfd19ea705a4152225eddcf2f66c0edffc249b9b892a5c",
|
| 307 |
+
"eval-framework/tests/test_server_lifecycle.py": "316880f817c995f012ba2ed3fd158776db1cd424b0f9e7af359d2a3e4af4b906",
|
| 308 |
+
"eval-framework/tests/test_sft_final_checkpoint.py": "416a179105907016dfdbdd3bbd4668af7663271070b2271c4609db889e292ef4",
|
| 309 |
+
"eval-framework/tests/test_suite_scheduling.py": "759376108d58a0e9ba290d47c6dabe1933da2d166b02c1f56aca2ee025c83099",
|
| 310 |
+
"eval-framework/tests/test_terminal_bench_queue.py": "ef42c06385334aff5d08c73af1331459b2b7ff94a439580af78c351852f391fb",
|
| 311 |
+
"eval-framework/tests/test_terminal_bench_sharding.py": "c91336fdf26c8ef513eb86e865aec74d6feb707357456ea04c456cd5765a091c",
|
| 312 |
+
"eval-framework/vendor/SCI_EVAL_SOURCE_COMMIT": "c3c8007c60b8e203023028263beb848e5780cdfeaf528aa7131b36a91ed4a29d",
|
| 313 |
+
"eval-framework/vendor/src/scieval/__init__.py": "19d667e1beb23f97356ee5fd5bffbbf28929d1580cdd407ee8fe25c7c2fe4510",
|
| 314 |
+
"eval-framework/vendor/src/scieval/_vendor/__init__.py": "686340d0be9cf7d81f553cba5d6fce041444a17a46d92d8cc2bdfcc640710ed9",
|
| 315 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/EED.py": "ccabaa6dd7793f7fb1295a9436c365abb3806c8c42b486e192a59c8b49c9c955",
|
| 316 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/LICENSE": "8d4f6779f96c77767b40ff51fc3b8123505e548718576772af7d5ec595c610dd",
|
| 317 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/__init__.py": "4d1e465078b502aaba5c7730bcb95eb20e6c5897afbed5753c04179e7998905f",
|
| 318 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/extended_zss.py": "7fbfbd623f49d2bb3ebd4fd73150f335aa9237ce4f3a3f13832e8bace4ee6624",
|
| 319 |
+
"eval-framework/vendor/src/scieval/_vendor/phybench_eed/latex_pre_process.py": "b8766b59bdadcee1a995e9e582e88877ee11fec83b8b829054e599d018e598a5",
|
| 320 |
+
"eval-framework/vendor/src/scieval/agentic.py": "358744588fe07c21c277557bd9122ea56fb9d9992f7a01868cf7220fd5f032da",
|
| 321 |
+
"eval-framework/vendor/src/scieval/backends/__init__.py": "01430eaff184aca2a3e4a5099bf360770ed92d3b35f7aafe2da70191135d8454",
|
| 322 |
+
"eval-framework/vendor/src/scieval/backends/base.py": "108e00432b8ed8baaf2c5da94b6cd9f979aecd46a2d356760cd086a84119b059",
|
| 323 |
+
"eval-framework/vendor/src/scieval/backends/bigbang_tools.py": "ca56cc65ac79d8fd91fc25670bb095f0eb00b7f26ca167685b591dd949538ece",
|
| 324 |
+
"eval-framework/vendor/src/scieval/backends/kubernetes_code_exec.py": "6548ea6964ce38777e7a274938b73e31ae18dae4aff045ec6d9783513654dadc",
|
| 325 |
+
"eval-framework/vendor/src/scieval/backends/openai_compatible.py": "0409964b292218e13b2953d6a063bf81221b3e2b432db8c5b75596ad96e4f368",
|
| 326 |
+
"eval-framework/vendor/src/scieval/backends/registry.py": "30cc109218351ba08a6631e98daf05d7cd3a840cb8f5e84494a73e46895ff05c",
|
| 327 |
+
"eval-framework/vendor/src/scieval/backends/usage_recording.py": "ed2610f830ce7f19aaa01a50fa1631af8e85580fb89492934589ae4d6fbc65d0",
|
| 328 |
+
"eval-framework/vendor/src/scieval/benchmarks/__init__.py": "a805a2a219c5c420c70606f54f2a3bb7c4dc7b414bcdc08c22fa235909d6645f",
|
| 329 |
+
"eval-framework/vendor/src/scieval/benchmarks/base.py": "6de50922aaa9c8c83a4b4f566fdddc63e5df4b0e1f8f2da00c649be880b4b5dc",
|
| 330 |
+
"eval-framework/vendor/src/scieval/benchmarks/boxed_math.py": "b635652f5b443909ac9bea2469ef1968a60a7e0ae037e10014fadfd1ca004767",
|
| 331 |
+
"eval-framework/vendor/src/scieval/benchmarks/data.py": "3cbe22d485592eaddcacca92b84621e155ace93e53974c76992a7068addcfd6a",
|
| 332 |
+
"eval-framework/vendor/src/scieval/benchmarks/frontierscience.py": "d1acab1dd5d53f2a4e1c64d8106e2ada3bdeb614734d90ddc3c7157e668a9c3a",
|
| 333 |
+
"eval-framework/vendor/src/scieval/benchmarks/hle.py": "7fe7ad72875a12c392860630210010a1e18c65840c52b53e5796d632bc8cf35e",
|
| 334 |
+
"eval-framework/vendor/src/scieval/benchmarks/matharena_arxivmath.py": "65113884207d1694b026bd9aaea49c906359bbd51f15f5cbbca9f77d0d0471d1",
|
| 335 |
+
"eval-framework/vendor/src/scieval/benchmarks/registry.py": "10879afe7f8f83e521ff5b225959b53606af7d062531f5d82f44187f3879c1a3",
|
| 336 |
+
"eval-framework/vendor/src/scieval/benchmarks/scoring.py": "7c2af930faf2b0cdab62253e9c1c443b89c66a97e3ca04d2e5a6866616e73ca5",
|
| 337 |
+
"eval-framework/vendor/src/scieval/cli.py": "5e0cb762bc7428418433876b14b648b152314a55e7faf9e15a86ebdca3a2b709",
|
| 338 |
+
"eval-framework/vendor/src/scieval/config.py": "95043b5cd443615ede853d5a925583470ed989361bdaf3b4761c17e3e61907ea",
|
| 339 |
+
"eval-framework/vendor/src/scieval/judge_defaults.py": "25e0b9065ae231d78382b5b4f07619c78ca95cf6e7e142747d6461e7165c9939",
|
| 340 |
+
"eval-framework/vendor/src/scieval/native.py": "f51fc840c23a438799ed98a3e17183ee61df9fdc859580d414dd9e110a7eb917",
|
| 341 |
+
"eval-framework/vendor/src/scieval/polar_evaluator.py": "1bca84bf1e3ca1eb34ce50f66ad094ca560e5215d5d2667a7c75ece493d433eb",
|
| 342 |
+
"eval-framework/vendor/src/scieval/polar_harness.py": "db21c93ddd0d58982abfd2ca421635551005b73ca462f85a2b18dea813b0f92f",
|
| 343 |
+
"eval-framework/vendor/src/scieval/polar_harness_command.py": "69fefbc78137a52dca374768e72ddad2cba3f93f8b3a8196896a588a793e018c",
|
| 344 |
+
"eval-framework/vendor/src/scieval/polar_harness_v2.py": "78fbf1f8891b3b6947057b2135d6bf6d50cf2dfb9ef2e0021e7ecbdee7f9015c",
|
| 345 |
+
"eval-framework/vendor/src/scieval/runner.py": "f4d95b14f29779cde6db604051a401c09e8f2fa9b584999cbb2752e4984e9d8a",
|
| 346 |
+
"eval-framework/vendor/src/scieval/suite.py": "e8b3b4996edd481896167334db704b8ee77620c8f488c50bc34c9a42b52ecadf",
|
| 347 |
+
"eval-framework/vendor/src/scieval/token_cap.py": "b177220aace70b8b60d75f2928b7d1f669f15e0d988853bb649b484d49bc996b",
|
| 348 |
+
"eval-framework/vendor/src/scieval/types.py": "41eda562665e33e5fa07f3cb6be1a4f933110c2d89d3cb2139cbe73165a45120"
|
| 349 |
+
},
|
| 350 |
+
"scope": "Current origin/main plus isolated reviewed profiles and prior evaluation runtime fixes."
|
| 351 |
+
},
|
| 352 |
+
"artifact_manifest": {
|
| 353 |
+
"sft-agentic-stage2/hle-verified-gold/configs.json": {
|
| 354 |
+
"size": 97117,
|
| 355 |
+
"sha256": "e47530b3d95ee8e5058347d6a7c496f650a6075c8cbf390894cb4a7b04b56403"
|
| 356 |
+
},
|
| 357 |
+
"sft-agentic-stage2/hle-verified-gold/events.jsonl.gz": {
|
| 358 |
+
"size": 40849329,
|
| 359 |
+
"sha256": "6354006b0987e42aa8623cf384aceb56fe4ded2453242a1f7d0df0775c0e7774"
|
| 360 |
+
},
|
| 361 |
+
"sft-agentic-stage2/hle-verified-gold/judge_trajectories.jsonl": {
|
| 362 |
+
"size": 808092,
|
| 363 |
+
"sha256": "1966a9542e50ad4611569df699749f49fe6c7ab38348fe9a4878101dcd06935e"
|
| 364 |
+
},
|
| 365 |
+
"sft-agentic-stage2/hle-verified-gold/summary.json": {
|
| 366 |
+
"size": 1511,
|
| 367 |
+
"sha256": "12864d1f6cb71017dfd770bedeb26332721afc6f3762e9d5681907506f24af93"
|
| 368 |
+
},
|
| 369 |
+
"sft-agentic-stage2/hle-verified-gold/trajectories.jsonl": {
|
| 370 |
+
"size": 67049194,
|
| 371 |
+
"sha256": "f0db38689f2018b6192d85e48a8b669a7ea017e703f0d33f4f2de1b18476fc05"
|
| 372 |
+
},
|
| 373 |
+
"sft-agentic-stage2/hle-verified-revision/configs.json": {
|
| 374 |
+
"size": 165549,
|
| 375 |
+
"sha256": "fb902ff8e06c3fc9fbdcaa5d9c3e0fad16fc64a806970868fb29c2bb433ca9ac"
|
| 376 |
+
},
|
| 377 |
+
"sft-agentic-stage2/hle-verified-revision/events.jsonl.gz": {
|
| 378 |
+
"size": 75369864,
|
| 379 |
+
"sha256": "d8a9d38adaed854200ec0c7a7cb534740deca65e8db71df68ac151d68f60d622"
|
| 380 |
+
},
|
| 381 |
+
"sft-agentic-stage2/hle-verified-revision/judge_trajectories.jsonl": {
|
| 382 |
+
"size": 1459606,
|
| 383 |
+
"sha256": "8277bf3b4053d472660fde2105a30fdcb5e7acabfb7f58385800a42e05dc6dfd"
|
| 384 |
+
},
|
| 385 |
+
"sft-agentic-stage2/hle-verified-revision/summary.json": {
|
| 386 |
+
"size": 1518,
|
| 387 |
+
"sha256": "07d966e568f850d54878e50b80158a1d4f26ca4392b23c7fa84de07cb8d1f737"
|
| 388 |
+
},
|
| 389 |
+
"sft-agentic-stage2/hle-verified-revision/trajectories.jsonl": {
|
| 390 |
+
"size": 122766620,
|
| 391 |
+
"sha256": "cb3df4e9d3f79da9ad894fe892273d9d66dd1ada40200f7b61946f010d866465"
|
| 392 |
+
}
|
| 393 |
+
}
|
| 394 |
+
}
|
sft-agentic-stage2/summary.json
CHANGED
|
@@ -1 +1,472 @@
|
|
| 1 |
-
{"checkpoint_step": 1144, "full_suite_complete": true, "benchmarks": {"arxivmath": {"benchmark": "arxivmath", "releases": [{"total_records": 68, "completed": 68, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.3235294117647059, "accuracy": 0.3235294117647059, "mean_latency_seconds": 1057.760202811014, "finish_reason_counts": {"length": 9, "stop": 59}, "length_limited_response_count": 9, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.3235294117647059, "competition_metrics": {"arxiv/december": {"n_problems_scored": 17, "n_trajectories_scored": 68, "reported_accuracy": 0.3235294117647059}}}, "benchmark": "matharena-arxivmath-december", "model": "final-906a58d3597d", "execution_mode": "non_agent", "generation_hash": "9d0c325edaabbad88f8ff1a7b8e57cd2ed8cf47979a80569dfbb946ec4c32b07", "scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661", "selected": 68, "incomplete": 0, "generated_now": 68, "scored_now": 68, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 68, "generation_concurrency": 1536, "judge_concurrency": 256, "updated_at": "2026-10-27T16:20:49.471305+00:00"}, {"total_records": 128, "completed": 128, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.203125, "accuracy": 0.203125, "mean_latency_seconds": 909.8116732325425, "finish_reason_counts": {"length": 15, "stop": 113}, "length_limited_response_count": 15, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.203125, "competition_metrics": {"arxiv/february": {"n_problems_scored": 32, "n_trajectories_scored": 128, "reported_accuracy": 0.203125}}}, "benchmark": "matharena-arxivmath-february", "model": "final-906a58d3597d", "execution_mode": "non_agent", "generation_hash": "40fd8c27589b994327f32df17472fc57d08352b08ca0b3d7323ca728f30b88c3", "scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5", "selected": 128, "incomplete": 0, "generated_now": 128, "scored_now": 128, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 128, "generation_concurrency": 1536, "judge_concurrency": 256, "updated_at": "2026-10-27T16:21:25.513362+00:00"}, {"total_records": 92, "completed": 92, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.45652173913043476, "accuracy": 0.45652173913043476, "mean_latency_seconds": 1134.4229743046883, "finish_reason_counts": {"length": 13, "stop": 79}, "length_limited_response_count": 13, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.45652173913043476, "competition_metrics": {"arxiv/january": {"n_problems_scored": 23, "n_trajectories_scored": 92, "reported_accuracy": 0.45652173913043476}}}, "benchmark": "matharena-arxivmath-january", "model": "final-906a58d3597d", "execution_mode": "non_agent", "generation_hash": "dc8eb2459c44b34550eababb7c1a8b71403e3ca9f571f2c689152783ff6e20c4", "scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8", "selected": 92, "incomplete": 0, "generated_now": 92, "scored_now": 92, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 92, "generation_concurrency": 1536, "judge_concurrency": 256, "updated_at": "2026-10-27T16:19:56.930413+00:00"}, {"total_records": 124, "completed": 124, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.23387096774193547, "accuracy": 0.23387096774193547, "mean_latency_seconds": 818.3343730721108, "finish_reason_counts": {"length": 7, "stop": 117}, "length_limited_response_count": 7, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 0, "request_count": 0, "retry_request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "judge_api_usage": {"request_count": 0, "prompt_tokens": 0, "cached_prompt_tokens": 0, "cache_write_prompt_tokens": 0, "completion_tokens": 0, "reasoning_tokens": 0, "total_tokens": 0}, "benchmark_metrics": {"competition": "arxivmath", "aggregation": "avg@4", "n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.23387096774193547, "competition_metrics": {"arxiv/march": {"n_problems_scored": 31, "n_trajectories_scored": 124, "reported_accuracy": 0.23387096774193547}}}, "benchmark": "matharena-arxivmath-march", "model": "final-906a58d3597d", "execution_mode": "non_agent", "generation_hash": "a2960b182a433b8fb322d71183e52bb2199b198c3b15790c65f3e5a2870340f7", "scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8", "selected": 124, "incomplete": 0, "generated_now": 124, "scored_now": 124, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 124, "generation_concurrency": 1536, "judge_concurrency": 256, "updated_at": "2026-10-27T16:20:29.602872+00:00"}], "total_records": 412, "completed": 412, "accuracy": 0.28883495145631066}, "scicode-verified": {"updated_at": 1793119295.3793085, "expected_problems": 64, "expected_steps": 287, "generated_steps": 287, "scored_steps": 287, "passed_steps": 127, "completed_problems": 64, "passed_problems": 8, "subproblem_accuracy": 0.4425087108013937, "main_problem_accuracy": 0.125, "complete": true, "cap_hits": 0, "mean_output_tokens": 15117.979094076654, "grading_infrastructure_errors": 0, "errors": []}, "scicoding-agent": {"tasks": 50, "attempts": 100, "attempts_per_task": 2, "metric": "avg@2", "passed": 41, "avg_at_2": 0.41, "unscored": 0, "agent_timeouts": 0, "reward_positive_with_exception": 11, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1"}, "frontierscience-research": {"total_records": 1800, "completed": 1800, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.29590305555555557, "accuracy": 0.07333333333333333, "mean_latency_seconds": 505.6271946268312, "finish_reason_counts": {"length": 25, "stop": 1775}, "length_limited_response_count": 25, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 1800, "request_count": 1820, "retry_request_count": 20, "prompt_tokens": 9522776, "cached_prompt_tokens": 124123, "cache_write_prompt_tokens": 9371007, "completion_tokens": 2939042, "reasoning_tokens": 1718732, "total_tokens": 12461818}, "judge_api_usage": {"request_count": 1820, "prompt_tokens": 9522776, "cached_prompt_tokens": 124123, "cache_write_prompt_tokens": 9371007, "completion_tokens": 2939042, "reasoning_tokens": 1718732, "total_tokens": 12461818}, "benchmark_metrics": {"track": "research", "n_problems_scored": 60, "n_trajectories_scored": 1800, "aggregation": "avg@30", "reported_accuracy": 0.07333333333333332, "mean_rubric_score": 0.29590305555555557, "success_threshold": 0.7, "community_profile": null}, "benchmark": "frontierscience-research", "model": "final-906a58d3597d", "execution_mode": "non_agent", "generation_hash": "9a53ae203c2f853346e633dc54dcc75c5a80b034f5c7c0dfcdc7a3157ca88a61", "scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750", "selected": 1800, "incomplete": 0, "generated_now": 1800, "scored_now": 1800, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 1800, "generation_concurrency": 1536, "judge_concurrency": 256, "updated_at": "2026-10-27T16:08:48.021990+00:00"}, "frontierscience-olympiad": {"total_records": 3000, "completed": 3000, "query_errors": 0, "score_errors": 0, "generated_unscored": 0, "mean_score": 0.614, "accuracy": 0.614, "mean_latency_seconds": 341.9275932600237, "finish_reason_counts": {"length": 18, "stop": 2982}, "length_limited_response_count": 18, "observed_max_completion_tokens": 98304, "judge_usage": {"judged_record_count": 3000, "request_count": 3000, "retry_request_count": 0, "prompt_tokens": 2460555, "cached_prompt_tokens": 4873, "cache_write_prompt_tokens": 625442, "completion_tokens": 822086, "reasoning_tokens": 596849, "total_tokens": 3282641}, "judge_api_usage": {"request_count": 3000, "prompt_tokens": 2460555, "cached_prompt_tokens": 4873, "cache_write_prompt_tokens": 625442, "completion_tokens": 822086, "reasoning_tokens": 596849, "total_tokens": 3282641}, "benchmark_metrics": {"track": "olympiad", "n_problems_scored": 100, "n_trajectories_scored": 3000, "aggregation": "avg@30", "reported_accuracy": 0.614, "mean_rubric_score": null, "success_threshold": 0.5, "community_profile": null}, "benchmark": "frontierscience-olympiad", "model": "final-906a58d3597d", "execution_mode": "non_agent", "generation_hash": "3cf730a629fac5a44682882a563d9765f9effe71282b54a3b4257c954f2b94bb", "scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016", "selected": 3000, "incomplete": 0, "generated_now": 3000, "scored_now": 3000, "cached_completed": 0, "stage": "all", "worker_count": 1, "worker_index": 0, "worker_owned_predictions": 3000, "generation_concurrency": 1536, "judge_concurrency": 256, "updated_at": "2026-10-27T16:17:54.676278+00:00"}, "terminal-bench-2.1": {"tasks": 89, "attempts": 178, "attempts_per_task": 2, "metric": "avg@2", "passed": 42, "avg_at_2": 0.23595505617977527, "unscored": 0, "agent_timeouts": 105, "reward_positive_with_exception": 4, "scoring": "native_verifier", "scoring_policy": "native-verifier-after-agent-stop-v1"}}, "pending_benchmarks": []}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"checkpoint_step": 1144,
|
| 3 |
+
"full_suite_complete": true,
|
| 4 |
+
"benchmarks": {
|
| 5 |
+
"arxivmath": {
|
| 6 |
+
"benchmark": "arxivmath",
|
| 7 |
+
"releases": [
|
| 8 |
+
{
|
| 9 |
+
"total_records": 68,
|
| 10 |
+
"completed": 68,
|
| 11 |
+
"query_errors": 0,
|
| 12 |
+
"score_errors": 0,
|
| 13 |
+
"generated_unscored": 0,
|
| 14 |
+
"mean_score": 0.3235294117647059,
|
| 15 |
+
"accuracy": 0.3235294117647059,
|
| 16 |
+
"mean_latency_seconds": 1057.760202811014,
|
| 17 |
+
"finish_reason_counts": {
|
| 18 |
+
"length": 9,
|
| 19 |
+
"stop": 59
|
| 20 |
+
},
|
| 21 |
+
"length_limited_response_count": 9,
|
| 22 |
+
"observed_max_completion_tokens": 98304,
|
| 23 |
+
"judge_usage": {
|
| 24 |
+
"judged_record_count": 0,
|
| 25 |
+
"request_count": 0,
|
| 26 |
+
"retry_request_count": 0,
|
| 27 |
+
"prompt_tokens": 0,
|
| 28 |
+
"cached_prompt_tokens": 0,
|
| 29 |
+
"cache_write_prompt_tokens": 0,
|
| 30 |
+
"completion_tokens": 0,
|
| 31 |
+
"reasoning_tokens": 0,
|
| 32 |
+
"total_tokens": 0
|
| 33 |
+
},
|
| 34 |
+
"judge_api_usage": {
|
| 35 |
+
"request_count": 0,
|
| 36 |
+
"prompt_tokens": 0,
|
| 37 |
+
"cached_prompt_tokens": 0,
|
| 38 |
+
"cache_write_prompt_tokens": 0,
|
| 39 |
+
"completion_tokens": 0,
|
| 40 |
+
"reasoning_tokens": 0,
|
| 41 |
+
"total_tokens": 0
|
| 42 |
+
},
|
| 43 |
+
"benchmark_metrics": {
|
| 44 |
+
"competition": "arxivmath",
|
| 45 |
+
"aggregation": "avg@4",
|
| 46 |
+
"n_problems_scored": 17,
|
| 47 |
+
"n_trajectories_scored": 68,
|
| 48 |
+
"reported_accuracy": 0.3235294117647059,
|
| 49 |
+
"competition_metrics": {
|
| 50 |
+
"arxiv/december": {
|
| 51 |
+
"n_problems_scored": 17,
|
| 52 |
+
"n_trajectories_scored": 68,
|
| 53 |
+
"reported_accuracy": 0.3235294117647059
|
| 54 |
+
}
|
| 55 |
+
}
|
| 56 |
+
},
|
| 57 |
+
"benchmark": "matharena-arxivmath-december",
|
| 58 |
+
"model": "final-906a58d3597d",
|
| 59 |
+
"execution_mode": "non_agent",
|
| 60 |
+
"generation_hash": "9d0c325edaabbad88f8ff1a7b8e57cd2ed8cf47979a80569dfbb946ec4c32b07",
|
| 61 |
+
"scoring_hash": "a3e2b1125bcb225515024853daf2e70da3b83f9b59856ada480339b2eb926661",
|
| 62 |
+
"selected": 68,
|
| 63 |
+
"incomplete": 0,
|
| 64 |
+
"generated_now": 68,
|
| 65 |
+
"scored_now": 68,
|
| 66 |
+
"cached_completed": 0,
|
| 67 |
+
"stage": "all",
|
| 68 |
+
"worker_count": 1,
|
| 69 |
+
"worker_index": 0,
|
| 70 |
+
"worker_owned_predictions": 68,
|
| 71 |
+
"generation_concurrency": 1536,
|
| 72 |
+
"judge_concurrency": 256,
|
| 73 |
+
"updated_at": "2026-10-27T16:20:49.471305+00:00"
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"total_records": 128,
|
| 77 |
+
"completed": 128,
|
| 78 |
+
"query_errors": 0,
|
| 79 |
+
"score_errors": 0,
|
| 80 |
+
"generated_unscored": 0,
|
| 81 |
+
"mean_score": 0.203125,
|
| 82 |
+
"accuracy": 0.203125,
|
| 83 |
+
"mean_latency_seconds": 909.8116732325425,
|
| 84 |
+
"finish_reason_counts": {
|
| 85 |
+
"length": 15,
|
| 86 |
+
"stop": 113
|
| 87 |
+
},
|
| 88 |
+
"length_limited_response_count": 15,
|
| 89 |
+
"observed_max_completion_tokens": 98304,
|
| 90 |
+
"judge_usage": {
|
| 91 |
+
"judged_record_count": 0,
|
| 92 |
+
"request_count": 0,
|
| 93 |
+
"retry_request_count": 0,
|
| 94 |
+
"prompt_tokens": 0,
|
| 95 |
+
"cached_prompt_tokens": 0,
|
| 96 |
+
"cache_write_prompt_tokens": 0,
|
| 97 |
+
"completion_tokens": 0,
|
| 98 |
+
"reasoning_tokens": 0,
|
| 99 |
+
"total_tokens": 0
|
| 100 |
+
},
|
| 101 |
+
"judge_api_usage": {
|
| 102 |
+
"request_count": 0,
|
| 103 |
+
"prompt_tokens": 0,
|
| 104 |
+
"cached_prompt_tokens": 0,
|
| 105 |
+
"cache_write_prompt_tokens": 0,
|
| 106 |
+
"completion_tokens": 0,
|
| 107 |
+
"reasoning_tokens": 0,
|
| 108 |
+
"total_tokens": 0
|
| 109 |
+
},
|
| 110 |
+
"benchmark_metrics": {
|
| 111 |
+
"competition": "arxivmath",
|
| 112 |
+
"aggregation": "avg@4",
|
| 113 |
+
"n_problems_scored": 32,
|
| 114 |
+
"n_trajectories_scored": 128,
|
| 115 |
+
"reported_accuracy": 0.203125,
|
| 116 |
+
"competition_metrics": {
|
| 117 |
+
"arxiv/february": {
|
| 118 |
+
"n_problems_scored": 32,
|
| 119 |
+
"n_trajectories_scored": 128,
|
| 120 |
+
"reported_accuracy": 0.203125
|
| 121 |
+
}
|
| 122 |
+
}
|
| 123 |
+
},
|
| 124 |
+
"benchmark": "matharena-arxivmath-february",
|
| 125 |
+
"model": "final-906a58d3597d",
|
| 126 |
+
"execution_mode": "non_agent",
|
| 127 |
+
"generation_hash": "40fd8c27589b994327f32df17472fc57d08352b08ca0b3d7323ca728f30b88c3",
|
| 128 |
+
"scoring_hash": "17f11f51e8e8063d80cf440ac0ca8e971e8fee7ed79f96a8d9867e872d10faa5",
|
| 129 |
+
"selected": 128,
|
| 130 |
+
"incomplete": 0,
|
| 131 |
+
"generated_now": 128,
|
| 132 |
+
"scored_now": 128,
|
| 133 |
+
"cached_completed": 0,
|
| 134 |
+
"stage": "all",
|
| 135 |
+
"worker_count": 1,
|
| 136 |
+
"worker_index": 0,
|
| 137 |
+
"worker_owned_predictions": 128,
|
| 138 |
+
"generation_concurrency": 1536,
|
| 139 |
+
"judge_concurrency": 256,
|
| 140 |
+
"updated_at": "2026-10-27T16:21:25.513362+00:00"
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"total_records": 92,
|
| 144 |
+
"completed": 92,
|
| 145 |
+
"query_errors": 0,
|
| 146 |
+
"score_errors": 0,
|
| 147 |
+
"generated_unscored": 0,
|
| 148 |
+
"mean_score": 0.45652173913043476,
|
| 149 |
+
"accuracy": 0.45652173913043476,
|
| 150 |
+
"mean_latency_seconds": 1134.4229743046883,
|
| 151 |
+
"finish_reason_counts": {
|
| 152 |
+
"length": 13,
|
| 153 |
+
"stop": 79
|
| 154 |
+
},
|
| 155 |
+
"length_limited_response_count": 13,
|
| 156 |
+
"observed_max_completion_tokens": 98304,
|
| 157 |
+
"judge_usage": {
|
| 158 |
+
"judged_record_count": 0,
|
| 159 |
+
"request_count": 0,
|
| 160 |
+
"retry_request_count": 0,
|
| 161 |
+
"prompt_tokens": 0,
|
| 162 |
+
"cached_prompt_tokens": 0,
|
| 163 |
+
"cache_write_prompt_tokens": 0,
|
| 164 |
+
"completion_tokens": 0,
|
| 165 |
+
"reasoning_tokens": 0,
|
| 166 |
+
"total_tokens": 0
|
| 167 |
+
},
|
| 168 |
+
"judge_api_usage": {
|
| 169 |
+
"request_count": 0,
|
| 170 |
+
"prompt_tokens": 0,
|
| 171 |
+
"cached_prompt_tokens": 0,
|
| 172 |
+
"cache_write_prompt_tokens": 0,
|
| 173 |
+
"completion_tokens": 0,
|
| 174 |
+
"reasoning_tokens": 0,
|
| 175 |
+
"total_tokens": 0
|
| 176 |
+
},
|
| 177 |
+
"benchmark_metrics": {
|
| 178 |
+
"competition": "arxivmath",
|
| 179 |
+
"aggregation": "avg@4",
|
| 180 |
+
"n_problems_scored": 23,
|
| 181 |
+
"n_trajectories_scored": 92,
|
| 182 |
+
"reported_accuracy": 0.45652173913043476,
|
| 183 |
+
"competition_metrics": {
|
| 184 |
+
"arxiv/january": {
|
| 185 |
+
"n_problems_scored": 23,
|
| 186 |
+
"n_trajectories_scored": 92,
|
| 187 |
+
"reported_accuracy": 0.45652173913043476
|
| 188 |
+
}
|
| 189 |
+
}
|
| 190 |
+
},
|
| 191 |
+
"benchmark": "matharena-arxivmath-january",
|
| 192 |
+
"model": "final-906a58d3597d",
|
| 193 |
+
"execution_mode": "non_agent",
|
| 194 |
+
"generation_hash": "dc8eb2459c44b34550eababb7c1a8b71403e3ca9f571f2c689152783ff6e20c4",
|
| 195 |
+
"scoring_hash": "e09d955ed6a3fccc283a4ea838cd7802e2927196a3fa75d3fb16070088ce2bb8",
|
| 196 |
+
"selected": 92,
|
| 197 |
+
"incomplete": 0,
|
| 198 |
+
"generated_now": 92,
|
| 199 |
+
"scored_now": 92,
|
| 200 |
+
"cached_completed": 0,
|
| 201 |
+
"stage": "all",
|
| 202 |
+
"worker_count": 1,
|
| 203 |
+
"worker_index": 0,
|
| 204 |
+
"worker_owned_predictions": 92,
|
| 205 |
+
"generation_concurrency": 1536,
|
| 206 |
+
"judge_concurrency": 256,
|
| 207 |
+
"updated_at": "2026-10-27T16:19:56.930413+00:00"
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"total_records": 124,
|
| 211 |
+
"completed": 124,
|
| 212 |
+
"query_errors": 0,
|
| 213 |
+
"score_errors": 0,
|
| 214 |
+
"generated_unscored": 0,
|
| 215 |
+
"mean_score": 0.23387096774193547,
|
| 216 |
+
"accuracy": 0.23387096774193547,
|
| 217 |
+
"mean_latency_seconds": 818.3343730721108,
|
| 218 |
+
"finish_reason_counts": {
|
| 219 |
+
"length": 7,
|
| 220 |
+
"stop": 117
|
| 221 |
+
},
|
| 222 |
+
"length_limited_response_count": 7,
|
| 223 |
+
"observed_max_completion_tokens": 98304,
|
| 224 |
+
"judge_usage": {
|
| 225 |
+
"judged_record_count": 0,
|
| 226 |
+
"request_count": 0,
|
| 227 |
+
"retry_request_count": 0,
|
| 228 |
+
"prompt_tokens": 0,
|
| 229 |
+
"cached_prompt_tokens": 0,
|
| 230 |
+
"cache_write_prompt_tokens": 0,
|
| 231 |
+
"completion_tokens": 0,
|
| 232 |
+
"reasoning_tokens": 0,
|
| 233 |
+
"total_tokens": 0
|
| 234 |
+
},
|
| 235 |
+
"judge_api_usage": {
|
| 236 |
+
"request_count": 0,
|
| 237 |
+
"prompt_tokens": 0,
|
| 238 |
+
"cached_prompt_tokens": 0,
|
| 239 |
+
"cache_write_prompt_tokens": 0,
|
| 240 |
+
"completion_tokens": 0,
|
| 241 |
+
"reasoning_tokens": 0,
|
| 242 |
+
"total_tokens": 0
|
| 243 |
+
},
|
| 244 |
+
"benchmark_metrics": {
|
| 245 |
+
"competition": "arxivmath",
|
| 246 |
+
"aggregation": "avg@4",
|
| 247 |
+
"n_problems_scored": 31,
|
| 248 |
+
"n_trajectories_scored": 124,
|
| 249 |
+
"reported_accuracy": 0.23387096774193547,
|
| 250 |
+
"competition_metrics": {
|
| 251 |
+
"arxiv/march": {
|
| 252 |
+
"n_problems_scored": 31,
|
| 253 |
+
"n_trajectories_scored": 124,
|
| 254 |
+
"reported_accuracy": 0.23387096774193547
|
| 255 |
+
}
|
| 256 |
+
}
|
| 257 |
+
},
|
| 258 |
+
"benchmark": "matharena-arxivmath-march",
|
| 259 |
+
"model": "final-906a58d3597d",
|
| 260 |
+
"execution_mode": "non_agent",
|
| 261 |
+
"generation_hash": "a2960b182a433b8fb322d71183e52bb2199b198c3b15790c65f3e5a2870340f7",
|
| 262 |
+
"scoring_hash": "2dd97483dbb1256f5c3074e2aedd129a0b86523ee141a54afcf1e6d115ea8db8",
|
| 263 |
+
"selected": 124,
|
| 264 |
+
"incomplete": 0,
|
| 265 |
+
"generated_now": 124,
|
| 266 |
+
"scored_now": 124,
|
| 267 |
+
"cached_completed": 0,
|
| 268 |
+
"stage": "all",
|
| 269 |
+
"worker_count": 1,
|
| 270 |
+
"worker_index": 0,
|
| 271 |
+
"worker_owned_predictions": 124,
|
| 272 |
+
"generation_concurrency": 1536,
|
| 273 |
+
"judge_concurrency": 256,
|
| 274 |
+
"updated_at": "2026-10-27T16:20:29.602872+00:00"
|
| 275 |
+
}
|
| 276 |
+
],
|
| 277 |
+
"total_records": 412,
|
| 278 |
+
"completed": 412,
|
| 279 |
+
"accuracy": 0.28883495145631066
|
| 280 |
+
},
|
| 281 |
+
"scicode-verified": {
|
| 282 |
+
"updated_at": 1793119295.3793085,
|
| 283 |
+
"expected_problems": 64,
|
| 284 |
+
"expected_steps": 287,
|
| 285 |
+
"generated_steps": 287,
|
| 286 |
+
"scored_steps": 287,
|
| 287 |
+
"passed_steps": 127,
|
| 288 |
+
"completed_problems": 64,
|
| 289 |
+
"passed_problems": 8,
|
| 290 |
+
"subproblem_accuracy": 0.4425087108013937,
|
| 291 |
+
"main_problem_accuracy": 0.125,
|
| 292 |
+
"complete": true,
|
| 293 |
+
"cap_hits": 0,
|
| 294 |
+
"mean_output_tokens": 15117.979094076654,
|
| 295 |
+
"grading_infrastructure_errors": 0,
|
| 296 |
+
"errors": []
|
| 297 |
+
},
|
| 298 |
+
"scicoding-agent": {
|
| 299 |
+
"tasks": 50,
|
| 300 |
+
"attempts": 100,
|
| 301 |
+
"attempts_per_task": 2,
|
| 302 |
+
"metric": "avg@2",
|
| 303 |
+
"passed": 41,
|
| 304 |
+
"avg_at_2": 0.41,
|
| 305 |
+
"unscored": 0,
|
| 306 |
+
"agent_timeouts": 0,
|
| 307 |
+
"reward_positive_with_exception": 11,
|
| 308 |
+
"scoring": "native_verifier",
|
| 309 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1"
|
| 310 |
+
},
|
| 311 |
+
"frontierscience-research": {
|
| 312 |
+
"total_records": 1800,
|
| 313 |
+
"completed": 1800,
|
| 314 |
+
"query_errors": 0,
|
| 315 |
+
"score_errors": 0,
|
| 316 |
+
"generated_unscored": 0,
|
| 317 |
+
"mean_score": 0.29590305555555557,
|
| 318 |
+
"accuracy": 0.07333333333333333,
|
| 319 |
+
"mean_latency_seconds": 505.6271946268312,
|
| 320 |
+
"finish_reason_counts": {
|
| 321 |
+
"length": 25,
|
| 322 |
+
"stop": 1775
|
| 323 |
+
},
|
| 324 |
+
"length_limited_response_count": 25,
|
| 325 |
+
"observed_max_completion_tokens": 98304,
|
| 326 |
+
"judge_usage": {
|
| 327 |
+
"judged_record_count": 1800,
|
| 328 |
+
"request_count": 1820,
|
| 329 |
+
"retry_request_count": 20,
|
| 330 |
+
"prompt_tokens": 9522776,
|
| 331 |
+
"cached_prompt_tokens": 124123,
|
| 332 |
+
"cache_write_prompt_tokens": 9371007,
|
| 333 |
+
"completion_tokens": 2939042,
|
| 334 |
+
"reasoning_tokens": 1718732,
|
| 335 |
+
"total_tokens": 12461818
|
| 336 |
+
},
|
| 337 |
+
"judge_api_usage": {
|
| 338 |
+
"request_count": 1820,
|
| 339 |
+
"prompt_tokens": 9522776,
|
| 340 |
+
"cached_prompt_tokens": 124123,
|
| 341 |
+
"cache_write_prompt_tokens": 9371007,
|
| 342 |
+
"completion_tokens": 2939042,
|
| 343 |
+
"reasoning_tokens": 1718732,
|
| 344 |
+
"total_tokens": 12461818
|
| 345 |
+
},
|
| 346 |
+
"benchmark_metrics": {
|
| 347 |
+
"track": "research",
|
| 348 |
+
"n_problems_scored": 60,
|
| 349 |
+
"n_trajectories_scored": 1800,
|
| 350 |
+
"aggregation": "avg@30",
|
| 351 |
+
"reported_accuracy": 0.07333333333333332,
|
| 352 |
+
"mean_rubric_score": 0.29590305555555557,
|
| 353 |
+
"success_threshold": 0.7,
|
| 354 |
+
"community_profile": null
|
| 355 |
+
},
|
| 356 |
+
"benchmark": "frontierscience-research",
|
| 357 |
+
"model": "final-906a58d3597d",
|
| 358 |
+
"execution_mode": "non_agent",
|
| 359 |
+
"generation_hash": "9a53ae203c2f853346e633dc54dcc75c5a80b034f5c7c0dfcdc7a3157ca88a61",
|
| 360 |
+
"scoring_hash": "ceea6075bb3e7655a14d6ad2f07750f7b1b588c24da569c70a81bc6665831750",
|
| 361 |
+
"selected": 1800,
|
| 362 |
+
"incomplete": 0,
|
| 363 |
+
"generated_now": 1800,
|
| 364 |
+
"scored_now": 1800,
|
| 365 |
+
"cached_completed": 0,
|
| 366 |
+
"stage": "all",
|
| 367 |
+
"worker_count": 1,
|
| 368 |
+
"worker_index": 0,
|
| 369 |
+
"worker_owned_predictions": 1800,
|
| 370 |
+
"generation_concurrency": 1536,
|
| 371 |
+
"judge_concurrency": 256,
|
| 372 |
+
"updated_at": "2026-10-27T16:08:48.021990+00:00"
|
| 373 |
+
},
|
| 374 |
+
"frontierscience-olympiad": {
|
| 375 |
+
"total_records": 3000,
|
| 376 |
+
"completed": 3000,
|
| 377 |
+
"query_errors": 0,
|
| 378 |
+
"score_errors": 0,
|
| 379 |
+
"generated_unscored": 0,
|
| 380 |
+
"mean_score": 0.614,
|
| 381 |
+
"accuracy": 0.614,
|
| 382 |
+
"mean_latency_seconds": 341.9275932600237,
|
| 383 |
+
"finish_reason_counts": {
|
| 384 |
+
"length": 18,
|
| 385 |
+
"stop": 2982
|
| 386 |
+
},
|
| 387 |
+
"length_limited_response_count": 18,
|
| 388 |
+
"observed_max_completion_tokens": 98304,
|
| 389 |
+
"judge_usage": {
|
| 390 |
+
"judged_record_count": 3000,
|
| 391 |
+
"request_count": 3000,
|
| 392 |
+
"retry_request_count": 0,
|
| 393 |
+
"prompt_tokens": 2460555,
|
| 394 |
+
"cached_prompt_tokens": 4873,
|
| 395 |
+
"cache_write_prompt_tokens": 625442,
|
| 396 |
+
"completion_tokens": 822086,
|
| 397 |
+
"reasoning_tokens": 596849,
|
| 398 |
+
"total_tokens": 3282641
|
| 399 |
+
},
|
| 400 |
+
"judge_api_usage": {
|
| 401 |
+
"request_count": 3000,
|
| 402 |
+
"prompt_tokens": 2460555,
|
| 403 |
+
"cached_prompt_tokens": 4873,
|
| 404 |
+
"cache_write_prompt_tokens": 625442,
|
| 405 |
+
"completion_tokens": 822086,
|
| 406 |
+
"reasoning_tokens": 596849,
|
| 407 |
+
"total_tokens": 3282641
|
| 408 |
+
},
|
| 409 |
+
"benchmark_metrics": {
|
| 410 |
+
"track": "olympiad",
|
| 411 |
+
"n_problems_scored": 100,
|
| 412 |
+
"n_trajectories_scored": 3000,
|
| 413 |
+
"aggregation": "avg@30",
|
| 414 |
+
"reported_accuracy": 0.614,
|
| 415 |
+
"mean_rubric_score": null,
|
| 416 |
+
"success_threshold": 0.5,
|
| 417 |
+
"community_profile": null
|
| 418 |
+
},
|
| 419 |
+
"benchmark": "frontierscience-olympiad",
|
| 420 |
+
"model": "final-906a58d3597d",
|
| 421 |
+
"execution_mode": "non_agent",
|
| 422 |
+
"generation_hash": "3cf730a629fac5a44682882a563d9765f9effe71282b54a3b4257c954f2b94bb",
|
| 423 |
+
"scoring_hash": "c15a7c1751abfa79a462596bae6c2f2b2b33d2a1f26b30753d0cff221feee016",
|
| 424 |
+
"selected": 3000,
|
| 425 |
+
"incomplete": 0,
|
| 426 |
+
"generated_now": 3000,
|
| 427 |
+
"scored_now": 3000,
|
| 428 |
+
"cached_completed": 0,
|
| 429 |
+
"stage": "all",
|
| 430 |
+
"worker_count": 1,
|
| 431 |
+
"worker_index": 0,
|
| 432 |
+
"worker_owned_predictions": 3000,
|
| 433 |
+
"generation_concurrency": 1536,
|
| 434 |
+
"judge_concurrency": 256,
|
| 435 |
+
"updated_at": "2026-10-27T16:17:54.676278+00:00"
|
| 436 |
+
},
|
| 437 |
+
"terminal-bench-2.1": {
|
| 438 |
+
"tasks": 89,
|
| 439 |
+
"attempts": 178,
|
| 440 |
+
"attempts_per_task": 2,
|
| 441 |
+
"metric": "avg@2",
|
| 442 |
+
"passed": 42,
|
| 443 |
+
"avg_at_2": 0.23595505617977527,
|
| 444 |
+
"unscored": 0,
|
| 445 |
+
"agent_timeouts": 105,
|
| 446 |
+
"reward_positive_with_exception": 4,
|
| 447 |
+
"scoring": "native_verifier",
|
| 448 |
+
"scoring_policy": "native-verifier-after-agent-stop-v1"
|
| 449 |
+
},
|
| 450 |
+
"hle-verified": {
|
| 451 |
+
"gold": {
|
| 452 |
+
"correct": 165,
|
| 453 |
+
"total": 575,
|
| 454 |
+
"accuracy": 0.28695652173913044,
|
| 455 |
+
"cap_hits": 27
|
| 456 |
+
},
|
| 457 |
+
"revision": {
|
| 458 |
+
"correct": 335,
|
| 459 |
+
"total": 1025,
|
| 460 |
+
"accuracy": 0.32682926829268294,
|
| 461 |
+
"cap_hits": 56
|
| 462 |
+
},
|
| 463 |
+
"correct": 500,
|
| 464 |
+
"total": 1600,
|
| 465 |
+
"overall_accuracy": 0.3125,
|
| 466 |
+
"all_complete": true,
|
| 467 |
+
"metric": "question-weighted accuracy; one response per question",
|
| 468 |
+
"provenance": "hle-verified-provenance.json"
|
| 469 |
+
}
|
| 470 |
+
},
|
| 471 |
+
"pending_benchmarks": []
|
| 472 |
+
}
|