File size: 9,276 Bytes
182c036 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 | {
"attempt_id": "b4432d6d-0e2d-413d-a4dd-28720b0d335b",
"claims": [
{
"challenge_claim_sha256": "3e23512497a9642162ac69ed2bd7cbcef00962324c530b34e93bc60475893862",
"evidence": {
"me2_asset_present": true,
"me2_dimensions": [
"Macro-Efficiency",
"Macro-Effectiveness",
"Micro-Efficiency",
"Micro-Effectiveness"
]
},
"status": "verified",
"summary": "The pinned arXiv source hash and GitHub revision expose the four ME2 dimensions and the ME2 figure asset.",
"text": "The ME2 principle characterizes reasoning traces along macro/micro granularity and efficiency/effectiveness axes (Figure 2)."
},
{
"challenge_claim_sha256": "6ccf7ef3305d4a60c6456012a0ed4fdd4d59ad28cc2c7dca5f4c0583a848a370",
"evidence": {
"dag_steps": [
"partition",
"build_dag",
"merge_view"
],
"merged_parent_for_merge": [
1,
2
],
"raw_actions": {
"0": "root",
"1": "continue",
"2": "backtrack",
"3": "merge"
},
"raw_parents": {
"0": [],
"1": [
0
],
"2": [
0
],
"3": [
1,
2
]
},
"usage_calls": 3
},
"status": "verified",
"summary": "The released dag_construction package was imported at the pinned revision and produced a local DAG containing continue, backtrack, and merge actions with merged parents [1, 2].",
"text": "The paper represents reasoning traces as DAGs with progression, branching, and merging structures for pairwise evaluation (Figure 3)."
},
{
"challenge_claim_sha256": "db021919e89a8a7b8485b63b85c4e2d00a1ff7477d23fbfaf8310b1a09832d92",
"evidence": {
"architecture": "LlamaForSequenceClassification",
"card_tags": [
"generated_from_trainer",
"trl",
"reward-trainer"
],
"dataset_revision": "0d0752035ea0e8f7d5c28e1e7a7d8f27e2e45d61",
"eval_accuracy": 0.8835227272727273,
"model_revision": "b84f02bf6b4227675284538a4deb82822371ebbd",
"pipeline_tag": "text-classification",
"sample_keys": [
"chosen",
"prompt",
"rejected"
],
"test_examples": 1500,
"training_script": {
"center_rewards_coefficient_present": true,
"load_best_model_at_end": true,
"metric_for_best_model": "accuracy",
"script": "train_rm.sh",
"train_file_arg": "TRM-preference-train.json",
"train_py_present": false,
"validation_file_arg": "TRM-preference-test.json"
}
},
"status": "toy",
"summary": "TRM-Preference exposes pairwise prompt/chosen/rejected records, TRM-8B is a released reward-trainer sequence-classification model, and train_rm.sh references the pairwise train/test JSON files and accuracy selection, but train.py is not present in the pinned repository, so this does not independently rerun the stated Bradley-Terry training implementation.",
"text": "TRM is trained from a TRM-Preference dataset with a Bradley-Terry preference loss to score reasoning trace quality at scale (Section 5.1)."
}
],
"commands": [
"git clone https://github.com/Simplified-Reasoning/TRM.git",
"git checkout --detach 82ac3778aaba9cf63b237b3db434dc2ba813ef29",
"download arXiv e-print 2602.08498 and verify SHA256",
"huggingface_hub dataset_info/model_info at pinned revisions",
"hf_hub_download TRM-preference-test.json, eval_results.json, config.json",
"import released dag_construction package and run deterministic DAG smoke"
],
"generated_at": "2026-07-29T04:05:00+00:00",
"observations": {
"github": {
"dag_package_tests": [
"test_dag_validity.py",
"test_merged_view.py",
"test_parser.py",
"test_partition.py",
"test_pipeline_mocked.py",
"test_voting.py"
],
"dag_steps": [
"partition",
"build_dag",
"merge_view"
],
"file_count": 1058,
"license": "MIT",
"me2_asset_present": true,
"me2_dimensions": [
"Macro-Efficiency",
"Macro-Effectiveness",
"Micro-Efficiency",
"Micro-Effectiveness"
],
"repo": "https://github.com/Simplified-Reasoning/TRM.git",
"revision": "82ac3778aaba9cf63b237b3db434dc2ba813ef29"
},
"paper_source": {
"arxiv_id": "2602.08498",
"source_bytes": 5721189,
"source_sha256": "6c01d28b97ecbc850b2813c3e0af85ce2b2ec57e6fc9173ab6a0e5ac3dfb3f7f"
},
"released_dag_smoke": {
"merged_nodes": [
{
"id": 0,
"last_raw": 0,
"raw": [
0
]
},
{
"id": 1,
"last_raw": 1,
"raw": [
1
]
},
{
"id": 2,
"last_raw": 2,
"raw": [
2
]
},
{
"id": 3,
"last_raw": 3,
"raw": [
3
]
}
],
"merged_parent_for_merge": [
1,
2
],
"raw_actions": {
"0": "root",
"1": "continue",
"2": "backtrack",
"3": "merge"
},
"raw_leaves": [
3
],
"raw_parents": {
"0": [],
"1": [
0
],
"2": [
0
],
"3": [
1,
2
]
},
"usage_calls": 3
},
"tooling": {
"downloads_avoided": [
"TRM-preference-train.json",
"TRM-8B safetensors shards"
],
"git_available": true,
"network_required": true
},
"training_script": {
"center_rewards_coefficient_present": true,
"load_best_model_at_end": true,
"metric_for_best_model": "accuracy",
"script": "train_rm.sh",
"train_file_arg": "TRM-preference-train.json",
"train_py_present": false,
"validation_file_arg": "TRM-preference-test.json"
},
"trm_model": {
"architecture": "LlamaForSequenceClassification",
"card_tags": [
"generated_from_trainer",
"trl",
"reward-trainer"
],
"eval_accuracy": 0.8835227272727273,
"eval_loss": 0.2938343286514282,
"files": {
".gitattributes": 1570,
"README.md": 2204,
"all_results.json": 408,
"chat_template.jinja": 4614,
"config.json": 952,
"eval_results.json": 408,
"model-00001-of-00007.safetensors": 4886466168,
"model-00002-of-00007.safetensors": 4832007448,
"model-00003-of-00007.safetensors": 4999813112,
"model-00004-of-00007.safetensors": 4999813128,
"model-00005-of-00007.safetensors": 4832007496,
"model-00006-of-00007.safetensors": 4999813120,
"model-00007-of-00007.safetensors": 469828232,
"model.safetensors.index.json": 23984,
"special_tokens_map.json": 325,
"tokenizer.json": 17209920,
"tokenizer_config.json": 50554,
"training_args.bin": 6353
},
"license": "apache-2.0",
"model_type": "llama",
"pipeline_tag": "text-classification",
"repo_id": "zzzhr97/TRM-8B",
"revision": "b84f02bf6b4227675284538a4deb82822371ebbd"
},
"trm_preference": {
"files": {
".gitattributes": 2627,
"README.md": 1536,
"TRM-preference-test.json": 28973005,
"TRM-preference-train.json": 2046546117
},
"repo_id": "zzzhr97/TRM-Preference",
"revision": "0d0752035ea0e8f7d5c28e1e7a7d8f27e2e45d61",
"sample_keys": [
"chosen",
"prompt",
"rejected"
],
"sample_pair_roles": {
"chosen": [
"assistant"
],
"rejected": [
"assistant"
]
},
"sample_prompt_roles": [
"user"
],
"test_examples": 1500
},
"webinstruct_processed": {
"files": {
".gitattributes": 2504,
"README.md": 1112,
"test.parquet": 383950,
"train.parquet": 89879756
},
"repo_id": "zzzhr97/WebInstruct-Verified-Processed",
"revision": "7ad04734d39b71dcba52dc399213288a7602e56c"
}
},
"paper_id": "IMFgiWw4jd",
"source_urls": {
"arxiv": "https://arxiv.org/abs/2602.08498",
"github": "https://github.com/Simplified-Reasoning/TRM.git",
"trm_model": "https://huggingface.co/zzzhr97/TRM-8B",
"trm_preference": "https://huggingface.co/datasets/zzzhr97/TRM-Preference",
"webinstruct_processed": "https://huggingface.co/datasets/zzzhr97/WebInstruct-Verified-Processed"
},
"upstream_revision": "arxiv:2602.08498+arxiv-source-sha256:6c01d28b97ecbc850b2813c3e0af85ce2b2ec57e6fc9173ab6a0e5ac3dfb3f7f+github:Simplified-Reasoning/TRM@82ac3778aaba9cf63b237b3db434dc2ba813ef29+hf-model:zzzhr97/TRM-8B@b84f02bf6b4227675284538a4deb82822371ebbd+hf-dataset:zzzhr97/TRM-Preference@0d0752035ea0e8f7d5c28e1e7a7d8f27e2e45d61+hf-dataset:zzzhr97/WebInstruct-Verified-Processed@7ad04734d39b71dcba52dc399213288a7602e56c"
}
|