Build latest book artifacts / build (push) Canceled after 0s
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s
285 lines
6.6 KiB
JSON
285 lines
6.6 KiB
JSON
{
|
|
"experiment": "curly-quote-sft",
|
|
"model": "Qwen/Qwen3-8B",
|
|
"model_open_weight": true,
|
|
"hardware": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
|
|
"data_sha256": {
|
|
"boundary.jsonl": "e342edf5d28257907bbe486c99a91bd40c482838613c59e09f445b3634a7356c",
|
|
"eval.jsonl": "7c54b2bf1f146e6a22fb72460c0a7b67a7a2ad5301e5a05dcf4828b6962a2049",
|
|
"train.jsonl": "9e27158265ab9782951d81a4838e77128061f492466beeddf0b35a27f46cedd6"
|
|
},
|
|
"training_receipt": "validation/train_20260808T040644Z/training_receipt.json",
|
|
"evaluations": {
|
|
"eval_adapted_boundary.json": {
|
|
"label": "adapted",
|
|
"split": "boundary",
|
|
"count": 256,
|
|
"exact": 0.9765625,
|
|
"protected_preservation": 1.0,
|
|
"curly_count_match": 0.9765625,
|
|
"by_kind": {
|
|
"comment": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"english": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"go": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"java": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"javascript": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"json": {
|
|
"count": 16,
|
|
"exact": 0.6875,
|
|
"curly_count_match": 0.6875
|
|
},
|
|
"markdown": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"mixed": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"nested": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"python": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"quote": {
|
|
"count": 16,
|
|
"exact": 0.9375,
|
|
"curly_count_match": 0.9375
|
|
},
|
|
"rust": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"shell": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"sql": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"yaml": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"zh": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
}
|
|
}
|
|
},
|
|
"eval_adapted_eval.json": {
|
|
"label": "adapted",
|
|
"split": "eval",
|
|
"count": 256,
|
|
"exact": 0.96875,
|
|
"protected_preservation": 1.0,
|
|
"curly_count_match": 0.96875,
|
|
"by_kind": {
|
|
"comment": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"english": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"go": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"java": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"javascript": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"json": {
|
|
"count": 16,
|
|
"exact": 0.6875,
|
|
"curly_count_match": 0.6875
|
|
},
|
|
"markdown": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"mixed": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"nested": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"python": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"quote": {
|
|
"count": 16,
|
|
"exact": 0.8125,
|
|
"curly_count_match": 0.8125
|
|
},
|
|
"rust": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"shell": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"sql": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"yaml": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"zh": {
|
|
"count": 16,
|
|
"exact": 1.0,
|
|
"curly_count_match": 1.0
|
|
}
|
|
}
|
|
},
|
|
"eval_base_eval.json": {
|
|
"label": "base",
|
|
"split": "eval",
|
|
"count": 256,
|
|
"exact": 0.0,
|
|
"protected_preservation": 0.9868421052631579,
|
|
"curly_count_match": 0.67578125,
|
|
"by_kind": {
|
|
"comment": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"english": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 0.75
|
|
},
|
|
"go": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"java": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"javascript": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"json": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 0.0
|
|
},
|
|
"markdown": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"mixed": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 0.0
|
|
},
|
|
"nested": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 0.0
|
|
},
|
|
"python": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"quote": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 0.0625
|
|
},
|
|
"rust": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"shell": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"sql": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"yaml": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 1.0
|
|
},
|
|
"zh": {
|
|
"count": 16,
|
|
"exact": 0.0,
|
|
"curly_count_match": 0.0
|
|
}
|
|
}
|
|
}
|
|
},
|
|
"quality_audit": "validation/quality_audit.json"
|
|
} |