tracked progres report and data folder also and added correct evaluation results, Add graph-seeded multi-hop QA generator and --qa-file eval; document in README
784726a | [ | |
| { | |
| "model": "claude-sonnet-4-6", | |
| "provider": "anthropic", | |
| "books": [ | |
| "finkbeiner", | |
| "guinee", | |
| "hauschild", | |
| "ilcd" | |
| ], | |
| "n": 770, | |
| "no_rag": { | |
| "bertscore_recall": 0.8795992732048035, | |
| "cosine": 0.6885953807205166, | |
| "rouge_l": 0.5283131673537708 | |
| }, | |
| "rag": { | |
| "bertscore_recall": 0.9214715361595154, | |
| "cosine": 0.7446262060565944, | |
| "rouge_l": 0.7603977655346544 | |
| }, | |
| "rag_graph": { | |
| "bertscore_recall": 0.9182685613632202, | |
| "cosine": 0.7386851505928546, | |
| "rouge_l": 0.7851236931112503 | |
| }, | |
| "target_bertscore": 0.8 | |
| }, | |
| { | |
| "model": "gpt-5.4", | |
| "provider": "openai", | |
| "books": [ | |
| "finkbeiner", | |
| "guinee", | |
| "hauschild", | |
| "ilcd" | |
| ], | |
| "n": 770, | |
| "no_rag": { | |
| "bertscore_recall": 0.8864094018936157, | |
| "cosine": 0.7055306658812076, | |
| "rouge_l": 0.49230725341137455 | |
| }, | |
| "rag": { | |
| "bertscore_recall": 0.9262104630470276, | |
| "cosine": 0.7631553358339636, | |
| "rouge_l": 0.6898600680079529 | |
| }, | |
| "rag_graph": { | |
| "bertscore_recall": 0.9236271977424622, | |
| "cosine": 0.7551015681341352, | |
| "rouge_l": 0.7226493422081696 | |
| }, | |
| "target_bertscore": 0.8 | |
| } | |
| ] |