| { | |
| "do_not_pr_as_full_board": true, | |
| "note": "Budget slice (n=1 scored). Schema matches ContextBench/contextbench.github.io src/data/agent_results.json. Not comparable to 1,136-task Pass@1.", | |
| "source": "https://huggingface.co/datasets/pragnyanramtha/splice-contextbench", | |
| "row": { | |
| "model": "Splice + DeepSeek-V4-Pro-0813", | |
| "performance": { | |
| "file": { | |
| "recall": 1.0, | |
| "precision": 0.667, | |
| "f1": 0.8 | |
| }, | |
| "block": { | |
| "recall": 0.946, | |
| "precision": 0.127, | |
| "f1": 0.224 | |
| }, | |
| "line": { | |
| "recall": 0.943, | |
| "precision": 0.125, | |
| "f1": 0.221 | |
| }, | |
| "pass_at_1": 1.0 | |
| }, | |
| "patterns": { | |
| "avg_steps_per_instance": 40, | |
| "avg_cost_per_instance": 0.25 | |
| }, | |
| "n_scored": 2, | |
| "instances": [ | |
| "scikit-learn__scikit-learn-25232", | |
| "django__django-16263" | |
| ], | |
| "django_note": "RuntimeError at $2 cap; no PATCH_CONTEXT; Pass@1 no" | |
| } | |
| } | |