Files
yao-meta-skill/evals/history/description_optimization/2026-03-31-description-optimization-suite.json
T

46 lines
1.4 KiB
JSON

{
"snapshot_id": "description-optimization-suite",
"date": "2026-03-31",
"commit": "e78602c",
"label": "Description Optimization Suite",
"targets": [
{
"name": "yao-meta-skill",
"winner_label": "Current",
"winner_tokens": 65,
"winner_holdout_fp": 0,
"winner_holdout_fn": 0,
"winner_blind_holdout_fp": null,
"winner_blind_holdout_fn": null,
"winner_blind_holdout_total_errors": null,
"drift_note": "initial description optimization suite without blind holdout"
},
{
"name": "team-frontend-review",
"winner_label": "Current",
"winner_tokens": 50,
"winner_holdout_fp": 0,
"winner_holdout_fn": 0,
"winner_blind_holdout_fp": null,
"winner_blind_holdout_fn": null,
"winner_blind_holdout_total_errors": null,
"drift_note": "compressed example description and stored first holdout report"
},
{
"name": "governed-incident-command",
"winner_label": "Current",
"winner_tokens": 37,
"winner_holdout_fp": 0,
"winner_holdout_fn": 1,
"winner_blind_holdout_fp": null,
"winner_blind_holdout_fn": null,
"winner_blind_holdout_total_errors": null,
"drift_note": "compressed governed description and reduced visible holdout misses"
}
],
"notes": [
"added dev and holdout description optimization reports",
"connected description optimization to make test"
]
}