pluto-1 / eval_summary.json
henryphilf's picture
Pluto-1 private release candidate: dual-encoder ensemble, calibrated Typed Decisions specialist
2778af7 verified
Raw History Blame Contribute Delete
1.96 kB
{
"model": "artifacts/pluto-typed-ensemble/config.json",
"benchmark": "LocalLLaMA/typed-decisions",
"split": "test",
"mode": "fitted",
"n_cases": 400,
"n_decisions": 2000,
"accuracy": 0.7515,
"kl_from_gold": 0.11825331703553803,
"brier": 0.06491284081552073,
"ece": 0.14000920995022426,
"by_kind": {
"choice": {
"n_decisions": 600,
"accuracy": 0.7166666666666667,
"kl_from_gold": 0.1750736015750848,
"brier": 0.09019795780750824,
"ece": 0.14079999397409826
},
"noul": {
"n_decisions": 600,
"accuracy": 0.8416666666666667,
"kl_from_gold": 0.04831293153979232,
"brier": 0.03659910507461351,
"ece": 0.08494916795437828
},
"score": {
"n_decisions": 800,
"accuracy": 0.71,
"kl_from_gold": 0.12809339275268725,
"brier": 0.06718430487721054,
"ece": 0.1819052732641962
}
},
"by_workflow": {
"agent_trace_observability": {
"n_decisions": 500,
"accuracy": 0.7,
"kl_from_gold": 0.09569835841938794,
"brier": 0.05901824673184943,
"ece": 0.12393117519401484
},
"customer_service": {
"n_decisions": 500,
"accuracy": 0.754,
"kl_from_gold": 0.17659215574016174,
"brier": 0.09179313928678065,
"ece": 0.1660653832562083
},
"invoice_processing": {
"n_decisions": 500,
"accuracy": 0.834,
"kl_from_gold": 0.11278324721085041,
"brier": 0.06002676292418177,
"ece": 0.12675089177637838
},
"security_incidents": {
"n_decisions": 500,
"accuracy": 0.718,
"kl_from_gold": 0.0879395067717521,
"brier": 0.04881321431927111,
"ece": 0.14328938957429524
}
},
"latency_ms_per_case_amortized": 66.0180790000004,
"latency_ms_single_p50": 54.230400000051304,
"latency_ms_single_p95": 66.72152500023005,
"params_m": 332.84
}