Watch
1
0
Fork
You've already forked Seyyed_arc
0
forked from hesabix/arc
Seyyed_arc/docs/ai-agent/evals/phase-8/latest.json
2026-08-21 00:15:57 +00:00

218 lines
No EOL
5 KiB
JSON

{
"dataset_version": "2026-08-19.v1",
"target_recall": 0.95,
"fixed": {
"k10": {
"queries": 443,
"recall": 0.9549,
"mrr": 0.7856,
"ndcg": 0.8069,
"precision": 0.2164,
"avg_k": 9.43,
"avg_initial_schema_tokens": 2605,
"avg_schema_tokens_per_request": 13958,
"k": 10,
"mode": "fixed_10"
},
"k15": {
"queries": 443,
"recall": 0.9752,
"mrr": 0.787,
"ndcg": 0.8141,
"precision": 0.1706,
"avg_k": 13.51,
"avg_initial_schema_tokens": 3706,
"avg_schema_tokens_per_request": 19931,
"k": 15,
"mode": "fixed_15"
},
"k20": {
"queries": 443,
"recall": 0.9819,
"mrr": 0.7873,
"ndcg": 0.8167,
"precision": 0.1496,
"avg_k": 17.05,
"avg_initial_schema_tokens": 4621,
"avg_schema_tokens_per_request": 24837,
"k": 20,
"mode": "fixed_20"
}
},
"adaptive": [
{
"queries": 443,
"recall": 0.9707,
"mrr": 0.7866,
"ndcg": 0.8099,
"precision": 0.1999,
"avg_k": 11.23,
"avg_initial_schema_tokens": 3112,
"avg_schema_tokens_per_request": 16776,
"mode": "adaptive_gap_v1",
"policy": "gap_v1",
"levels": {
"high": 308,
"low": 87,
"normal": 37,
"none": 10,
"fail": 1
},
"recall_ok": true
},
{
"queries": 443,
"recall": 0.9661,
"mrr": 0.7863,
"ndcg": 0.8086,
"precision": 0.208,
"avg_k": 10.19,
"avg_initial_schema_tokens": 2841,
"avg_schema_tokens_per_request": 15330,
"mode": "adaptive_tight",
"policy": "tight",
"levels": {
"high": 308,
"low": 87,
"normal": 37,
"none": 10,
"fail": 1
},
"recall_ok": true
},
{
"queries": 443,
"recall": 0.9707,
"mrr": 0.7866,
"ndcg": 0.8101,
"precision": 0.198,
"avg_k": 11.72,
"avg_initial_schema_tokens": 3235,
"avg_schema_tokens_per_request": 17441,
"mode": "adaptive_wide",
"policy": "wide",
"levels": {
"high": 308,
"low": 87,
"normal": 37,
"none": 10,
"fail": 1
},
"recall_ok": true
}
],
"best_adaptive": {
"queries": 443,
"recall": 0.9661,
"mrr": 0.7863,
"ndcg": 0.8086,
"precision": 0.208,
"avg_k": 10.19,
"avg_initial_schema_tokens": 2841,
"avg_schema_tokens_per_request": 15330,
"mode": "adaptive_tight",
"policy": "tight",
"levels": {
"high": 308,
"low": 87,
"normal": 37,
"none": 10,
"fail": 1
},
"recall_ok": true
},
"hybrid": [
{
"queries": 443,
"recall": 0.9458,
"mrr": 0.7845,
"ndcg": 0.8023,
"precision": 0.1808,
"avg_k": 10.0,
"avg_initial_schema_tokens": 2754,
"avg_schema_tokens_per_request": 14462,
"k": 10,
"weights": "lexical_dominant",
"delete_family_false_positives": 1,
"delete_family_queries": 20
},
{
"queries": 443,
"recall": 0.9639,
"mrr": 0.7858,
"ndcg": 0.8091,
"precision": 0.1251,
"avg_k": 15.0,
"avg_initial_schema_tokens": 4025,
"avg_schema_tokens_per_request": 20944,
"k": 15,
"weights": "lexical_dominant",
"delete_family_false_positives": 1,
"delete_family_queries": 20
},
{
"queries": 443,
"recall": 0.9458,
"mrr": 0.7847,
"ndcg": 0.7989,
"precision": 0.1779,
"avg_k": 10.0,
"avg_initial_schema_tokens": 2739,
"avg_schema_tokens_per_request": 14325,
"k": 10,
"weights": "balanced",
"delete_family_false_positives": 1,
"delete_family_queries": 20
},
{
"queries": 443,
"recall": 0.9616,
"mrr": 0.7857,
"ndcg": 0.8068,
"precision": 0.124,
"avg_k": 15.0,
"avg_initial_schema_tokens": 4013,
"avg_schema_tokens_per_request": 20872,
"k": 15,
"weights": "balanced",
"delete_family_false_positives": 1,
"delete_family_queries": 20
},
{
"queries": 443,
"recall": 0.9413,
"mrr": 0.7838,
"ndcg": 0.7956,
"precision": 0.1758,
"avg_k": 10.0,
"avg_initial_schema_tokens": 2722,
"avg_schema_tokens_per_request": 14221,
"k": 10,
"weights": "semantic_dominant",
"delete_family_false_positives": 1,
"delete_family_queries": 20
},
{
"queries": 443,
"recall": 0.9571,
"mrr": 0.7853,
"ndcg": 0.8043,
"precision": 0.1228,
"avg_k": 15.0,
"avg_initial_schema_tokens": 3956,
"avg_schema_tokens_per_request": 20556,
"k": 15,
"weights": "semantic_dominant",
"delete_family_false_positives": 1,
"delete_family_queries": 20
}
],
"best_hybrid": {
"weights": "lexical_dominant",
"k": 15,
"recall": 0.9639,
"mrr": 0.7858,
"ndcg": 0.8091
},
"hybrid_vs_fixed_k10": null
}