eval/results/generalization_grpo_v2.json

[
 {
  "system": "GRPO v2 (KL-anchored, union)",
  "sources": [
   "flights",
   "rayyan",
   "ed2_restaurants"
  ],
  "gen_f1": 0.054831358307998644,
  "variant_recall": 0.10838756027134193,
  "variant_precision": 0.3999487967229903,
  "other_recall": 0.027985877895461985,
  "damage": 0.07449272188217769,
  "per_source": [
   {
    "source": "flights",
    "f1": 0.16395283548568218,
    "precision": 0.26497277676951,
    "recall": 0.11869918699186992,
    "damage": 0.08393101092896176,
    "variant_errors": 1049,
    "variant_fixed": 259,
    "variant_changed": 465,
    "variant_good": 259,
    "other_errors": 3871,
    "other_fixed": 325
   },
   {
    "source": "rayyan",
    "f1": 0.0,
    "precision": 0.0,
    "recall": 0.0,
    "damage": 0.06356943891762833,
    "variant_errors": 171,
    "variant_fixed": 0,
    "variant_changed": 4,
    "variant_good": 0,
    "other_errors": 777,
    "other_fixed": 0
   },
   {
    "source": "ed2_restaurants",
    "f1": 0.0005412394383137384,
    "precision": 0.00027448229589191497,
    "recall": 0.019230769230769232,
    "damage": 0.07597771579994297,
    "variant_errors": 115,
    "variant_fixed": 9,
    "variant_changed": 14,
    "variant_good": 9,
    "other_errors": 353,
    "other_fixed": 0
   }
  ]
 }
]