{ "package": "qwen3_5_4b_natural_memory_v2_1", "eval_file": "data/realistic_v2/eval.jsonl", "episodes": 24, "per_category": { "alias_paraphrase": { "episodes": 3, "candidates_written": 15, "active_after_write": 15, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 33.33, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 }, "long_fact": { "episodes": 3, "candidates_written": 12, "active_after_write": 11, "write_survival_pct": 91.67, "candidates_parsed_as_attribute_pct": 0.0, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 }, "multi_entity": { "episodes": 3, "candidates_written": 9, "active_after_write": 9, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 0.0, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 }, "multi_hop": { "episodes": 3, "candidates_written": 15, "active_after_write": 15, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 0.0, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 }, "near_miss": { "episodes": 3, "candidates_written": 12, "active_after_write": 12, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 75.0, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 }, "noise_context": { "episodes": 3, "candidates_written": 27, "active_after_write": 27, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 11.11, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 }, "unknown_attribute": { "episodes": 3, "candidates_written": 15, "active_after_write": 15, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 100.0, "answerable_episodes": 0, "target_survived": 0, "target_survival_pct": 0.0 }, "update_conflict": { "episodes": 3, "candidates_written": 15, "active_after_write": 6, "write_survival_pct": 40.0, "candidates_parsed_as_attribute_pct": 80.0, "answerable_episodes": 3, "target_survived": 3, "target_survival_pct": 100.0 } }, "target_missing_examples": [], "overall": { "candidates_written": 120, "active_after_write": 110, "write_survival_pct": 91.67, "answerable_episodes": 21, "target_survival_pct": 100.0 } }