{ "package": "qwen3_5_4b_natural_memory_v2_1", "eval_file": "data/realistic_v2/eval.jsonl", "episodes": 40, "per_category": { "alias_paraphrase": { "episodes": 5, "candidates_written": 25, "active_after_write": 25, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 40.0, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 }, "long_fact": { "episodes": 5, "candidates_written": 20, "active_after_write": 19, "write_survival_pct": 95.0, "candidates_parsed_as_attribute_pct": 0.0, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 }, "multi_entity": { "episodes": 5, "candidates_written": 15, "active_after_write": 15, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 0.0, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 }, "multi_hop": { "episodes": 5, "candidates_written": 25, "active_after_write": 25, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 0.0, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 }, "near_miss": { "episodes": 5, "candidates_written": 20, "active_after_write": 20, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 65.0, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 }, "noise_context": { "episodes": 5, "candidates_written": 45, "active_after_write": 45, "write_survival_pct": 100.0, "candidates_parsed_as_attribute_pct": 11.11, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 }, "unknown_attribute": { "episodes": 5, "candidates_written": 25, "active_after_write": 25, "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": 5, "candidates_written": 25, "active_after_write": 10, "write_survival_pct": 40.0, "candidates_parsed_as_attribute_pct": 80.0, "answerable_episodes": 5, "target_survived": 5, "target_survival_pct": 100.0 } }, "target_missing_examples": [], "overall": { "candidates_written": 200, "active_after_write": 184, "write_survival_pct": 92.0, "answerable_episodes": 35, "target_survival_pct": 100.0 } }