Knowledge-Gap Localization on Overcooked
100Loc RecallRandom-Loc
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Random-Loc2026.06 | 100 | 4 | 7.7 | 92.5 | 0 | 92.5 | 3.6 | 6.9 | 100 | |
| Random-H-Loc2026.06 | 100 | 33.3 | 50 | 92.5 | 0 | 92.5 | 30.8 | 46.2 | 8.2 | |
| Oracle-Loctype=Upper2026.06 | 100 | 100 | 100 | 92.4 | 0 | 92.4 | 92.4 | 92.4 | 0 | |
| SENSEI2026.06 | 82.8 | 64.6 | 72.6 | 100 | 0 | 82.8 | 64.6 | 72.6 | 1.9 | |
| Llama-3-70B2026.06 | 33.5 | 17.9 | 23.3 | 10.4 | 30 | 3.5 | 1.9 | 2.5 | 3.8 | |
| GPT-3.52026.06 | 27.5 | 22.4 | 24.7 | 30.9 | 22 | 8.5 | 6.9 | 7.6 | 2.6 | |
| GPT-4o2026.06 | 18.5 | 19.6 | 19 | 2.7 | 53.9 | 0.5 | 0.5 | 0.5 | 1.2 | |
| End-to-End2026.06 | 4.2 | 2.7 | 3.3 | 85.3 | 0 | 3.6 | 2.3 | 2.8 | 6.1 |