Theory of Mind reasoning on MMToM
83.3AccuracyThinking-RFT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Thinking-RFTModality=Lan.+Vis.2026.06 | 83.3 | 82.2 | |
| Thinking-RFTModality=Lan.2026.06 | 78.5 | — | |
| SFTModality=Lan.+Vis.2026.06 | 74.3 | 74.75 | |
| SFTModality=Lan.2026.06 | 73.02 | — | |
| AutoToMModality=Lan.+Vis.2026.06 | 56.9 | 58 | |
| SimToMModality=Lan.+Vis.2026.06 | 50.6 | 50.1 | |
| Zero-shotModality=Lan.+Vis.2026.06 | 45 | 44.15 | |
| Zero-shotModality=Lan.2026.06 | 39.4 | — |