Tutor Robustness Evaluation on Multi-Agent Student
11Student LeakageMathDial-SFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MathDial-SFTTutor Setting=Base in-context2026.04 | 11 | 5.73 | 41 | 5.96 | |
| SocraticLMTutor Setting=Base in-context2026.04 | 12 | 5.65 | 50 | 4.04 | |
| TutorRL-7BTutor Setting=Base in-context2026.04 | 25 | 5.72 | 53 | 8.43 |