Tutor Leakage Evaluation on MathDial fine-tuned adversarial student setting (test)
8Tutor LeakageMulti-Agent Tutor
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Multi-Agent TutorTutor Type=Multi-Agent Tutor, Tutor Model=Qwen2.5-72B-Instruct, Adversarial Student Setting=fine-tuned2026.04 | 8 | 3 | |
| Tutor w/ ReasoningTutor Type=Tutor w/ Reasoning, Tutor Model=Qwen2.5-72B-Instruct, Adversarial Student Setting=fine-tuned2026.04 | 8 | 2 | |
| Base In-Context TutorTutor Type=Base In-Context Tutor, Tutor Model=Qwen2.5-72B-Instruct, Adversarial Student Setting=fine-tuned2026.04 | 80 | 1 |