Intelligent Tutoring on Intelligent Tutoring Environment Full-task
100Success RateGemini-Flash Oracle
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-Flash OracleObs.=GT†, n=302026.06 | 100 | 8.4 | 100 | |
| GPT-5-mini OracleObs.=GT†, n=302026.06 | 100 | 3.4 | 100 | |
| PPOObs.=Binary, n=3002026.06 | 93.3 | 72.4 | 98.8 | |
| PPO + TutorRLObs.=NL, n=502026.06 | 82 | 105.6 | 98 | |
| PPO + Qwen2.5Obs.=NL, n=502026.06 | 78 | 120.1 | 95.4 | |
| RandomObs.=Binary, n=3002026.06 | 28.3 | 217.4 | 90.6 | |
| Gemini-FlashObs.=NL, n=302026.06 | 23.3 | 223.3 | 83.9 | |
| GPT-5-miniObs.=NL, n=302026.06 | 3.3 | 234.2 | 77.1 | |
| TutorRL OracleObs.=GT†, n=3002026.06 | 0 | 236.2 | 32.6 | |
| TutorRLObs.=NL, n=3002026.06 | 0 | 236.2 | 30.5 | |
| ThinkingKT + Qwen2.5Obs.=NL, n=502026.06 | 0 | 236.2 | 48.1 | |
| ThinkingKT + TutorRLObs.=NL, n=502026.06 | 0 | 236.2 | 49.3 |