Game Theory Agent Evaluation on GTBench (test)
53.3TS ScoreMAFP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MAFPRound type=Multiple Rounds, Base model=Q3.5-35B2026.06 | 53.3 | 42.1 | |
| ToMRound type=Multiple Rounds, Base model=Q3.5-35B2026.06 | 52.7 | 36.9 | |
| Q3.5-35BRound type=Single Round2026.06 | 50 | 35.5 | |
| Q3-1.7BRound type=Single Round2026.06 | 49.6 | 35.9 | |
| DebateRound type=Multiple Rounds, Base model=Q3.5-35B2026.06 | 49.6 | 36.3 | |
| MAFP-LastRound type=Multiple Rounds, Base model=Q3.5-35B2026.06 | 49.1 | 37.4 | |
| SRRound type=Multiple Rounds, Base model=Q3.5-35B2026.06 | 47.7 | 34.8 | |
| Llama-3.1-8BRound type=Single Round2026.06 | 45.2 | 31.6 | |
| GPT-5-nanoRound type=Single Round2026.06 | 42 | 30.9 |