Iterated Prisoner's Dilemma (IPD)
2.745Payoff per RoundSEPO
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SEPOBackbone=Gemma 4 E4B-it, Step=252026.05 | 2.745 | 0.703 | 15.9 | 0.772 | -0.09 | |
| BaseBackbone=Gemma 4 E4B-it2026.05 | 2.741 | 0.703 | 15.5 | 0.761 | -0.091 | |
| SFTBackbone=Gemma 4 E4B-it2026.05 | 2.719 | 0.828 | 15.3 | 0.455 | -0.105 | |
| SEPOBackbone=Qwen 3.5-4B, Step=final2026.05 | 2.379 | 1.25 | 26.1 | -1.847 | -0.08 | |
| SFTBackbone=Qwen 3.5-4B2026.05 | 1.773 | 3.125 | 27.8 | -7.894 | -0.225 | |
| BaseBackbone=Qwen 3.5-4B2026.05 | 1.641 | 5 | 14.6 | -13.109 | -0.372 |