Preference Adaptation on CALCONFLICTBENCH (test)
0.12AERPEARL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PEARLbase_model=Qwen3-4B, decision_rounds=1042026.01 | 0.12 | 0.761 | |
| SFTbase_model=Qwen3-4B, decision_rounds=104, protocol=Supervised Fine-Tuning2026.01 | 0.27 | 0.325 | |
| Zero-shot + StrategyHubbase_model=Qwen3-4B, decision_rounds=104, protocol=Zero-shot with external Strategy Hub2026.01 | 0.41 | 0.048 | |
| Zero-shotbase_model=Qwen3-4B, decision_rounds=104, protocol=Zero-shot2026.01 | 0.45 | -0.029 |