Code Generation on LiveCodeBench v6 (test) (Avg@4)
21.1Avg@4AIPO
Evaluation Results
| Method | Links | |
|---|---|---|
| AIPOPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 21.1 | |
| AIPOPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen2.5-7B-Instruct2026.05 | 19.2 | |
| LUFFYPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 19.2 | |
| Dr.GRPOPolicy Model=Qwen2.5-7B-Instruct2026.05 | 18.4 | |
| GRPOPolicy Model=Qwen2.5-7B-Instruct2026.05 | 18.1 | |
| OPSDPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 18.1 | |
| LUFFYPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen2.5-7B-Instruct2026.05 | 17.3 | |
| OPSDPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen2.5-7B-Instruct2026.05 | 17 | |
| SFTPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 16.4 | |
| PRIMEPolicy Model=Qwen2.5-7B-Instruct2026.05 | 16.3 | |
| SFTPolicy Model=Qwen2.5-7B-Instruct, Collaborator Model=Qwen2.5-7B-Instruct2026.05 | 15.9 | |
| OriginalPolicy Model=Qwen2.5-7B-Instruct2026.05 | 15.7 | |
| AIPOPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 12.4 | |
| LUFFYPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 10.8 | |
| AIPOPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Llama3.2-3B-Instruct2026.05 | 10.1 | |
| OPSDPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 9.6 | |
| LUFFYPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Llama3.2-3B-Instruct2026.05 | 8.7 | |
| SFTPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Qwen3-30B-A3B-Instruct-25072026.05 | 8 | |
| OPSDPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Llama3.2-3B-Instruct2026.05 | 7.5 | |
| SFTPolicy Model=Llama3.2-3B-Instruct, Collaborator Model=Llama3.2-3B-Instruct2026.05 | 6.2 |