Competitive Coding on OJBench
45.9Best@8 ScoreAgentic Verifier
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Agentic VerifierPolicy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 45.9 | 48.6 | |
| Random GeneratorPolicy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 43.8 | 44.2 | |
| CodeTPolicy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 41 | 40.6 | |
| MBR-Exec (soft)Policy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 40.5 | 41.4 | |
| CodeRMPolicy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 40.4 | 39.5 | |
| MBR-Exec (hard)Policy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 37.7 | 39.6 | |
| Grading RMPolicy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 37.6 | 34.9 | |
| VanillaPolicy Model=Qwen3-235B-A22B-Thinking-25072026.02 | 37.1 | 37.1 | |
| Agentic VerifierPolicy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 33.5 | 34.9 | |
| Random GeneratorPolicy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 33.4 | 34.1 | |
| CodeRMPolicy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 32 | 31.3 | |
| CodeTPolicy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 31.6 | 32.1 | |
| MBR-Exec (soft)Policy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 31.4 | 31.8 | |
| MBR-Exec (hard)Policy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 29.6 | 32.4 | |
| VanillaPolicy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 28.9 | 28.9 | |
| Grading RMPolicy Model=Qwen3-30B-A3B-Thinking-25072026.02 | 28.7 | 29.3 |