Code Generation on LiveCodeBench v6 (Pass Ratio)
52.2Pass RatioREASONING ARENA-Live
Evaluation results
| Method | Links | |
|---|---|---|
| REASONING ARENA-Livetype=adaptive routing2026.06 | 52.2 | |
| REASONING ARENAtype=adaptive routing2026.06 | 51.6 | |
| RLAIFtype=judge-only2026.06 | 50.7 | |
| ArenaRLtype=judge-only2026.06 | 50.4 | |
| Adaptive Pointwisetype=adaptive routing2026.06 | 48.5 | |
| RLVRtype=verifier-only RL2026.06 | 46.7 |