Code Generation on LiveCodeBench 8/1/24-2/1/25 (test)
61.2Pass@1Sync.AReaL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Sync.AReaLModel Size=32B, # Nodes=48, PPO Steps=602025.05 | 61.2 | 51.1 | |
| AReaLModel Size=32B, # Nodes=48, PPO Steps=602025.05 | 61 | 31.1 | |
| AReaLModel Size=14B, # Nodes=32, PPO Steps=802025.05 | 58.1 | 21.9 | |
| VeRLModel Size=14B, # Nodes=32, PPO Steps=802025.05 | 57.9 | 44.4 | |
| BasemodelModel Size=32B2025.05 | 57.4 | — | |
| Sync.AReaLModel Size=14B, # Nodes=32, PPO Steps=802025.05 | 56.7 | 48.8 | |
| BasemodelModel Size=14B2025.05 | 53.4 | — | |
| VeRLModel Size=32B, # Nodes=48, PPO Steps=602025.05 | — | 46.4 |