Code Generation on LiveCodeBench (Avg@16)
32.9Avg@16TRD
Evaluation Results
| Method | Links | |
|---|---|---|
| TRDModel=Qwen3-1.7B (w/ thinking), Traj.=yr2026.06 | 32.9 | |
| Forward KLModel=Qwen3-1.7B (w/ thinking), Traj.=yo2026.06 | 32.8 | |
| Reverse KL w/ Top-KModel=Qwen3-1.7B (w/ thinking), Traj.=yo2026.06 | 32.6 | |
| Forward KL w/ ClipModel=Qwen3-1.7B (w/ thinking), Traj.=yo2026.06 | 32.4 | |
| BaseModel=Qwen3-1.7B (w/ thinking), Traj.=–2026.06 | 32.3 | |
| Reverse KLModel=Qwen3-4B-Instruct-2507, Traj.=yo2026.06 | 32.2 | |
| BaseModel=Qwen3-4B-Instruct-2507, Traj.=–2026.06 | 32.1 | |
| Forward KL w/ ClipModel=Qwen3-4B-Instruct-2507, Traj.=yo2026.06 | 32 | |
| Reverse KL w/ Top-KModel=Qwen3-4B-Instruct-2507, Traj.=yo2026.06 | 32 | |
| Reverse KLModel=Qwen3-1.7B (w/ thinking), Traj.=yo2026.06 | 31.9 | |
| TRDModel=Qwen3-4B-Instruct-2507, Traj.=yr2026.06 | 31.9 | |
| Forward KLModel=Qwen3-4B-Instruct-2507, Traj.=yo2026.06 | 31.5 |