Code Generation on LiveCodeBench v6 (Pass@16)
55.2Pass@16Base
Evaluation Results
| Method | Links | |
|---|---|---|
| BaseModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=-, Teacher Model=Qwen3-8B2026.06 | 55.2 | |
| Reverse KLModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 55.1 | |
| Reverse KL w/ Top-KModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 54.5 | |
| Forward KL w/ ClipModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 54.1 | |
| TRDModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yr, Teacher Model=Qwen3-8B2026.06 | 54.1 | |
| Forward KLModel Backbone=Qwen3-4B-Instruct-2507, Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 53.3 | |
| BaseModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=-, Teacher Model=Qwen3-8B2026.06 | 48.5 | |
| Forward KL w/ ClipModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 47.2 | |
| TRDModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yr, Teacher Model=Qwen3-8B2026.06 | 46.8 | |
| Reverse KL w/ Top-KModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 46.6 | |
| Reverse KLModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 46.3 | |
| Forward KLModel Backbone=Qwen3-1.7B (w/ thinking), Trajectory Type (Traj.)=yo, Teacher Model=Qwen3-8B2026.06 | 41.4 |