Code Generation on LiveCodeBench v5 (test)
72.1Pass@1BF16 (Teacher)
Evaluation Results
| Method | Links | |
|---|---|---|
| BF16 (Teacher)sampling=8 samples per problem2026.06 | 72.1 | |
| CKA-QADsampling=8 samples per problem2026.06 | 70.3 | |
| NVFP4 PTQsampling=8 samples per problem2026.06 | 68.9 | |
| QAD (KL-only)sampling=8 samples per problem2026.06 | 68.9 | |
| CAPSModel=GPT-OSS-20B, N=16, Temperature=0.62026.05 | 61.6 | |
| V1-InferModel=GPT-OSS-20B, N=16, Temperature=0.62026.05 | 58.4 | |
| PointwiseModel=GPT-OSS-20B, N=16, Temperature=0.62026.05 | 52.8 | |
| CAPSModel=Qwen3-4B-Thinking-2507, N=16, Temperature=0.62026.05 | 51.6 | |
| VanillaModel=GPT-OSS-20B, N=16, Temperature=0.62026.05 | 49.8 | |
| CAPSModel=Qwen3-4B-Instruct-2507, N=16, Temperature=0.62026.05 | 48.7 | |
| V1-InferModel=Qwen3-4B-Thinking-2507, N=16, Temperature=0.62026.05 | 45.2 | |
| V1-InferModel=Qwen3-4B-Instruct-2507, N=16, Temperature=0.62026.05 | 42.3 | |
| PointwiseModel=Qwen3-4B-Thinking-2507, N=16, Temperature=0.62026.05 | 39.8 | |
| PointwiseModel=Qwen3-4B-Instruct-2507, N=16, Temperature=0.62026.05 | 38 | |
| CAPSModel=Qwen3-14B, N=16, Temperature=0.62026.05 | 37.6 | |
| VanillaModel=Qwen3-4B-Thinking-2507, N=16, Temperature=0.62026.05 | 36.9 | |
| VanillaModel=Qwen3-4B-Instruct-2507, N=16, Temperature=0.62026.05 | 35.3 | |
| V1-InferModel=Qwen3-14B, N=16, Temperature=0.62026.05 | 34.8 | |
| PointwiseModel=Qwen3-14B, N=16, Temperature=0.62026.05 | 29.3 | |
| VanillaModel=Qwen3-14B, N=16, Temperature=0.62026.05 | 26.2 |