Code Generation on HumanEval (pass@2, pass@4, pass@8)
59.15Pass@2DLE (top-p+top-k)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DLE (top-p+top-k)Model=Llama3.2-3B-Instruct, Sampling method=top-p+top-k2026.04 | 59.15 | 67.07 | 76.83 | |
| DLE (min-p)Model=Llama3.2-3B-Instruct, Sampling method=min-p2026.04 | 59.15 | 68.29 | 76.83 | |
| DLE (ε-sampling)-PROBFIRSTModel=Llama3.2-3B-Instruct, Sampling method=ε-sampling, Branching strategy=PROBFIRST2026.04 | 59.15 | 69.51 | 78.66 | |
| DLE (ε-sampling)-RANDBRANCHModel=Llama3.2-3B-Instruct, Sampling method=ε-sampling, Branching strategy=RANDBRANCH2026.04 | 57.93 | 65.24 | 73.17 | |
| Self-consistency (ε-sampling)Model=Llama3.2-3B-Instruct, Sampling method=ε-sampling2026.04 | 55.49 | 64.46 | 75.61 | |
| DLE (ε-sampling)-DIVFIRSTModel=Llama3.2-3B-Instruct, Sampling method=ε-sampling, Branching strategy=DIVFIRST2026.04 | 55.49 | 65.24 | 71.95 | |
| Self-consistency (min-p)Model=Llama3.2-3B-Instruct, Sampling method=min-p2026.04 | 51.83 | 65.85 | 76.83 | |
| Self-consistency (top-p+top-k)Model=Llama3.2-3B-Instruct, Sampling method=top-p+top-k2026.04 | 49.39 | 64.02 | 73.78 | |
| Self-consistencyModel=Llama3.2-3B-Instruct2026.04 | 46.34 | 57.93 | 70.12 |