Code Generation on Django
75.4WS Prefix MatchDLLM-JEPA
Evaluation results
| Method | Links | |
|---|---|---|
| DLLM-JEPABackbone=LLaDA-8B, Evaluation Protocol=4-shot, lambda=task-optimal2026.05 | 75.4 | |
| BaselineBackbone=LLaDA-8B, Evaluation Protocol=4-shot, lambda=02026.05 | 74.4 | |
| DLLM-JEPABackbone=Dream-7B, Evaluation Protocol=4-shot, lambda=task-optimal2026.05 | 72.35 | |
| BaselineBackbone=Dream-7B, Evaluation Protocol=4-shot, lambda=02026.05 | 69.58 |