Dream 7B: Diffusion Large Language Models

Benchmark Model Rank Results
code-generation-on-humanevalDream 7B–Pass@1: 57.9
common-sense-reasoning-on-winograndeDream 7B (5-shot)–Accuracy: 74.5
multi-task-language-understanding-on-mmluDream 7B (5-shot)–Average (%): 69.5
multi-task-language-understanding-on-mmluDream-Instruct 7B (4-shot)–Average (%): 67.0
multiple-choice-question-answering-mcqa-on-arc-cDream 7B–Accuracy: 59.8
multiple-choice-question-answering-mcqa-on-arc-eDream 7B–Accuracy: 83.9
question-answering-on-piqaDream 7B–Accuracy: 75.8
sentence-completion-on-hellaswagDream 7B–Accuracy: 73.3