Large Language Diffusion Models

Benchmark Model Rank Results
multiple-choice-question-answering-mcqa-on-arc-cLLaDA 8B Instruct#1Accuracy: 88.5
multiple-choice-question-answering-mcqa-on-arc-cLLaDA 8B Base#8Accuracy: 45.9
multi-task-language-understanding-on-c-evalLLaDA 8B Base (5-shot)#22Accuracy: 70.5
multi-task-language-understanding-on-mmluLLaDA 8B Base (5-shot)#38Average (%): 65.9
multi-task-language-understanding-on-mmluLLaDA 8B Instruct (5-shot)#39Average (%): 65.5
common-sense-reasoning-on-winograndeLLaDA 8B Base (5-shot)#57Accuracy: 74.8
sentence-completion-on-hellaswagLLaDA 8B Instruct#96Accuracy: 74.6
sentence-completion-on-hellaswagLLaDA 8B Base#107Accuracy: 70.5
code-generation-on-humanevalLLaDA 8B Instruct#102Pass@1: 49.4
question-answering-on-piqaLLaDA 8B Base#118Accuracy: 73.6