ReFusion: A Diffusion Large Language Model with Parallel Autoregressive Decoding

Benchmark Model Rank Results
code-generation-on-mbppReFusion#63Accuracy: 68.2
code-generation-on-humanevalReFusion#67Pass@1: 78.66
math-word-problem-solving-on-mathReFusion#70Accuracy: 54.22
arithmetic-reasoning-on-gsm8kReFusion#111Accuracy: 84.91
multi-task-language-understanding-on-mmlu-proReFusion#121Accuracy: 45.94