Multi-Block Diffusion Language Models

Chat with MBD-Math-SDAR-8B-Chat-b32, a Multi-Block Diffusion Language Model from SJTU-DENG-Lab. This model generates text via iterative block diffusion (non-autoregressive) instead of standard token-by-token generation. Trained on math reasoning data.

📄 Paper: Multi-Block Diffusion Language Models 🐙 GitHub: SJTU-DENG-Lab/mbd-lms 🤗 Model: SJTU-DENG-Lab/MBD-Math-SDAR-8B-Chat-b32

0 1
1 16
8 64