ByteDance Releases 8B Diffusion Model iLLaDA Matching Qwen2.5

Researchers from Renmin University and ByteDance introduced iLLaDA, an 8B parameter language model that uses diffusion for text generation instead of standard autoregressive methods. The base model reaches performance parity with Qwen2.5, though it lags after fine-tuning. The release demonstrates continued experimentation with alternative architectures for large language models.