diffusionvl-ar-to-diffusion

【数据与研究】Convert pre-trained autoregressive vision-language models into diffusion VLMs without architectural modifications. Use block diffusion strategy enabling arbitrary-length generation and KV-cache reuse. Hybrid attention enforces bidirectional within blocks, causal between blocks. Requires less than 5% of data compared to prior diffusion VLM methods.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准