From AR to Diffusion: Efficiently Adapting Large Language Models with Strictly Causal and Elastic Horizons

In three linesFLUID efficiently adapts autoregressive (AR) language models to diffusion-based generation through strictly causal alignment and elastic horizons. The framework reduces training costs by orders of magnitude by reusing existing GPT checkpoints while maintaining state-of-the-art performance.

Summary generated by Claude — human-verified