This was previously reported 5 months ago: <a href="https://news.ycombinator.com/item?id=42415122">https://news.ycombinator.com/item?id=42415122</a> (84 comments).<p>As an aside - I am a big fan of Luke Zettlemoyer and his team at the University of Washington. They've been doing cool NLP research for years!
Great to see our paper here again! Since the paper release, we've also released model weights here for anyone interesting in building on top of it: <a href="https://huggingface.co/facebook/blt" rel="nofollow">https://huggingface.co/facebook/blt</a>. We also added HF Hub code to easily load the model <a href="https://github.com/facebookresearch/blt?tab=readme-ov-file#load-weights-via-hf-hub">https://github.com/facebookresearch/blt?tab=readme-ov-file#l...</a>.
This BLT approach is why "AI research is stalling" takes are wrong. Dynamic byte-level patches instead of tokens seems genuinely innovative, not just scaling up the same architecture. Better efficiency AND handling edge cases better? Actual progress. The field is still finding clever ways to rethink fundamentals.