New in @berkeley_ai blog. Cao et al. create new kernel optimizer K-search and use it to auto adapt CUDA to MLX. Their attention translation measures 0.97x speed to Apple’s native implementation. Translate your code with their OSS tool. Available now.
bair.berkeley.edu/blog/2026/07/2… 🐻📄
bair.berkeley.edu/blog/2026/07/2… 🐻📄
1
2
9
3,722
New on @berkeley_ai blog. What actually affects the output of LLMs? New work by Butler et al. (NeurIPS 2025) introduces a model-agnostic method of measuring input, data, and component attribution, giving insight on what really steers generation.
bair.berkeley.edu/blog/2026/03/1… 🐻📄
bair.berkeley.edu/blog/2026/03/1… 🐻📄
3
12
41
9,814
addison retweeted
Why and how do diffusion models memorize vs generalize? Can we have scaling laws for memorization? This is increasingly relevant scientifically and pragmatically (e.g. Sora 2).
🚨 Our new preprint "On the Edge of Memorization in Diffusion Models" addresses this timely question!
🚨 Our new preprint "On the Edge of Memorization in Diffusion Models" addresses this timely question!
5
66
369
30,565
addison retweeted
Now we can have self-refinement for VLAs in the real world (with the aid of a big VLM)! VLM critiques VLA rollouts and iteratively refines the commands to make it perform better.
LLMs have shown a remarkable ability to “self-refine” and learn from their mistakes via in-context learning. But in robotics, most methods are single-shot. How can we bring inference-time adaptation to robot learning? A 🧵:
5
22
242
27,146