The story’s score and claims

Comments on No. 46 of 137 · leapquant.github.io · edition of Friday 13 November 2026

LeapQuant: Efficient Linear Attention with Accurate Recurrent State Quantization

Summary

LeapQuant, a paper from researchers at UC Berkeley, the University of Washington, MIT, Perplexity AI, and NVIDIA, proposes a training-free way to quantize the recurrent state of linear attention layers to 8 bits.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules