UT Austin researchers introduced LazyWeights, using lossless entropy coding to cut edge LLM memory by up to 31% and accelerate on-device token generation.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results