Better performance than TQ and better quality than FP16?

Am I reading this right??

• qeternity

  today at 5:04 PM

  It's not better quality: 59.3% vs 59.4% fp16 on AIME 25

• thefox96

  today at 5:02 PM

  Faster than Fp16, not better quality i guess

• pbich

  today at 4:55 PM

  [dead]

Why this is not a PR for vLLM ?

• esafak

  today at 4:00 PM

  It's the output of a research paper; the authors are not trying to build up vLLM, and they probably have no incentive to do so. You can submit a PR, though! It's easier now while the divergence is low, so don't wait. Since there are six authors, I bet you could get help with the inevitable review chores if you just take the step of creating the PR.

  edit: It might not be clear that it is based on vLLM 0.22, which is the current version: https://github.com/huawei-csl/KVarN/commit/d6290e99098d7426d.... All you have to do is create a diff off it; it's fairly straightforward.

  • jmalicki

    today at 4:14 PM

    And with the help of AI, pointing at AI at this paper and saying "making a vLLM PR from this paper" tends to work surprisingly well, even if you need to nudge it a little bit along the way.

• thefox96

  today at 5:28 PM

  it should be easy to do btw

[dead]