alt.hn

2/20/2026 at 4:46:37 AM

Fast KV Compaction via Attention Matching

https://arxiv.org/abs/2602.16284

by cbracketdash

2/20/2026 at 2:46:44 PM

Considering the insanity of the AI arms race going on now, and the incredible sums of money be thrown at any slight advantage, is there any reason to believe that any meaningful AI breakthrough would be openly published for anyone to leverage?

by WarmWash

2/20/2026 at 3:12:55 PM

These folks are MIT, so citations are valuable to them. Citations convert into prestige, academic career progression, or a favorable exit from academia into industry.

Also, I don't see why you couldn't patent this if you wanted to monetize it.

by 542458

2/20/2026 at 5:55:31 PM

> Also, I don't see why you couldn't patent this if you wanted to monetize it.

We all just saw the prior art published for the public. That will preclude patenting this work. Further reduction to practice is required.

(I am not a lawyer).

by BetaDeltaAlpha

2/20/2026 at 6:26:47 PM

Yes there is. Lots of researchers are more interested in making a contribution to societal flourishing than in making incredible sums of money. That’s why there’s still lots of top AI researchers in academia.

by jph00

2/20/2026 at 3:06:06 PM

I do sometimes wonder -- if the transformers paper wasn't published, what would the industry be like? Would the same ideas have been put together in almost the same way weeks or months later somewhere else?

by abeppu

2/20/2026 at 2:57:07 PM

I would say yes.

The reality is that the money being thrown = the time of humans. I guess compute as well, but in terms of people doing innovation - openly published things are the same thing, minus the money.

by mikodin

2/20/2026 at 5:55:59 PM

I know the frontier “labs” are holding back publications.

I don’t think it will last among researchers who think beyond production LLMs

by gdiamos

2/20/2026 at 4:08:41 PM

The inventor's grace period under first to file changes still gives them/their university a year to file if they publish openly.

by cma

2/20/2026 at 11:56:11 AM

Superficially it sounds like this could create a bit more of a move toward doing compaction on some continuous basis, or compacting in batches once you hit the context limit, rather than starting fresh with a summary and system prompt..

Feels like high fidelity, fast compaction could be a path to “solving” long context.

by cadamsdotcom

2/20/2026 at 3:09:48 PM

This looks promising. I've added it to my reading list.

by cs702

2/20/2026 at 12:55:27 PM

This is big for long-horizon tasks

by speedping

2/20/2026 at 2:30:10 PM

None of the compaction accuracies look impressive.

by esafak

2/20/2026 at 2:39:47 PM

I think matching or exceeding the original cache at 20% compacted size is fairly impressive.

by yorwba

2/20/2026 at 3:11:57 PM

The original cache had 70% accuracy, and the alternatives were only worse.

by esafak

2/20/2026 at 4:30:05 PM

It sounds like you looked at figure 1 but not figure 3.

by yorwba