Birbla

Join the discussion

Write your take first — we'll ask for email only when you're ready to publish.

Sign in
  • Hacker News
  • Very cool stuff.
    by zephyroszhou
  • I like the visualisation. Pretty cool
    by itsnasme
  • neat, combining info from two phrases is hard to see without such a tool.

    are you worried later-layer attention gets drowned out by earlier layers just because there are more of them contributing to the sum?

    by wopak
  • Hmm, I might try to add some controls to limit which layers get summed up. It might be able to reveal more patterns.

    Right now only simple correlations are visible.

    by ifz
  • I highly question this simplistic idea of high vector magnitude = high influence.
    by sva_
  • Same I dont get it just, could you clarify it
    by apnabhidu47
  • I don't disagree with that. I did add an entire caveat paragraph there.

    To me, it's more of a neat visualization, not something that can be used to interpret LLM behavior. Even with a lot of simplification, it can show some interesting patterns.

    by ifz
  • You get what you pay for. If you want to think harder and get more, https://transformer-circuits.pub/2025/attention-qk/index.htm...
    by smallmancontrov
  • You can also mine attention from image models, it's a lot of fun and very interesting.
    by scottcodie
  • I don't know much about LLMs but does that mean you have N^2 computation with the context size since every token needs to track how it relates to every other token?
    by ex-aws-dude
  • Yes, except no with the KV cache. Because tokens aren't modified by future tokens you can cache the meaning of previous tokens. This makes the total effort linear over the entire context (or constant per forward pass).
    by TomatoCo
  • For full self attention yes
    by acedTrex
  • Is the attention explanation of why the model tells like this? I've seen that there are many discussions about this. (Image attention visualizations were not that good I think)
    by lhk931122
  • This is very cool. It's simpler than Bertviz for understanding inference and surface level and a good starting people for new learners as well.
    by sshussain270
  • Oh, I actually didn't know about Bertviz before. That looks interesting as well!
    by ifz
  • The pause-on-load problem is a real UX trap. Stepping through one token at a time is how people actually learn attention; single-step controls would help more than autoplay.
    by Mugshelf
  • This is great, I've read multiple books and watched videos about the attention mechanism. Now that I understand it, this is the clearest example I've seen on how attention works.
    by fuddle
  • UX report. I wished to examine attention state step by step, but I found the animation moved along too fast for that. So I tried pausing...

    On Chromium/linux, pressing pause doesn't pause, instead resetting the animation to it's pre-play state - the current attention highlighting disappears. Pressing play again, restarts at the beginning. Having a commonplace "pause pauses, and play resumes" UI, could allow more time to look over state. A youtube-like slow playback 0.25? option might similarly help. Or perhaps even better, buttons for single stepping. Tnx for your work.

    by mncharity
  • Thanks for trying it out! You can hover / click / tap on any word to focus it. I actually intended for this to be main way to use the visualization, I just added the auto-playback later on so it's more eye-catching on page load.
    by ifz
  • This is great, thank you. I have to teach this stuff on Friday so perfect timing. It's hard to explain the attention mechanism in a way that becomes intuitive because the weighting scheme does not help much with the intuition. Having a visualization like this helps a lot. Don't move that page please since I'll link to it!
    by MCP123