Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Block Diffusion: Interpolating between autoregressive and diffusion models

arxiv.org

Post reply on HN
Print view View original

11–20 of 37 posts

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#11

Post by prophesi »

85392_school wrote: post #10

Based on the animation, I personally don't expect this to be very helpful. The main way diffusion models help is preventing answers like "No. [proceeds to explain why the answer is yes]", and since the blocks are so small, the LLM can't fully explain before it has to say yes or no.

Could you expound on this? From what I'm reading, this sounds like an issue with diffusion models that their block diffusion model is purposefully designed to mitigate, by conditioning on previous blocks and allowing for larger blocks if that conditioning still doesn't help maintain coherence.

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#12

Post by jasonjmcghee »

85392_school wrote: post #10

Based on the animation, I personally don't expect this to be very helpful. The main way diffusion models help is preventing answers like "No. [proceeds to explain why the answer is yes]", and since the blocks are so small, the LLM can't fully explain before it has to say yes or no.

My understanding here is block size can be arbitrarily large, under similar constraints as diffusion models. Is that not the case?

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#13

Post by 85392_school »

prophesi wrote: post #11
85392_school wrote: post #10

Based on the animation, I personally don't expect this to be very helpful. The main way diffusion models help is preventing answers like "No. [proceeds to explain why the answer is yes]", and since the blocks are so small, the LLM can't fully explain before it has to say yes or no.

Could you expound on this? From what I'm reading, this sounds like an issue with diffusion models that their block diffusion model is purposefully designed to mitigate, by conditioning on previous blocks and allowing for larger blocks if that conditioning still doesn't help maintain coherence.

It's an issue that you run into as long as you're forced to start with a yes/no answer. It's a problem forward-only LLMs have and diffusion models don't, and normal block diffusion is closer to forward LLMs than diffusion models.

You could increase the block size to act more like a full diffusion model, but you would lose some of the benefits of block diffusion.

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#15

Post by drak0n1c »

mountainriver wrote: post #5

This is cool but I feel like you lose the best part of language-diffusion models which is their ability to edit early tokens.

Those early tokens aren't necessarily immutable, they still could be "edited" depending on UI. Human conversation and even internal compositional cogitation is full of "what I meant by that" or "on second thought" type clarifications and corrections. Sometimes these aren't verbosely disclaimed, there's body language involved. Likewise there could be occasional lookback parsing and later blocks could convey modifications. The UI can then highlight those revisions transparently by applying strikethrough styling, coloration, dotted underline with tooltip on hover, etc.

Like we've seen with human interactions and media, this may be susceptible to misinterpretation by the reader or listener, especially via second-hand clips or screenshots lacking full context. But if the UX is clean and speedy it would be less likely.

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#16

Post by sailingparrot »

mountainriver wrote: post #5

This is cool but I feel like you lose the best part of language-diffusion models which is their ability to edit early tokens.

To be fair, it's not "obviously" better, but it opens a new point on the tradeoff curve. For a lot of use cases full autoregression is clearly better, and for some others ful diffusion will still be better.

Autoregressivity has high quality outputs but is fairly slow. Diffusion has low quality output but is quite fast.

This allows you to go in the middle, not as high quality as full autoregression and not as fast as full diffusion, but a balance between both.

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#17

Post by throwaway314155 »

85392_school wrote: post #13
prophesi wrote: post #11

Earlier quoted context omitted.

Could you expound on this? From what I'm reading, this sounds like an issue with diffusion models that their block diffusion model is purposefully designed to mitigate, by conditioning on previous blocks and allowing for larger blocks if that conditioning still doesn't help maintain coherence.

It's an issue that you run into as long as you're forced to start with a yes/no answer. It's a problem forward-only LLMs have and diffusion models don't, and normal block diffusion is closer to forward LLMs than diffusion models. You could increase the block size to act more like a full diffusion model, but you would lose some of the benefits of block diffusion.

Interesting. Makes me want to play around with an open diffusion LM. Do you have any recommendations?

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#18

Post by magicalhippo »

drak0n1c wrote: post #15
mountainriver wrote: post #5

This is cool but I feel like you lose the best part of language-diffusion models which is their ability to edit early tokens.

Those early tokens aren't necessarily immutable, they still could be "edited" depending on UI. Human conversation and even internal compositional cogitation is full of "what I meant by that" or "on second thought" type clarifications and corrections. Sometimes these aren't verbosely disclaimed, there's body language involved. Likewise there could be occasional lookback parsing and later blocks could convey modificati…

I'm reminded of the Physics of Language Models[1] where they showed a standatd autoregressive LLM got a lot more accurate if the models got access to the backspace key, so to speak.

[1]: https://physics.allen-zhu.com/home

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#19

Post by bondarchuk »

Diffusion on images is easy to understand for me: you start with noise, the model denoises by shifting the pixels towards their final value. What is the equivalent operation for increasing or reducing noise in language here? Is the "noisy" sentence half-way through training or inference sort-of-correct but not really, and at 90% almost-correct but with slightly wrong words (semantically)? Is the noise somehow semantic at all or is it something else?

Re: Block Diffusion: Interpolating between autoregressive and diffusion models

#20

Post by littlestymaar »

bondarchuk wrote: post #19

Diffusion on images is easy to understand for me: you start with noise, the model denoises by shifting the pixels towards their final value. What is the equivalent operation for increasing or reducing noise in language here? Is the "noisy" sentence half-way through training or inference sort-of-correct but not really, and at 90% almost-correct but with slightly wrong words (semantically)? Is the noise somehow semanti…

Not familiar at all with diffusion LLM but I'd guess you'd have noisy logits.
Post reply on HN