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

Scaling up test-time compute with latent reasoning: A recurrent depth approach

arxiv.org

Post reply on HN
Print view View original

21–30 of 45 posts

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#21

Post by jonathanrmumm »

viraptor wrote: post #18
janalsncm wrote: post #8

Earlier quoted context omitted.

> seems a step in the right direction I can’t see why. I can’t think of any problems where recurrent loops with latent streams would be preferable to tokens. And the downsides are obvious. > externally specifying the number of recurrent iterations Yeah this seems wrong to me. At least with RL training you saw that the length of the CoT decreased dramatically before climbing again, as the model became more proficient.

> I can’t think of any problems where recurrent loops with latent streams would be preferable to tokens. Efficiency. The written language is extremely inefficient. By running through whole concepts at a time instead of parts of a word the reasoning time will be much more concise.

If we're talking conscious thought, millions of simultaneously firing neurons to form words. If we're unconscious intelligence, it's closer to latent space. A lot of intelligence that can't be articulated.

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#22

Post by nielsole »

janalsncm wrote: post #7

One of the benefits of using thinking tokens compared to “thinking in a latent” space is that you can directly observe the quality of the CoT. In R1 they saw it was mixing languages and fixed it with cold start data. It would be hard to SFT this because you can only SFT the final result not the latent space. I also notice the authors only had compute for a single full training run. It’s impressive they saw such good…

With a bit of fiddling you should be able to get the LLM to translate/summarize the thinking process. Not a 1:1 thing, but still

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#23

Post by WithinReason »

nielsole wrote: post #22
janalsncm wrote: post #7

One of the benefits of using thinking tokens compared to “thinking in a latent” space is that you can directly observe the quality of the CoT. In R1 they saw it was mixing languages and fixed it with cold start data. It would be hard to SFT this because you can only SFT the final result not the latent space. I also notice the authors only had compute for a single full training run. It’s impressive they saw such good…

With a bit of fiddling you should be able to get the LLM to translate/summarize the thinking process. Not a 1:1 thing, but still

how would you do it?

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#24

Post by viraptor »

jonathanrmumm wrote: post #21
viraptor wrote: post #18

Earlier quoted context omitted.

> I can’t think of any problems where recurrent loops with latent streams would be preferable to tokens. Efficiency. The written language is extremely inefficient. By running through whole concepts at a time instead of parts of a word the reasoning time will be much more concise.

If we're talking conscious thought, millions of simultaneously firing neurons to form words. If we're unconscious intelligence, it's closer to latent space. A lot of intelligence that can't be articulated.

(citation needed) It sounds fun and all, but we barely have any connection between human brain and llms as they exist today.

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#25

Post by mola »

anothermathbozo wrote: post #16
optimalsolver wrote: post #15

Earlier quoted context omitted.

Is there any actual evidence that the reasoning tokens output by current models actually represent the computation happening in the hidden layers? In both cases, the model is doing a ton of processing that you can't actually inspect, except here, you at least get some efficiency gains. Even more importantly, you're also less likely to convince yourself that you know what the model is thinking.

No and we’ve observed evidence to the contrary

Do you have some reading material on this? How did they understand the difference between stated cot and "actual processing"

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#26

Post by nielsole »

WithinReason wrote: post #23
nielsole wrote: post #22

Earlier quoted context omitted.

With a bit of fiddling you should be able to get the LLM to translate/summarize the thinking process. Not a 1:1 thing, but still

how would you do it?

my naive way would be to try to do seq2seq with the hidden state as input. Not sure how to replace the supervised samples though.

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#27

Post by WithinReason »

nielsole wrote: post #26
WithinReason wrote: post #23

Earlier quoted context omitted.

how would you do it?

my naive way would be to try to do seq2seq with the hidden state as input. Not sure how to replace the supervised samples though.

OK but what would you use as ground truth?

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#28

Post by porridgeraisin »

viraptor wrote: post #24
jonathanrmumm wrote: post #21

Earlier quoted context omitted.

If we're talking conscious thought, millions of simultaneously firing neurons to form words. If we're unconscious intelligence, it's closer to latent space. A lot of intelligence that can't be articulated.

(citation needed) It sounds fun and all, but we barely have any connection between human brain and llms as they exist today.

We need to reboot Bryan Cantrill's "Don't anthropomorphize the lawn mower" talk with a new edition titled "Don't anthropomorphize the internet document simulator"

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#29

Post by Imanari »

ckrapu wrote: post #9

My opinion is that opaque reasoning is a prerequisite for many of the worst possible AI outcomes. We should make reasoning fully visible in the output space.

maybe let it reason in latent space but have a method to transform and output it to text for inspection.

Re: Scaling up test-time compute with latent reasoning: A recurrent depth approach

#30

Post by Tostino »

thomasahle wrote: post #6
HarHarVeryFunny wrote: post #4

Latent / embedding-space reasoning seems a step in the right direction, but building recurrence into the model while still relying on gradient descent (i.e. BPTT) to train it seems to create more of a problem (training inefficiency) than it solves, especially since they still end up externally specifying the number of recurrent iterations (r=4, 8, etc) for a given inference. Ideally having recurrence internal to the…

> Latent / embedding-space reasoning seems a step in the right direction Might be good for reasoning, but it's terrible for interpretation / AI-safety.

Why is it any different to do 4 recurrent passes than having a model that is 4x deeper?
Post reply on HN