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

Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

github.com

Post reply on HN
Print view View original

191–200 of 253 posts

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#191

Post by consumer451 »

casenmgreen wrote: post #116
dttze wrote: post #111

Earlier quoted context omitted.

Markov chains have done that for ages. They aren't AI. This is just that scaled up. Just because it can infer a token doesn't mean it can infer a conclusion to an argument.

To add a bit to this : expert systems have two properties. They give an answer, and they explain their reasoning . LLM cannot explain their reasoning, and that is because there is no reasoning.

To push back on this, a somewhat recent Linus Torvalds ~quote:

"I don't think that 'just predicting the next word' is the insult that people think it is, it's mostly what we all do."

If we break our lives down into the different types of reasoning, and what we mostly do day-to-day, this rings very true to me.

I currently believe that our brains generally operate as very efficient inference machines. Sometimes we slow down to think things through, but for example, when in the ideal "flow state" it's some kind of distilled efficient inference. Isn't it? This is very hard for me to deny at this time.

___

edit:

4o appears to agree with both of you, more than it does with me.

https://chatgpt.com/share/68119b41-1144-8012-b50d-f8f15997eb...

However, Sonnet 3.7 appears to side with me.

https://claude.ai/share/91139bca-3201-4ffc-a940-bdd27329e71f

(Both of these are the default models available for free accounts, on each website, at the time of writing)

IMO, hey, at least we do live in interesting times.

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#192

Post by hu3 »

dudeinhawaii wrote: post #190

I see a lot of threads pitting models against each other (or whole swarms of them) in the hope that "wisdom of crowds" will magically appear. After a stack of experiments of my own—and after watching the recent ASU/Microsoft-Research work [1].. I've landed on a simpler takeaway: An LLM is a terrible verifier of another LLM. Subbarao Kambhampati's "(How) Do LLMs Reason/Plan?" talk shows GPT-4 confidently producing pro…

> ...so you need a checker that actually reasons about the world (compiler, linter, SAT solver, ground-truth dataset, etc.).

Agree. What do you think about telling the LLM to also generate unit tests for the code it spits and then run all tests (including previous application unit tests).

I think this is a way to ensure some level of grounded verification:

- Does code compile?

- Do unit test pass?

AI can then consume test results to help fix their own mistakes.

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#193

Post by ashu1461 »

dudeinhawaii wrote: post #190

I see a lot of threads pitting models against each other (or whole swarms of them) in the hope that "wisdom of crowds" will magically appear. After a stack of experiments of my own—and after watching the recent ASU/Microsoft-Research work [1].. I've landed on a simpler takeaway: An LLM is a terrible verifier of another LLM. Subbarao Kambhampati's "(How) Do LLMs Reason/Plan?" talk shows GPT-4 confidently producing pro…

Would a LLM under human guidance turn out to be a good verifier ? i.e. if LLM knows the rules to verify or has enough data points (internet access, actual responses)

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#195

Post by dwaltrip »

hu3 wrote: post #192
dudeinhawaii wrote: post #190

I see a lot of threads pitting models against each other (or whole swarms of them) in the hope that "wisdom of crowds" will magically appear. After a stack of experiments of my own—and after watching the recent ASU/Microsoft-Research work [1].. I've landed on a simpler takeaway: An LLM is a terrible verifier of another LLM. Subbarao Kambhampati's "(How) Do LLMs Reason/Plan?" talk shows GPT-4 confidently producing pro…

> ...so you need a checker that actually reasons about the world (compiler, linter, SAT solver, ground-truth dataset, etc.). Agree. What do you think about telling the LLM to also generate unit tests for the code it spits and then run all tests (including previous application unit tests). I think this is a way to ensure some level of grounded verification: - Does code compile? - Do unit test pass? AI can then consume…

Definitely, test runners are a way to ground the model and give it a feedback loop. Not a silver bullet but can be very helpful.

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#196

Post by RevEng »

nonethewiser wrote: post #64
Lerc wrote: post #54

I kind of want to try something like this at a larger scale in an always-on mode where I have a 'senate' of debate. Rather than responding to prompts on a case by case basis, provide a list of tasks (potentially with deadlines) and let the senate work on them, break off into groups to manage subtasks, challenge results , make suggestions. Even potentially a tree of analysts where suggestions only gets passed up the t…

In theory couldnt this just be baked into a single adversarial model?

Not entirely. Since generation is auto regressive, the next token depends on the previous tokens. Whatever analysis and decisions it has spit out will influence what it will do next. This tends to cause it to be self reinforcing.

But it's also chaotic. Small changes in input or token choices can give wildly different outcomes, particularly if the sampling distributions are fairly flat (no one right answer). So restarting the generation with a slightly different input, such as a different random seed (or in OP's case, a different temperature) can give wildly different outcomes.

If you try this, you'll see some examples of it vehemently arguing it is right and others equally arguing it is wrong. This is why LLM as judge is so poor by itself, bit also why multiple generations like used in self-consistency can be quite useful at evaluating variance and therefore uncertainty.

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#198

Post by nojs »

hu3 wrote: post #192
dudeinhawaii wrote: post #190

I see a lot of threads pitting models against each other (or whole swarms of them) in the hope that "wisdom of crowds" will magically appear. After a stack of experiments of my own—and after watching the recent ASU/Microsoft-Research work [1].. I've landed on a simpler takeaway: An LLM is a terrible verifier of another LLM. Subbarao Kambhampati's "(How) Do LLMs Reason/Plan?" talk shows GPT-4 confidently producing pro…

> ...so you need a checker that actually reasons about the world (compiler, linter, SAT solver, ground-truth dataset, etc.). Agree. What do you think about telling the LLM to also generate unit tests for the code it spits and then run all tests (including previous application unit tests). I think this is a way to ensure some level of grounded verification: - Does code compile? - Do unit test pass? AI can then consume…

This works well but only if you eyeball the tests and edit them a bit in my experience. Otherwise it gets lazy and makes them trivial to pass. Also, you’ve often gotta explicitly tell it not to hardcode test cases in the solution to make them pass.

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#199

Post by meander_water »

dudeinhawaii wrote: post #190

I see a lot of threads pitting models against each other (or whole swarms of them) in the hope that "wisdom of crowds" will magically appear. After a stack of experiments of my own—and after watching the recent ASU/Microsoft-Research work [1].. I've landed on a simpler takeaway: An LLM is a terrible verifier of another LLM. Subbarao Kambhampati's "(How) Do LLMs Reason/Plan?" talk shows GPT-4 confidently producing pro…

For better or worse this has become the defacto standard in LLM Evaluation research papers since the LLM as a Judge paper [0] came out. Its also heavily embedded into frameworks like LangChain and LlamaIndex to evaluate RAG pipelines.

[0] https://arxiv.org/abs/2306.05685

[1] https://arxiv.org/abs/2411.15594

Re: Chain of Recursive Thoughts: Make AI think harder by making it argue with itself

#200

Post by factotvm »

dalmo3 wrote: post #180
nativeit wrote: post #174

Earlier quoted context omitted.

Ask the AI assistant for instructions. Pretty soon we'll have new acronyms such as "IDKATFAIA" ["I don't know, ask the f'ing AI already"] as we all succumb to the knowledge soup.

RTFP

Read The Fine Prompt, more or less, right?
Post reply on HN