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

Simple tasks showing reasoning breakdown in state-of-the-art LLMs

arxiv.org

Post reply on HN
Print view View original

61–70 of 393 posts

Re: Simple tasks showing reasoning breakdown in state-of-the-art LLMs

#61

Post by CooCooCaCha »

photon_lines wrote: post #44
nerdjon wrote: post #18

For anyone considering reading the paper and like me don't normally read papers like this, open the PDF and think you don't have time to read it due to its length. The main part of the paper is the first 10 pages and a fairly quick read. On to the topic here. This is an interesting example that they are using. It is fairly simplistic to understand as a human (even if we may be inclined to quickly jump to the wrong co…

There's actually a pretty simple solution to this that I thought about testing out and it involves asking the model to re-construct the problem using a logic language (like Prolog) and asking it to execute this type of program in order to come up with a solution rather than attempting simple chain-of-reason training / other methodologies of getting the model to 'reason' through some of these examples. People forget t…

I’m curious how this would work considering knowledge can be fuzzy.

Like if I’m out camping and I sit on a log or a rock those things are not what people usually think of as chairs but they can serve as chairs in that situation.

Re: Simple tasks showing reasoning breakdown in state-of-the-art LLMs

#62

Post by pawelmurias »

eranation wrote: post #11
Closi wrote: post #3

Question is: "Alice has 60 brothers and she also has 212 sisters. How many sisters does Alice’s brother have?" (nb: I have added numbers, it's phrased as X and N in the paper) I must confess, when I tried to answer the question I got it wrong...! (I feel silly). I only realised I got it wrong when I plugged it into GPT-4o and it came back with the correct answer: https://chatgpt.com/share/6eb5fa36-e0fd-4417-87d1-64ca…

Great observation. It would be really interesting to repeat this research with less limiting prompts. I assume they made the restriction to make it easier to parse the answers, perhaps make it two phase, let them respond unlimited, then a follow up prompt ask to compress it to a single answer. I wonder how the results will vary.

When I added a " to the end of the prompt by accident I got a wrong answer.

Re: Simple tasks showing reasoning breakdown in state-of-the-art LLMs

#63

Post by kragen »

colechristensen wrote: post #40
nerdjon wrote: post #18

For anyone considering reading the paper and like me don't normally read papers like this, open the PDF and think you don't have time to read it due to its length. The main part of the paper is the first 10 pages and a fairly quick read. On to the topic here. This is an interesting example that they are using. It is fairly simplistic to understand as a human (even if we may be inclined to quickly jump to the wrong co…

I don't think anybody who actually uses LLMs thinks they are general purpose AI. Like you ask it to do one thing it's amazing, but then you try to modify or do something with extra steps, or just anything with any complexity to it and it falls over.

they're pretty general-purpose; you can ask the same model for recipe suggestions, a fanfic, or verilog. like, they're far, far more general-purpose than any humans i know

it's true that they're not very reliable, but they seem to be not very reliable across many different domains. and they don't seem to be particularly less reliable than the average human, so i think possibly your standards for 'general purpose ai' are set high enough that you would declare humans to be unintelligent (or perhaps not 'general-purpose') if you applied them consistently

you can certainly find particular domains where humans can still do things llms can't, but i haven't seen a persuasive account of why those domains are the more important ones, and of course the converse is also true

Re: Simple tasks showing reasoning breakdown in state-of-the-art LLMs

#65

Post by Miraste »

llm_trw wrote: post #23
Closi wrote: post #3

Question is: "Alice has 60 brothers and she also has 212 sisters. How many sisters does Alice’s brother have?" (nb: I have added numbers, it's phrased as X and N in the paper) I must confess, when I tried to answer the question I got it wrong...! (I feel silly). I only realised I got it wrong when I plugged it into GPT-4o and it came back with the correct answer: https://chatgpt.com/share/6eb5fa36-e0fd-4417-87d1-64ca…

>Worth noting that the prompts from the experiment include "To answer the question, DO NOT OUTPUT ANY TEXT EXCEPT following format that contains final answer: ### Answer:" so it appears that they are stopping the models from 'thinking out loud'. If I add that to the prompt, GPT4o gets it consistently wrong... Yes this is a common thing I see people who think LLMs are idiots do. The more an LLM talks the smarter it ge…

As it says in the paper, they already did the research with less limiting prompts. All models were tested with another prompt that gave only the question and no further instruction, and a third prompt that asked the model to consider its answer carefully.

The correct response rate chart doesn't even use the results from the concise prompt.

Re: Simple tasks showing reasoning breakdown in state-of-the-art LLMs

#67

Post by gweinberg »

naikus wrote: post #8

I'm using chatGPT 4o and it gave me the following :) ME Alice has 40 brothers and she also has 50 sisters. How many sisters does Alice's brother have? ChatGPT To determine how many sisters Alice's brother has, we need to use the information provided: Alice has 40 brothers and 50 sisters. Since Alice's brother is one of the 40 brothers, he shares the same family structure as Alice. This means he has the same number of…

I don't know why you are getting downvoted. I was disappointed the paper didn't give examples of the LLM confidently expressing its incorrect reasoning.

From a human perspective, I think there are three ways to get the answer wrong: failure to realize that Alice's brother has pretty much the same number of sisters as Alice herself, failure to realize that the brother has one additional sister, namely Alice, and failure to successfully add one to the number of sisters. But that assumes that the LLM is more or less reasoning like a human. It may well be "reasoning" more along the lines of "I've seen lots of story problems like this, the modal answer was five, I'll say five"

Re: Simple tasks showing reasoning breakdown in state-of-the-art LLMs

#70

Post by pama »

Nothing original to see here and nothing reformulated to be of practical importance other than a clickbait attempt. If this is your first trap of reading absurdist/populist LLM literature consider yourselves lucky. If you care about these models you know how powerful they can be and you probably learned how to use them. If you care and dont know how to use them, it helps to get some background on earlier language model research (maybe 2010–2018) instead to grasp the probabilistic underpinnings of causal inference and then some of the early OpenAI papers about GPT.
Post reply on HN