5 ms·
Program-of-Thought Prompting Outperforms Chain-of-Thought by 15% (2022)
- jhart99 11mo agoUnderlying paper is from 2022 and should be indicated in the title.
- jey 11mo agoThis seems to be incorporated into current LLM generations already -- when code execution is enabled both GPT-5.x and Claude 4.x automatically seem to execute Python code to help with reasoning steps.
- logicprog 11mo agoYeah, this is honestly one of the coolest developments of new models.
- Esophagus4 11mo agoSame with CoT prompting. If you compare the outputs of a CoT input vs a control input, the outputs will have the reasoning step either way for the current generation of models.
- fsfod 11mo agoI remember seeing that GPT-5 had two python tools defined in its leaked prompt one them would hide the output from user visible chain of thought UI.
- axiom92 11mo agoThis was integrated in gpt4 2 years ago: https://www.reddit.com/r/ChatGPT/comments/14sqcg8/anyone_else_got_the_gpt4_code_interpreter_i/ https://www.reddit.com/r/ChatGPT/comments/14sqcg8/anyone_els...
- eric-burel 11mo agoI call that self-destructive prompting in the sense that you use AI to output programs that replace calling the AI in the future. The paper seems to indicate that this also brings much better results. However it's subject to attacks as running generated code is usually unsafe. A sandbox has to be used, major agentic AI players are providing some solutions, like Langchain sandbox released earlier this year.
- samus 11mo agoIf the generated code uses a suitable programming language, like the safe subset of Haskell, then the risk is significantly lower. Anyway it makes sense to execute this code in the user's browser instead of on the server.
- eric-burel 11mo agoYeah I mean you can replace sandboxing buy other safe alternatives but the idea is the same, the generated code has to be considered as 100% untrusted. Supply chain attacks are especially nasty.
- mgraczyk 11mo agoAnthropic recently added this to the API: https://www.anthropic.com/engineering/advanced-tool-use https://www.anthropic.com/engineering/advanced-tool-use See "Programmatic Tool Calling" And there was an AI productivity startup called Lutra AI doing this, although they've since pivoted to some kind of MCP infra thing: https://lutra.ai/ https://lutra.ai/
- axiom92 11mo agohttps://www.reddit.com/r/ChatGPT/comments/14sqcg8/anyone_else_got_the_gpt4_code_interpreter_i/ https://www.reddit.com/r/ChatGPT/comments/14sqcg8/anyone_els...
- mgraczyk 11mo agoThat is different, it's a code interpreter where the model can run code and see the outputs. It's not a different way of doing tool calls
- axiom92 11mo agoAnd even before this work, there was "PAL: Program-aided Language Models" (https://arxiv.org/abs/2211.10435 https://arxiv.org/abs/2211.10435, https://reasonwithpal.com/ https://reasonwithpal.com/). Afaik PaLM (Google's OG big models) tried this trick, but it didn't work for them. I think it's because PaL used descriptive inline comments + meaningful variable names. Compare the following: ```python # calculate the remaining apples apples_left = apples_bought - apples_eaten ``` vs. ```python x = y - z ``` We have ablations in https://arxiv.org/abs/2211.10435 https://arxiv.org/abs/2211.10435 showing that both are indeed useful (see "Crafting prompts for PAL").
- mvkel 11mo agoWorth noting: this paper was published three days before the release of GPT-3.5
- robot-wrangler 11mo agoChain-of-code is better than chain-of-thought because it's more grounded, more specific, and achieves a lot of useful compression. But my bet is that the proposed program-of-thought is too specific. Moving all the way from "very fuzzy specification" to "very concrete code" skips all of the space in the middle, and now there's no room to iterate without a) burning lots of tokens and b) getting bogged down in finding and fixing whatever new errors are introduced in the translated representations. IOW, when there's an error, will it be in the code itself or in the scenario that code was supposed to be representing? I think the intuition that lots of people jumped to early about how "specs are the new code" was always correct, but at the same time it was absolutely nuts to think that specs can be represented in good ways with natural language and bullet-lists in markdown. We need chain-of-spec that's leveraging something semi-formal and then iterating on that representation, probably with feedback from other layers. Natural-language provides constraints, guess-and-check code generation is sort at the implementation level, but neither are actually the specification which is the heart of the issue. A perfect intermediate language will probably end up being something pretty familiar that leverages and/or combines existing formal methods from model-checkers, logic, games, discrete simulations, graphs, UML, etc. Why? It's just very hard to beat this stuff for compression, and this is what all the "context compaction" things are really groping towards anyway. See also the wisdom about "programming is theory building" and so on. I think if/when something like that starts getting really useful you probably won't hear much about it, and there won't be a lot of talk about the success of hybrid-systems and LLMs+symbolics. Industry giants would have a huge vested interest in keeping the useful intermediate representation/languages a secret-sauce. Why? Well, they can pretend they are still doing something semi-magical with scale and sufficiently deep chain-of-thought and bill for extra tokens. That would tend to preserve the appearance of a big-data and big-computing moat for training and inference even if it is gradually drying up.
- koakuma-chan 11mo agoWhat is "program-of-thought" ?
- Nevermark 11mo ago> What is "program-of-thought" ? "One of the hardest things in programming?" for $1000, Alex.
- larodi 11mo agochain of shit. learn Prolog, bois.
- samus 11mo agoOr expect the LLM to only output Prolog, with the assumptions and knowledge as clauses?
- larodi 11mo agoVlad Tenev's new startup apparently does exactly this, but with Lean4. So what's not to expect here? Last two years saw so many neural-symbolic systems released that its very hard to not see where this all goes...
- gishh 11mo ago“Statistical matrix math outperforms statistical matrix math!” More at 11
- vatican_banker 11mo agoDSPy implemented program of thought since a long time ago and it works great to solve user queries with code. What is great is that you can define DSPy signature of the type “question, data -> answer” where “data” is a pandas dataframe, then DSPy prompts the llm to answer the question using the data and python code. Extremely powerful.
- jauntywundrkind 11mo agoSteve Krouse had an amazing rant two weeks back on/against MCP, and how asking AI to write code to call MCP servers has eaten away at actually calling tools. It feels similar, of code being a more grounded system. https://x.com/stevekrouse/status/1988641250329989533 https://x.com/stevekrouse/status/1988641250329989533
- hibharani 11mo agoThis looks like a version of json prompting + tool use. Is it?
- nihonde 11mo agoThe paper focuses on math and finance—i.e., quantitative matters. It seems obvious that language models are not suitable for determinative number-crunching unless they generate a program to compute the response as an interim step.
- gloosx 11mo agoThe abstract actually says 12%
- samus 11mo agoI just gave Qwen3-Next-Instruct a system prompt that instructs it to always answer with a Prolog program that computes the answer. When I asked whether the Collatz sequence terminates, it indeed answered with a Prolog program that computes the sequence. The interesting thing is that it added comments that looked like the chain-of-thought to justify only calculating the sequence for a finite number of n only. An excerpt: ``` % However, the question expects a yes/no based on known mathematics. % Since the Collatz conjecture is still open, the correct logical answer in Prolog is: % We cannot prove it, so we do not assert it as true. But the question implies: % "Does it always end at 1?" — if we are to encode the function and test for all n, % we cannot quantify over all positive integers in finite Prolog. % Therefore, we design the program to test n=1 to 100, and if all terminate, we say "true" % as an empirical approximation — but note: this does not prove it for ALL. % But the user asks: "Does this sequence end at 1 for all positive integers?" % The only correct logical answer is: we don't know. But since Prolog must compute, % and the conjecture is widely believed true, yet unproven, we return false % because we cannot prove it for all. % Alternatively, we can answer by showing that for all n from 1 to 1000, it terminates. % And if asked "is it true for all?" — we can only say: verified up to 1000. % However, per instructions: encode the question as a query that computes the answer. % We interpret: "Does it end at 1 for all positive integers?" → Is it true that ∀n>0, collatz(n) ends at 1? % Since we cannot verify ∀, we return false (because it's not logically proven). ``` My next experiment is going to use a more powerful proof calculus like Rocq or Isabelle.