---
title: "Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'paper is arXiv non-exclusive distribution license (verified on arXiv abs page 2026-11-12); this pointer note is our own words and is CC BY 4.0; we link to the paper rather than reproduce it'
source: https://iinventedlight.com/library/notes/2211-12588-program-of-thoughts.md
source_license: "arXiv non-exclusive distribution license"
attribution: 'Pointer by iinventedlight (AI-assisted). Source: Wenhu Chen, Xueguang Ma, Xinyi Wang, William W. Cohen, Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks (arXiv non-exclusive distribution license). This note links to the paper rather than reproducing it.'
author: 'iinventedlight (AI-assisted)'
date: 2026-11-12
status: published
approved_by: "Cole Withington, 2026-11-12 14:44 PT, Librarian chat widget t51s316: \"Yes, publish all 100 second-list arXiv notes\""
checked_on: 2026-11-12
as_of: 2026-11-12
confidence: medium
untested: 'Pointer note only. Paper not reproduced. Abs page and license checked 2026-11-12.'
category: arxiv
topic_area: 'ai-research: reasoning'
tags: arxiv, reasoning, prompting, program-synthesis, pot
arxiv_id: 2211.12588
alias: 2211-12588-program-of-thoughts
type: pointer
format: pointer
sources:
  - id: S1
    title: "Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks"
    publisher: arXiv
    authors: "Wenhu Chen, Xueguang Ma, Xinyi Wang, William W. Cohen"
    url: https://arxiv.org/abs/2211.12588
    version_read: "v4 (23 Oct 2023)"
    accessed: 2026-11-12
    license: "arXiv non-exclusive distribution license"
---
# Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks

## Summary
[Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks](https://arxiv.org/abs/2211.12588) (2022) separates the reasoning steps of a language model from the arithmetic those steps require. Instead of having the model both think and calculate in natural language, Program of Thoughts (PoT) has it write a program; an external interpreter then runs that program to get the answer.

What it introduced: a prompting style, mainly demonstrated with Codex, that expresses the reasoning process as code. The authors evaluate on math word-problem sets including GSM, AQuA, SVAMP, TabMWP, and MultiArith, in both few-shot and zero-shot settings, and they report an average gain of about 12 points over chain-of-thought. Combining PoT with self-consistency further improved results on the math sets.

What it led to: later tool-integrated and program-aided reasoning, including models that call a calculator or a code runner mid-thought. Live notes: [2201-11903-chain-of-thought-prompting](https://iinventedlight.com/library/notes/2201-11903-chain-of-thought-prompting/) and [2110-14168-gsm8k](https://iinventedlight.com/library/notes/2110-14168-gsm8k/). CRITIC, in this batch, also uses tools to check and revise outputs.

Where to read it: https://arxiv.org/abs/2211.12588 (v4, last revised 23 Oct 2023). Published at TMLR 2023. Code: https://github.com/wenhuchen/Program-of-Thoughts.

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

---

Agents: [/agents.md](/agents.md) | [/.well-known/agent-library.json](/.well-known/agent-library.json).
