---
title: 'Large Language Models are Zero-Shot Reasoners'
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'This note is CC BY 4.0. It links to the paper and does not reproduce it. Paper license: arXiv non-exclusive distribution (verified on arXiv abs page 2026-11-11; http://arxiv.org/licenses/nonexclusive-distrib/1.0/).'
source: https://iinventedlight.com/library/notes/2205-11916-zero-shot-reasoners.md
source_license: 'arXiv default (non-exclusive distribution license)'
attribution: 'Pointer note by iinventedlight (AI-assisted). Links to: Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, Yusuke Iwasawa, Large Language Models are Zero-Shot Reasoners (arXiv). This note does not reproduce the paper.'
author: 'iinventedlight (AI-assisted)'
date: 2026-11-11
status: published
approved_by: "Cole Withington, 2026-11-11 20:58 PT, Grok Bot chat widget t1717s7407: \"Publish all 100 including the LLMs Encode Clinical Knowledge note\""
checked_on: 2026-11-11
as_of: 2026-11-11
confidence: medium
untested: 'Pointer note only. Abs page read 2026-11-11. The paper is not reproduced here.'
category: arxiv
topic_area: 'ai-research: reasoning'
tags: arxiv, pointer, reasoning
arxiv_id: 2205.11916
alias: 2205.11916-zero-shot-reasoners
format: pointer
type: pointer
sources:
  - id: S1
    title: 'Large Language Models are Zero-Shot Reasoners'
    publisher: arXiv
    authors: 'Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, Yusuke Iwasawa'
    url: https://arxiv.org/abs/2205.11916
    version_read: 'v4 (29 Jan 2023)'
    accessed: 2026-11-11
    license: 'arXiv default (non-exclusive distribution license)'
---
# Large Language Models are Zero-Shot Reasoners

## Summary
Large Language Models are Zero-Shot Reasoners (2022), by Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa, is at https://arxiv.org/abs/2205.11916. The paper shows that LLMs are decent zero-shot reasoners by adding Let's think step by step before each answer. The authors evaluate a single template, which they call Zero-shot-CoT, on arithmetic tasks such as MultiArith, GSM8K, AQUA-RAT, and SVAMP, on symbolic tasks such as Last Letter and Coin Flip, and on other logical tasks such as Date Understanding and Tracking Shuffled Objects. With the large InstructGPT model text-davinci-002, they report MultiArith accuracy moving from 17.7% to 78.7% and GSM8K from 10.4% to 40.7%, and they report similar gains with a 540B parameter PaLM model. Using the same single prompt template, they report that Zero-shot-CoT significantly outperforms zero-shot LLM performances on diverse benchmark reasoning tasks including arithmetics and symbolic reasoning, without any hand-crafted few-shot examples. They report increasing the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with a large InstructGPT model. Title, authors, abstract, and comments were taken from the arXiv abs page. The version is v4 (29 Jan 2023), submitted 24 May 2022.

## Claims
Pointer notes do not quote the paper. See Verification and Sources for the link and license.

## Verification
- Format: pointer note only. The paper is not reproduced here.
- Why it matters: It matters because it showed that a short, generic cue can elicit multi-step reasoning without hand-written few-shot examples.
- Reproduced: not reproduced (2026-11-11).
- Related work: Related notes already in the library include [Chain-of-Thought Prompting](https://iinventedlight.com/library/notes/2201-11903-chain-of-thought-prompting/) and [Plan-and-Solve Prompting](https://iinventedlight.com/library/notes/2305-04091-plan-and-solve-prompting/).
- Limits: This is a pointer note only. The MultiArith and GSM8K figures are the authors' reported results with text-davinci-002 in the v4 paper (29 Jan 2023).
- License: arXiv default (non-exclusive distribution license) on the abs page (2026-11-11; http://arxiv.org/licenses/nonexclusive-distrib/1.0/). Submitted 24 May 2022 (v1); last revised 29 Jan 2023 (v4). Sources: [S1] https://arxiv.org/abs/2205.11916, v4 (29 Jan 2023). checked_on: 2026-11-11. Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it. The note itself is CC BY 4.0.

## When to apply
Use when you need the original zero-shot chain-of-thought paper that studies the Let's think step by step cue.

## Facts
- id: F1
  claim: 'The paper shows that adding a Let''s think step by step cue before the answer can elicit zero-shot chain-of-thought reasoning.'
  confidence: medium
  date: 2026-11-11
  sources: [S1]
  license: CC-BY-4.0
- id: F2
  claim: 'With text-davinci-002, the authors report MultiArith accuracy moving from 17.7% to 78.7% and GSM8K from 10.4% to 40.7%.'
  confidence: medium
  date: 2026-11-11
  sources: [S1]
  license: CC-BY-4.0
- id: F3
  claim: 'They report similar magnitudes of improvement with a 540B parameter PaLM model.'
  confidence: medium
  date: 2026-11-11
  sources: [S1]
  license: CC-BY-4.0

## Sources
- [S1] arXiv 2205.11916, https://arxiv.org/abs/2205.11916. Paper license: arXiv default (non-exclusive distribution license). v4 (29 Jan 2023). Accessed 2026-11-11. This note links to the paper and does not reproduce it.

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).