# noahshinn/reflexion

[NeurIPS 2023] Reflexion: Language Agents with Verbal Reinforcement Learning

Repository: https://github.com/noahshinn/reflexion
Canonical: https://ross.abutalabs.com/products/reflexion
Language: Python
License: MIT
License Family: permissive
Topics: ai, artificial-intelligence, llm
Last push: 2025-01-14T07:54:02+00:00

## Health v2 (maintenance only)
Score: 31/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 1, release rhythm 35, longevity 90
- inputs: {"age_days": 1260, "days_push": 596, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3242, forks 316 (observed 2026-08-28T04:07:50.512488+00:00)

## What it is
Reference implementation of the NeurIPS 2023 Reflexion paper, where language agents improve through verbal self-reflection instead of weight updates. It includes notebooks and scripts for running reasoning (HotPotQA), decision-making (AlfWorld), and programming (HumanEval) experiments.

## Use cases
- reproduce the reflexion paper experiments
- build llm agents that learn from self-reflection
- compare react and chain-of-thought agent baselines
- run verbal reinforcement learning experiments
- study agent memory and self-critique strategies

## When to choose
- you want the official reference code for the Reflexion paper
- you are researching agent self-reflection and memory techniques
- you need reproducible HotPotQA, AlfWorld, or HumanEval agent benchmarks

## When to avoid
- you need a production-ready agent framework
- you want maintained tooling rather than research code
- you need agents without OpenAI API dependencies

## Facets
- artifact type: learning-resource
- maturity: maintenance
- function: agent-framework, machine-learning, llm-inference, prompt-engineering
- domain: artificial-intelligence, large-language-models, reinforcement-learning
- platform: python, cli
- tags: reflexion, verbal-reinforcement-learning, research-code, neurips-2023, react-agents, self-reflection, hotpotqa, alfworld, human-eval, ai-agents, research

## Member repositories
- noahshinn/reflexion (main) score 31

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:50.512488+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:24:30.942587+00:00, confidence not recorded.
  - readme: https://github.com/noahshinn/reflexion (fetched 2026-08-28T04:07:50.512488+00:00, sha 466639069f6c)
- Data as of 2026-08-30T08:39:29.467469+00:00.