Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
146 lines (101 loc) · 5.81 KB

CONTRIBUTING.md

File metadata and controls

146 lines (101 loc) · 5.81 KB
Raw

Contributing

Thank you for helping keep this list focused and useful.

How to Contribute

  • Suggest a resource — open an issue using the Suggest a resource form. It collects everything needed to vet the resource against this document.
  • Report a broken link — use the Broken link form.
  • Open a pull request — add the entry yourself following the rules below. The PR template includes the quality checklist, and CI validates formatting, Markdown style, awesome-list compliance, and link integrity.

This repository also ships an agent-assisted curation pipeline (discovery, vetting, review, and audit agents) described in AGENTS.md; it applies the same rules in this document.

Purpose

This repository curates high-signal resources on recursive language models, recursive inference, recursive reasoning architectures, self-calling AI systems, and adjacent work where recursion is central to the method.

It also supports research into learned simulation engines for society: AI systems that can model social, institutional, and behavioural dynamics through recursive agents, simulated populations, evaluation loops, reinforcement learning, and inspectable long-horizon reasoning.

Resource Inclusion Criteria

Contributions should satisfy all of the following:

  • The resource is directly relevant to recursion in AI systems.
  • Recursion is central to the method, architecture, inference process, evaluation design, agent loop, or self-improvement mechanism.
  • The resource is technically credible and useful to researchers or builders.
  • The source is preferably primary: arXiv, official project page, official GitHub repository, official documentation, author website, or conference page.
  • The description is accurate, concise, and non-promotional.

Time-Window Rule

Resources should be from 15 May 2022 onwards.

Older resources may only be proposed for a short historical context section if they are necessary for understanding the field. Mark them clearly as outside the main inclusion window.

Link Verification

Before submitting a resource:

  • Open the link.
  • Confirm the page loads.
  • Confirm the title on the page matches the proposed title.
  • Confirm the linked source is the intended resource, not a secondary summary unless no primary source exists.

Description Accuracy

Descriptions should state what the resource is and why it matters for recursive language models or recursive agent systems.

Avoid:

  • vague praise;
  • unsupported claims;
  • exaggerated importance;
  • invented venues, authors, dates, benchmarks, or results;
  • descriptions copied from abstracts without checking relevance.

Duplicate Prevention

Before adding a resource:

  • Search the README for the title, project name, arXiv identifier, repository name, and common abbreviation.
  • Do not add the same paper, repository, or project in multiple sections unless there is a strong reason.
  • If a resource is relevant to another section, prefer a short cross-reference in prose rather than duplicating the entry.

Preferred Source Hierarchy

Use the most authoritative available source:

  1. arXiv paper, conference page, or official proceedings page.
  2. Official project page.
  3. Official GitHub repository.
  4. Official documentation.
  5. Author or lab page.
  6. High-quality secondary explanation, only when it adds clear value and the primary source is also listed or unavailable.

Suggested Contribution Format

Use this format:

- [Resource title](URL) - One concise sentence explaining what it is and why it matters. (Year)

Example:

- [Tree of Thoughts: Deliberate Problem Solving with Large Language Models](https://arxiv.org/abs/2305.10601) - Frames reasoning as search over intermediate thought states with generation, self-evaluation, and selection. (2023)

Quality Checklist

Before submitting, confirm:

  • The resource is within the inclusion time window (15 May 2022 onwards) or clearly marked as historical context.
  • The link works.
  • The title matches the linked page.
  • The description is accurate.
  • The resource is not already listed.
  • The resource is directly relevant to recursion in AI systems.
  • The resource is not merely a general LLM, prompt engineering, RAG, or agents resource.
  • The resource supports the repository's technical purpose.
  • The year is correct.
  • No duplicate entry has been introduced.

Out-of-Scope Resources

Do not add:

  • generic LLM papers without an explicit recursive mechanism;
  • general prompt engineering resources;
  • general RAG resources where recursive retrieval, summarisation, or decomposition is not central;
  • broad AI agent lists or frameworks without recursive planning, reflection, self-correction, or tool-use loops;
  • unverified blog posts, newsletters, or social-media summaries;
  • resources with broken links;
  • resources whose date cannot be verified;
  • resources whose claims cannot be matched to the linked source.

Validation

Before opening a pull request, run the same checks CI runs:

npm ci                 # install dev tooling
npm run lint           # markdownlint
npm run lint:awesome   # awesome-list compliance for README.md
npm test               # link check (linkinator)

All three checks must pass. Keep diffs minimal: README.md and this file are hand-curated and must not be wholesale reformatted.

Review Standard

This list should remain concise, technically credible, and research-oriented. A smaller set of well-verified resources is preferable to broad coverage with weak relevance.

Licensing

This repository is released under CC0 1.0. By contributing, you agree that your contributions are released under the same terms.