Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
75 lines (51 loc) · 3.65 KB

CONTRIBUTING.md

File metadata and controls

75 lines (51 loc) · 3.65 KB
Raw

Contributing to LLM Reasoning Research

Thanks for considering a contribution. This repo gets better when people who actually know the literature point out what we've missed or gotten wrong.

What we're looking for

We welcome contributions in these forms:

  • Paper suggestions — Papers we should add to existing topic sections
  • Topic suggestions — New reasoning gaps that deserve their own section (see roadmap in main README)
  • Corrections — Annotations that misrepresent a paper, broken links, factual errors
  • Better organization — Restructuring within a topic, clearer categorization, improved indexing
  • Translations of key papers' contributions — Helping make dense academic work more accessible

What we're not looking for

To keep the repo focused and credible:

  • Not a comprehensive paper dump. We're aiming for curated quality over completeness. A topic section with 15 well-annotated foundational papers is more valuable than one with 200 papers and shallow notes.
  • Not auto-generated annotations. Please don't submit summaries written by LLMs. We can tell, and so can readers. Real human assessments are what makes this repo worth reading.
  • Not promotional content. We won't include papers solely because they're connected to a particular product, lab, or company (including ours). Inclusion is based on contribution to the field.
  • Not unpublished or non-peer-reviewed work without strong justification. ArXiv preprints are fine, especially in fast-moving areas, but we lean toward published or seriously-cited work.

How to suggest a paper

The lightest-weight path:

  1. Open an issue with the title [Paper Suggestion] <Paper Title>
  2. Include:
    • Title, authors, year, venue, link
    • Which topic section it belongs in
    • 2–3 sentences on why it matters and what it contributes
    • (Optional) Any specific limitations or caveats worth noting

If you want to write the full annotation yourself, even better — open a pull request with the addition to the relevant topic file using the standard format:

### [Paper Title]
**Authors** · **Year** · **Venue**
Links

**TL;DR:** One sentence.
**Why it matters:** 2–3 sentences.
**Key insight:** The most important takeaway.
**Limitations:** What this paper doesn't solve.

How to suggest a new topic

For new reasoning gaps that should have their own section:

  1. Open an issue with the title [Topic Proposal] <Topic Name>
  2. Include:
    • Why this is a distinct reasoning gap (rather than a subtopic of an existing one)
    • 3–5 foundational papers that would form the core of the section
    • Why this matters for builders or researchers

We'll discuss in the issue, and if it's a fit, we'll create the section structure together.

Editorial standards

For consistency:

  • Annotations are written from a curator's perspective. We're not the authors; we're explaining why a paper matters to a reader. First-person is fine sparingly.
  • Be honest about limitations. Every paper has them. Including the limitations section makes the annotation more useful, not less.
  • No marketing language. Adjectives like "groundbreaking," "revolutionary," or "game-changing" should be avoided. Let the work speak for itself.
  • Cite the venue. Knowing whether something is from NeurIPS, a workshop paper, or an ArXiv preprint helps readers calibrate.

Code of conduct

This is a small project run by a small team. We expect contributors to be respectful, intellectually honest, and willing to engage with disagreement constructively. That's the bar.

Questions

Open an issue, or reach out to the team at Krellix.