Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
35 lines (23 loc) · 2.95 KB

CONTRIBUTING.md

File metadata and controls

35 lines (23 loc) · 2.95 KB
Raw

Contributing to Proactive Atlas

Help keep the collection useful, traceable, and easy to maintain. Bibliographic content and the LowEntropyAI website share the same Markdown sources.

Add or correct a paper

  1. Check README.md for the title and canonical source URL to avoid duplicates.
  2. Explain the connection to proactive AI: a proactive decision or capability, a foundation or human-factor question, an evaluation target, or a supporting mechanism. Describe the actual contribution and available evidence; identify proactive subsets when the work covers broader tasks.
  3. Add a row under the best-fitting ## Papers section using the existing five-column schema: date, title, venue/source, tags, resources. Use existing tags from the vocabulary.
  4. Preserve publication uncertainty: a submission, preprint, repository claim, and verified conference acceptance are distinct statuses.
  5. Add a note under papers/arxiv/YYYY-MM/ or papers/conference/VENUE/. Use the exact headings Why It Matters, Proactivity Signal, Evaluation Setup, Key Limitations, and Use For. Ground results in primary sources and state evaluation limits.
  6. Link the note from the paper's README row using [Notes](papers/.../note.md) or the existing Notes badge format.
  7. Update RESEARCH_MAP.md, BENCHMARKS.md, or STREAMING.md when the work adds a useful research route or evaluation target.

For a benchmark, keep all nine columns in BENCHMARKS.md. Do not compare scores across incompatible datasets, hardware, protocols, or timing boundaries.

Add a project

Use PROJECTS.md. Record the proactive behavior or supporting role, activation mechanism, observed context, release/access status, evidence boundaries, canonical links, and checked date. For assistants, describe the behavior they choose; for components, describe how they support proactive systems. Keep these resource types explicit.

Place a resource in the collection

Use Applications and Personal Agents for selective reading routes. A paper keeps its canonical metadata in the README; projects keep implementation/access facts in PROJECTS; streaming mechanisms stay in STREAMING; benchmark protocols stay in BENCHMARKS. Link between these views instead of copying full records.

Record application, device/modality, decision mechanism, and resource type separately. Do not infer personalization from a camera interface or learned proactivity from a scheduler. Include supporting components through Infrastructure with their specific role. A model may support personal interaction without being a complete personal assistant. The restructuring design defines the scope and staged data plan.

Verify

npm ci
npm run check
git diff --check

Preview website changes with npm run dev. Verify desktop and mobile reading layouts and relevant interactions. See WEBSITE.md for architecture and deployment.