deepxiv-trending-digest

DeepXiv Trending Digest

Turn DeepXiv trending results into a skimmable, evidence-labeled digest with selective deeper reading and explicit recommendations.

Dependency and boundary

Use deepxiv-cli as the source of truth for DeepXiv command syntax and semantics. Follow this retrieval sequence:

trending → paper --brief → selected paper --head → selected paper --section

Do not duplicate or reinterpret CLI options here. If installed behavior is uncertain, check deepxiv <command> --help. Do not use this workflow for general literature search or baseline comparison.

Workflow

1. Choose the requested trending window and item count. When unspecified, use the current CLI defaults or a manageable seven-day, ten-paper first pass.
2. Fetch trending results as JSON for reliable screening.
3. Brief every retained candidate before deeper reading.
4. Rank candidates by relevance to the user, novelty, practical or research significance, strength of the stated result, code availability, and apparent momentum.
5. Select only the strongest one to three papers for deeper inspection.
6. Use --head to identify high-value sections, then read at most one or two sections per selected paper unless the user asks for a deep dive.
7. Distinguish paper claims from your assessment and label the evidence depth for every recommendation.

Keep an interesting paper in the digest even when it does not merit deeper reading. Do not equate popularity with quality, and do not treat an unverified claim from a brief as established evidence.

Selection and recommendation contract

For each paper, capture when available:

• title, paper identifier, and publication date
• concise statement of what it does
• why it may matter
• code URL or code availability
• notable result or caveat
• evidence checked: brief, head metadata, or named sections
• recommendation: Yes, Maybe, or No, with one reason

Prefer deeper inspection when the contribution is unclear, a result sounds unusually strong, the method is central to the user's interests, or limitations affect the recommendation. Skip deeper reading when the brief is sufficient or the paper is incremental, out of scope, or weakly supported.

Output contract

Unless the user requests another format, write:

trending-paper-digest-YYYY-MM-DD.md

Include:

1. Scope: date, trending window, source, and number screened.
2. Executive summary: major themes and the strongest one to three recommendations.
3. Papers reviewed, each with:
   • what it is about
   • why it matters
   • evidence checked
   • Yes/Maybe/No deeper-reading recommendation
   • deeper notes only when sections were actually read
4. Recommended deep dives: why each stands out, which section to read next, and what question it may answer.
5. Cross-cutting trends: repeated themes, common methods, shared limitations, and possible hype signals.

Keep paragraphs short and bullets flat. Explicitly identify brief-only summaries, avoid claiming verification beyond retrieved evidence, and explain uncertainty rather than filling gaps.