yzxoi/paperly-codex/arxiv-monitor

arXiv Monitor

Overview

Monitors arXiv for new papers matching a team's research interests. Runs on a schedule or on-demand, scores relevance, and surfaces the most important new work directly to the team — replacing manual daily browsing.

When to Use

• User asks "what's new on arXiv this week in [topic]?"
• User wants to set up automatic monitoring for a topic or author
• User asks "did anyone publish anything on X lately?"
• Lab wants a weekly digest of relevant preprints
• User wants to know if a competitor lab has posted new work

Key Capabilities

• Query arXiv API by category (cs.LG, q-bio, stat.ML, etc.), keyword, or author
• Score relevance against a team research profile or keyword list
• Deduplicate against papers already in the team's knowledge base
• Generate a ranked digest with one-line summaries per paper
• Watch specific author IDs for new submissions
• Alert on papers that cite a specific prior work

Usage Examples

Get this week's relevant papers

monitor.fetch_recent(
    categories=["cs.LG", "cs.AI", "q-bio.QM"],
    keywords=["protein language model", "few-shot learning", "benchmark"],
    days=7,
    top_k=15
)

Set up a recurring watch

monitor.watch(
    query="contrastive learning protein structure",
    author_ids=["0000-0002-1234-5678"],
    frequency="weekly",
    notify_channel="slack:#lab-papers"
)

Check if a topic has seen recent activity

monitor.activity_pulse(topic="diffusion models for molecular design", months=3)

Output Format

Returns ranked list of papers with: title, authors, arXiv ID, abstract summary (2 sentences), relevance score, and suggested tags. Optionally formats as a digest email or Slack message.

Best Paired With

• related_work_scout when the user asks for the latest papers or a recency check

Notes

• Relevance scoring uses keyword overlap plus lightweight semantic similarity against the query profile
• Rate-limited to arXiv API fair-use guidelines (3 requests/second max)