ArXiv Watcher

Research paper monitoring and summarization.

When to Use

• User wants to find recent papers on a topic
• User asks to monitor specific authors or subjects
• User needs paper summaries or trend analysis
• User maintains a research reading list

Procedure

1. Parse the user's research interest (topic, authors, keywords)
2. Query ArXiv API with appropriate search parameters
3. Filter by date range and relevance
4. Summarize each paper (title, authors, abstract, key contribution)
5. Score relevance to user's stated interests
6. Present results sorted by relevance

ArXiv API Usage

import urllib.request
import xml.etree.ElementTree as ET

base_url = 'http://export.arxiv.org/api/query?'
query = 'search_query=all:transformer+attention&sortBy=submittedDate&sortOrder=descending&max_results=10'
response = urllib.request.urlopen(base_url + query)
feed = ET.parse(response)

Output Format

For each paper:

• Title: [paper title]
• Authors: [author list]
• Submitted: [date]
• ArXiv ID: [id with link]
• Summary: 2-3 sentence summary of key contribution
• Relevance: High/Medium/Low

Monitoring (with Hermes cron)

Set up scheduled monitoring:

/cron daily at 8am: Check ArXiv for new papers on [topics] and summarize any relevant ones

Pitfalls

• ArXiv API has rate limits — max 1 request per 3 seconds
• Search results may include older revised papers
• Abstract summaries may miss nuanced contributions
• Some preprints are later withdrawn or significantly revised

Verification

• Paper links resolve to valid ArXiv pages
• Dates match the requested time range
• Relevance scoring matches user's stated interests