Paper Impact Analyzer#
Multi-source, fault-tolerant academic paper impact analysis.
When to Use#
- Evaluating a paper's academic influence or community adoption
- Comparing impact across multiple papers
- Deciding whether a paper is worth reading based on external signals
- Checking GitHub stars, citation counts, venue acceptance for a paper
- Assessing author credibility (h-index) for a paper
- Batch-analyzing papers in a survey or literature review
How to Use#
Single paper#
Run the analysis script with an arXiv ID:
bash
python scripts/analyze.py 2603.04948
Multiple papers#
Pass multiple arXiv IDs separated by spaces:
bash
python scripts/analyze.py 2603.04948 2602.15922 2603.05488 2602.22661
Output#
The script prints a structured Markdown impact report for each paper, including:
DimensionExamplePublication date2026-03-05 (20 days ago)Venue acceptanceICLR 2026GitHub repo2,263 stars / 214 forksCitation count12 (OpenAlex) / 15 (S2)Author h-indexFirst author h=23AffiliationsUC Berkeley, UT Austin
Plus a synthesized overall rating (S/A/B/C/D) with confidence level and data completeness.
Data Sources (Priority Order)#
- arXiv API — paper metadata, authors, abstract (always available)
- GitHub API — repo stars, forks, issues (most reliable external signal)
- OpenAlex API — citation count (free, no API key needed)
- Semantic Scholar API — citations, influential citations, author h-index (rate-limited)
Each source fails independently. The script always produces output using whatever data is available.
Design Philosophy#
- Graceful degradation: Every API call is wrapped in try/except with timeouts. If Semantic Scholar returns 429, the report still includes arXiv + GitHub + OpenAlex data.
- Age-aware scoring: Papers < 3 months old are scored primarily on GitHub + venue + team. Papers > 1 year old are scored primarily on citations.
- No API keys required: All data sources used are free and keyless.
- Single file: The entire implementation is in
scripts/analyze.py with zero external dependencies (stdlib only).