Conference paper metadata, cleaned and flattened into JSON

What it does

Paperlists is the data engine behind Paper Copilot. It aggregates paper metadata from major ML and graphics conferences—ICLR, NeurIPS, ICML, SIGGRAPH, and others—into yearly JSON files. A small local search tool lets you query these records without external APIs.

The interesting bit

The entire value proposition is janitorial: maintainers reconcile scattered sources so each entry has coherent metadata, which also powers the statistical analyses on the Paper Copilot site. It is essentially glue code wrapped around public data, and the cleanup is the product.

Key highlights

• Yearly JSON files for ICLR, NeurIPS (2000–2024), ICML, SIGGRAPH, and SIGGRAPH Asia, reconciled from multiple sources.
• Local Streamlit search tool and CLI extractor for filtering by keyword, title, topic, and primary area.
• Designed to eliminate tab-hopping between OpenReview, official conference pages, and open-access repositories.
• Powers downstream statistics and paper lists on papercopilot.com.
• Runs fully offline using the bundled JSON files.

Caveats

• Coverage has visible gaps: ICLR 2015 and 2016 JSON files are missing from the tables, and the README truncates before finishing the SIGGRAPH Asia year list.
• The search tooling is minimal—keyword filtering over static JSON rather than advanced ranking or live updates.

Verdict

Grab these JSON files if you are building literature reviews, meta-analyses, or conference dashboards. Look elsewhere if you need full-text search, real-time alerts, or paper PDFs.