About
What is this?
Most Valuable Papers ranks recent AI research papers by citation count with time decay. This helps surface the most impactful papers while giving more weight to recent work.
How does ranking work?
Papers are ranked using a time-decayed citation score:
score = citations × e^(-λ × years_ago)
Where λ is calculated from the half-life you select. A 1-year half-life means papers from 1 year ago get 50% weight, 2 years ago get 25%, etc.
Data sources
- Semantic Scholar - Citation data
- arXiv - Paper metadata and categories
Why build this?
Existing paper discovery tools either rank by popularity (upvotes) or show papers chronologically. Citation-based ranking with time decay provides a better signal for finding important recent work.
Feedback
This is an early MVP. Have suggestions or found issues? Let me know.