Skip to content

michaelnoi/alphascent

Repository files navigation

AlphaScent

A figure-first interface for computer science research papers from arXiv, live at alphascent.org. Automatically extracted figures from PDFs are shown inline, optimized for rapid understanding and decision-making in exploratory research tasks.

About

AlphaScent displays arXiv papers with automatically extracted figures shown inline. All data comes from arXiv, and dates shown are submitted dates (not announcement dates).

Important Disclaimer: Papers displayed here may not be the most up-to-date versions. Always check the actual arXiv abstract/PDF view for the latest version. Note that many published papers don't update their arXiv submissions with camera-ready versions from conferences/journals, so always verify you're viewing the most recent version.

Categories

Currently, AlphaScent supports three categories for now:

Design Philosophy

  1. Speed over features: No LLM summaries, no loading states
  2. Visual hooks: Automatically extracted figures shown inline (for now, we display first and second figures from each paper, eventually we will smartly select teaser/architecture/results/etc. figures)
  3. Zero friction: One-click expansion, keyboard-first navigation
  4. Progressive disclosure: Essential info first, details on demand

Keyboard Shortcuts

Navigation

Key Action
j / Navigate to next paper
k / Navigate to previous paper
Space Expand/collapse paper details

Search

Key Action
F / Ctrl F Focus search bar
Enter Execute search (disables text input)
Esc Blur search bar

Open Links

Key Action
u Open abstract page
i Open PDF
o Open code repository (if available)
p Open project page (if available)
Shift + u/i/o/p Open in new tab (otherwise navigates in same tab)

Figures

Key Action
f View first figure (expanded view only)
s View second figure (expanded view only)

Content

Key Action
Shift + M Load more papers

Help

Key Action
? Show/hide keyboard shortcuts
Esc Close shortcuts dialog

Features

  • Visual-First Representation: Automatically extracted figures from PDFs
  • Simple Figure Selection: First two valid figures from each paper
  • Expandable Details: Quick overview with inline detailed view showing full abstracts and high-resolution figures
  • Keyboard Navigation: Power-user shortcuts for rapid browsing and link opening
  • Fast Performance: No loading spinners, Cloudflare-edge delivery
  • Full-Text Search: full-text search across titles, abstracts and authors. Can be toggled to search either:
    • All papers: Search across all available historical papers for the category
    • Current view: Search only within the currently selected date or date range
  • Date Filtering: Browse papers by single date or date range
  • Link Detection: Automatically extracts code repositories and project pages from paper comments

For development setup, deployment, and technical details, see SETUP.md.

TODO

Category Views

  • HC views
  • CV views
  • GR views
    • Full history beyond this year

Paper Updates

  • Daily updates

Figure Processing

  • Better figure extraction
  • Better figure selection

Acknowledgments

Thank you to arXiv for use of its open access interoperability.

About

figure-first list view of arXiv computer science papers

Resources

Stars

Watchers

Forks