Hacker GitHub

← All repos

vishal-dehurdle/hypersae

View on GitHub ↗ ★ 13 Python

High-Performance Hyperbolic Sparse Autoencoders for Mechanistic Interpretability

5 mentions on Hacker News by 1 person: 5 stories, 0 comments · first 2026-08-11 · latest 2026-08-19

Mentions per year 5 mentions · busiest year 2026 (5)
20082026
AllStoriesComments NewestTopOldest
  1. ▲ 6
    Show HN: HyperSAE – Sparse Autoencoders, reimagined in hyperbolic space
    story by visha1v · 2026-08-19 · 0 comments
  2. ▲ 1
    Show HN: See what an LLM learned, organised as a browsable tree
    story by visha1v · 2026-08-18 · 0 comments
  3. ▲ 3
    Show HN: LLMs think in hierarchies. Their interpretability tools should too
    story by visha1v · 2026-08-15 · 0 comments
  4. ▲ 10
    Show HN: At 16K features, flat autoencoders break. Curved space doesn't
    story by visha1v · 2026-08-13 · 0 comments
  5. ▲ 3
    Show HN: HyperSAE – Hyperbolic Sparse Autoencoders for LLM Interpretability
    story by visha1v · 2026-08-11 · 0 comments