Bug in SAE paper's ICA baseline fixed after 3 years, making ICA highly competitive
TuhinChakr · x · 2026-10-22
Researchers found the ICA baseline in the original SAE paper had an evaluation bug. Once fixed, ICA becomes extremely competitive at finding interpretable LLM activation directions—taking just minutes per layer. The authors also released a video explaining why interpretable directions in LLMs look non-Gaussian, how ICA finds them, and how the bug kept ICA underrated for three years. A significant correction for interpretability research.
More from Research
- Looking Ahead: Predicting Future Chunks Independently Makes Video Generation More Physical — yshan2u · 2026-10-22
- Future Chunk Prediction: teaching autoregressive video models real physics — yshan2u · 2026-10-22
- Google's on-device AI matches expert sonographers in prenatal ultrasound study — ymatias · 2026-10-22
- Artios launches as a fully open-source system for de novo enzyme design — DeryaTR_ · 2026-10-22
- U of T launches project on culturally grounded AI for mental health care in Bangladesh — ishtiaqueSIA · 2026-10-22
- Obsession With AI May Be Starving the Research Fields It Needs to Survive — Symbiot10000 · 2026-10-22