TheZvi on new interpretability work: useful, but far from actually understanding models
TheZvi · x · 2026-10-17
TheZvi offered a measured take on a widely discussed interpretability post: if it holds up, it solves some problems and makes life easier, but he isn't convinced it tackles the 'hard problems' — it doesn't attempt to explain attention or the residual stream, and we remain far from understanding how it all works. He also wishes someone would write a shorter, clearer version of the post.
More from Research
- Plotly and Mila launch DataAnalyticsBench: 43 tasks, 28 models on a 20M-row warehouse — Mila_Quebec · 2026-10-17
- Terry Tao's Four Types of Mathematization: Heavy AI Use Needs Strong Human Understanding — ravithejads · 2026-10-17
- Princeton's LabOS, Featured in Nature: An AI Co-Scientist for Hypothesis and Experiment Design — MengdiWang10 · 2026-10-17
- 'Tip-of-the-tongue' queries: a blind spot for modern agentic retrieval — bclavie · 2026-10-17
- Tip-of-the-tongue queries emerge as a key test of agentic retrieval systems — ctjlewis · 2026-10-17
- micro1 launches LegalReasonBench to test models on attorney-authored legal matters — douwekiela · 2026-10-17