Toy Models of Superposition

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Toy Models of Superposition

Why it matters

Demonstrated that neurons store more features than concepts in superposition; the theoretical starting point of the entire SAE family

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2209.10652

Added in Week 35, 2026 · See that week

On the map — ⑩ Safety, Alignment line

  1. Previous stopFedAvg
  2. This record · stopToy Models of Superposition
  3. Next stopDPO
See it on the map

Same week

OpenAI open-sources WhisperOfficial

Nearby on the same axis

DPOPaperOpenAI launches Superalignment teamOfficialResponsible Scaling Policy v1.0 (ASL system)OfficialWeak-to-strong generalizationOfficial