Anthropic Discovers "Conscious-Accessible" Neural Patterns Inside Claude
10b0t0mized · reddit · 2026-07-07
In the human brain, some activity is "consciously accessible"—we can describe, control, and use it for deliberate reasoning, distinct from unconscious automatic processing. A new Anthropic paper finds that a similar distinction has emerged in modern language models like Claude: Claude has developed a small cluster of internal neural patterns that play a special role compared to other internal processing. The paper is published on transformer-circuits.pub and accompanied by an interactive demo on Neuronpedia.
Related event: Anthropic Discovers Global Workspace Inside Claude(102 posts)→
More from Research
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Terence Tao's SAIR Foundation revamps Mathathon with $1.5M compute per team — _sathvikr · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27