Transformer Circuits
Anthropic interpretability research articles.
Read from its RSS feedAppears in:AI
Latest from Transformer Circuits
- Characterizing interference weights in a tiny language model2026-08-21
- Verbalizable Representations Form a Global Workspace in Language Models2026-07-06
- Circuits Updates – June 20262026-07-02
- Circuits Updates – May 20262026-06-02
- Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations2026-05-07
- HeadVis: An Interactive Tool For Investigating Attention Heads2026-05-05
- Emotion Concepts and their Function in a Large Language Model2026-04-02
- Circuits Updates – November 20252025-11-26
- Emergent Introspective Awareness in Large Language Models2025-10-29
- Circuits Updates – October 20252025-10-25
- When Models Manipulate Manifolds: The Geometry of a Counting Task2025-10-21
- Circuits Updates - September 20252025-09-30
- Circuits Updates - August 20252025-08-28
- A Toy Model of Mechanistic (Un)Faithfulness2025-08-07
- Tracing Attention Computation Through Feature Interactions2025-07-31
- A Toy Model of Interference Weights2025-07-29
- Sparse mixtures of linear transforms2025-07-25
- Circuits Updates - July 20252025-07-25
- Circuits Updates - April 20252025-04-30
- Attribution Graphs2025-03-27
- On the Biology of a Large Language Model2025-03-27
- Circuit Tracing: Revealing Computational Graphs in Language Models2025-03-27
- Insights on Crosscoder Model Diffing2025-02-20
- Circuits Updates - January 20252025-01-28
- Stage-Wise Model Diffing2024-12-12
- Sparse Crosscoders for Cross-Layer Features and Model Diffing2024-10-25
- Using Dictionary Learning Features as Classifiers2024-10-16
- Circuits Updates - September 20242024-10-01
- Circuits Updates - August 20242024-09-06
- Circuits Updates - July 20242024-07-31