Mechanistic Interpretability

Mechanistic interpretability research has been done in NLP before the term was invented, under other names. See Mechanistic? for important historical context.

• Bereska & Gavves 2024 - Mechanistic Interpretability for AI Safety - A Review
• Rai et al 2024 - A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models paper list: github
• Sharkey et al 2025 - Open Problems in Mechanistic Interpretability
• Lin et al 2025 - A Survey on Mechanistic Interpretability for Multi-Modal Foundation Models

See also the papers in BERTology, Neural Network Psychology, Probing Experiments, Transformers - Analysis and Interpretation.

• Wang et al 2022 - Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 Small
• Conmy et al 2023 - Towards Automated Circuit Discovery for Mechanistic Interpretability
• Geva et al 2023 - Dissecting Recall of Factual Associations in Auto-Regressive Language Models
• Hanna et al 2023 - How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model
• Rajaram et al 2024 - Automatic Discovery of Visual Circuits
• 2025 - The role of Mechanistic Interpretability in ’unveiling’ the emergent representations of Large Language Models (This paper was rejected from TMLR, see reviews)
• Ge et al 2024 - Automatically Identifying Local and Global Circuits with Linear Computation Graphs
• Nie et al 2025 - Mechanistic Understanding and Mitigation of Language Confusion in English-Centric Large Language Models
• Induction Heads
  • Olsson et al 2022 - In-context Learning and Induction Heads See the first figure for the description of an induction head.
  • Singh et al 2024 - What needs to go right for an induction head? A mechanistic study of in-context learning circuits and their formation See figure 1a for an induction head example.
  • Bakalova et al 2025 - Contextualize-then-Aggregate: Circuits for In-Context Learning in Gemma-2 2B
  • Minegishi et al 2025 - Beyond Induction Heads: In-Context Meta Learning Induces Multi-Phase Circuit Emergence
  • Wu et al 2025 - How Do Transformers Learn Variable Binding in Symbolic Programs?
• Explaining Neurons
  • Choi et al 2024 - Scaling Automatic Neuron Description They released a database of descriptions of every neuron inside Llama-3.1-8B-Instruct.

This section should maybe be moved into its own page.

See also Language Model - Steering

• Cunningham et al 2023 - Sparse Autoencoders Find Highly Interpretable Features in Language Models
• 2023 - Towards Monosemanticity: Decomposing Language Models With Dictionary Learning
• 2024 - Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet

• Research Threads and Blogs
  • Transformer Circuits Thread
  • Transluce's Research Thread
• Companies
  • Anthropic
  • Transluce

• Alignment
• BERTology
• Explainability
• LLM Safety
• Neural Network Psychology
• Probing Experiments
• Trustworthy AI
• Transformers - Analysis and Interpretation