method · first seen 6 Oct 2026
causal ablation
Also: neuron ablation, causal neuron ablation, causal ablation of model components
Interpretability technique that measures a model component's causal contribution by removing or altering it and observing behavior changes.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Where Do LLMs Retrieve Network Information? Identifying the Responsible Heads and NeuronsPapers · 6 Oct 2026 · Md Abdul Kadir, Md Mohasin Hossain, Omair Shahzad Bhatti, Abdulrahman Mohamed Se…
- Neural networks are more modular than single neurons suggestPapers · 9 Oct 2026 · Shreya Saha, Benjamin Bergen, Meenakshi Khosla
- Language Models Have Interpretable Narrative MechanismsPapers · 9 Oct 2026 · David Yiyang Liu, Aditya Joshi, Paul Dawson, Andrew Piper
- Concentrated Causal Influence, Distributed Diagnostic Information: Cognitive Impairment Decisions in Large Audio-Language ModelsPapers · 23 Oct 2026 · Xiutian Zhao, Luqi Sun, Björn Schuller, Berrak Sisman
- Representational drift in a code model: behaviour settles, which neurons carry it does notPapers · 6 Nov 2026 · Piotr Wilam
Latest sources
- Representational drift in a code model: behaviour settles, which neurons carry it does notPapers · 6 Nov 2026 · Piotr Wilam
- Concentrated Causal Influence, Distributed Diagnostic Information: Cognitive Impairment Decisions in Large Audio-Language ModelsPapers · 23 Oct 2026 · Xiutian Zhao, Luqi Sun, Björn Schuller, Berrak Sisman
- Language Models Have Interpretable Narrative MechanismsPapers · 9 Oct 2026 · David Yiyang Liu, Aditya Joshi, Paul Dawson, Andrew Piper
- Neural networks are more modular than single neurons suggestPapers · 9 Oct 2026 · Shreya Saha, Benjamin Bergen, Meenakshi Khosla
- Where Do LLMs Retrieve Network Information? Identifying the Responsible Heads and NeuronsPapers · 6 Oct 2026 · Md Abdul Kadir, Md Mohasin Hossain, Omair Shahzad Bhatti, Abdulrahman Mohamed Se…