[2006.00995] Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs. arXiv Operational Status Get status notifications via email or slack

[2006.00995] Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals Skip to main content arXiv is now an independent nonprofit! Learn more × Search arXiv Press Enter to search · Advanced search --> Computer Science > Computation and Language arXiv:2006.00995 (cs) [Submitted on 1 Jun 2020 ( v1 ), last revised 19 Feb 2021 (this version, v3)] Title: Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals Authors: Yanai Elazar , Shauli Ravfogel , Alon Jacovi , Yoav Goldberg View a PDF of the paper titled Amnesic Probing: Behavioral Explanation with Amnesic C

related reading

• Structure and Interpretation of Deep Networkssidn.baulab.info
• A Pragmatic Vision for Interpretability — AI Alignment Forumalignmentforum.org
• How well do truth probes generalise? — LessWronglesswrong.com
• Faithful, Interpretable Model Explanations via Causal Abstraction | SAIL Blogai.stanford.edu
• Language models can explain neurons in language modelsopenaipublic.blob.core.windows.net
• A Pragmatic Vision for Interpretability — LessWronglesswrong.com
• Causal Scrubbing: a method for rigorously testing interpretability hypotheses [Redwood Research] — LessWronglesswrong.com
• Topicslearnmechinterp.com
• Chapter 1: Transformer Interpretability - ARENAlearn.arena.education
• [2512.11949] Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitorsarxiv.org
• Features as Rewards: Using Interpretability to Reduce Hallucinationsgoodfire.ai
• Discovering Language Model Behaviors with Model-Written Evaluations — LessWronglesswrong.com