Don't Forget It! Conditional Sparse Autoencoder Clamping Works for Unlearning
Recent developments in Large Language Model (LLM) capabilities have brought great potential but also posed new risks. For example, LLMs with knowledge of bioweapons, advanced chemistry, or cyberattacks could cause violence if placed in the wrong hands or during malfunctions. Because of their nature as near-black boxes, intuitive interpretation of LLM internals remains an open research question, preventing developers from easily controlling model behavior and capabilities. The use of Sparse Autoencoders (SAEs) has recently emerged as a potential method of unraveling representations of concepts in LLMs internals, and has allowed developers to steer model outputs by directly modifying the hidden activations. In this paper, we use SAEs to identify unwanted concepts from the Weapons of Mass Destruction Proxy (WMDP) dataset within gemma-2-2b internals and use feature steering to reduce the model's ability to answer harmful questions while retaining its performance on harmless queries. Our results bring back optimism to the viability of SAE-based explicit knowledge unlearning techniques.
Code (0)
등록된 구현이 없습니다.
Tasks
Language ModelingLanguage ModellingLarge Language ModelSimilar Papers 제목 키워드 기반
Aligning Sentence Embeddings to Human Concepts via Sparse Autoencoders
Dense sentence embeddings are fundamental to modern Retrieval-Augmented Generation (RAG) systems but suffer from a lack of interpretability due to feature superposition. This opacity hinders the alignment of retrieval pr…
Information RetrievalMulti-Modal Mean-Fields via Cardinality-Based Clamping
Mean Field inference is central to statistical physics. It has attracted much interest in the Computer Vision community to efficiently solve problems expressible in terms of large Conditional Random Fields. However, sinc…
DescriptiveSAE-FD: Sparse Autoencoder Feature Distillation for Continual Learning of Large Language Models
Continual learning enables large language models to adapt to evolving tasks without retraining from scratch, yet catastrophic forgetting remains a central obstacle. Among continual learning methods, regularization-based …
Continual LearningSelective Amnesia: A Continual Learning Approach to Forgetting in Deep Generative Models
The recent proliferation of large-scale text-to-image models has led to growing concerns that such models may be misused to generate harmful, misleading, and inappropriate content. Motivated by this issue, we derive a te…
Continual LearningSparse Autoencoders as a Steering Basis for Phase Synchronization in Graph-Based CFD Surrogates
Graph-based surrogate models provide fast alternatives to high-fidelity CFD solvers, but their opaque latent spaces and limited controllability restrict use in safety-critical settings. A key failure mode in oscillatory …