10 papers
Anatomy of Unlearning: The Dual Impact of Fact Salience and Model Fine-Tuning
Anna Borisiuk, Andrey Savchenko, Alexander Panchenko +1
Machine Unlearning (MU) enables Large Language Models (LLMs) to remove unsafe or outdated information. However, existing work assumes that all facts are equally forgettable and lar…
Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?
Anton Korznikov, Andrey Galichin, Alexey Dontsov +3
Sparse Autoencoders (SAEs) have emerged as a promising tool for interpreting neural networks by decomposing their activations into sparse sets of human-interpretable features. Rece…
The Rogue Scalpel: Activation Steering Compromises LLM Safety
Anton Korznikov, Andrey Galichin, Alexey Dontsov +3
Activation steering is a promising technique for controlling LLM behavior by adding semantically meaningful vectors directly into a model's hidden states during inference. It is of…
OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features
Anton Korznikov, Andrey Galichin, Alexey Dontsov +3
Sparse autoencoders (SAEs) are a technique for sparse decomposition of neural network activations into human-interpretable features. However, current SAEs suffer from feature absor…
HAMSA: Hijacking Aligned Compact Models via Stealthy Automation
Alexey Krylov, Iskander Vagizov, Dmitrii Korzh +6
Large Language Models (LLMs), especially their compact efficiency-oriented variants, remain susceptible to jailbreak attacks that can elicit harmful outputs despite extensive align…
I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders
Andrey Galichin, Alexey Dontsov, Polina Druzhinina +4
Recent LLMs like DeepSeek-R1 have demonstrated state-of-the-art performance by integrating deep thinking and complex reasoning during generation. However, the internal mechanisms b…