Alexey Dontsov

11 papers A* 1Journal 6Unranked 4

Year	Rank	Type	Title / Venue / Authors
2026	—	`conf`	Feature Drift: How Fine-Tuning Repurposes Representations in LLMs. EACL (Findings) Andrey V. Galichin, Anton Korznikov, Alexey Dontsov, Oleg Rogov, Elena Tutubalina, Ivan V. Oseledets
2026	A*	`conf`	I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders. AAAI Andrey V. Galichin, Alexey Dontsov, Polina Druzhinina, Anton Razzhigaev, Oleg Rogov, Elena Tutubalina, Ivan V. Oseledets
2026	—	`conf`	Out of Distribution, Out of Luck: Process Rewards Misguide Reasoning Models. EACL (Volume 2: Short Papers) Alexey Dontsov, Anton Korznikov, Andrey V. Galichin, Elena Tutubalina
2026	J	`jnl`	Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines? CoRR Anton Korznikov, Andrey V. Galichin, Alexey Dontsov, Oleg Rogov, Ivan V. Oseledets, Elena Tutubalina
2025	—	`conf`	CLEAR: Character Unlearning in Textual and Visual Modalities. ACL (Findings) Alexey Dontsov, Dmitrii Korzh, Alexey Zhavoronkin, Boris Mikheev, Denis Bobkov, Aibek Alanov, Oleg Rogov, Ivan V. Oseledets, Elena Tutubalina
2025	J	`jnl`	I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders. CoRR Andrey V. Galichin, Alexey Dontsov, Polina Druzhinina, Anton Razzhigaev, Oleg Y. Rogov, Elena Tutubalina, Ivan V. Oseledets
2025	J	`jnl`	Lightweight error mitigation strategies for post-training N:M activation sparsity in LLMs. CoRR Shirin Alanova, Kristina Kazistova, Ekaterina Galaeva, Alina Kostromina, Vladimir Smirnov, Dmitry Redko, Alexey Dontsov, Maxim Zhelnin, Evgeny Burnaev, Egor Shvetsov
2025	J	`jnl`	OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features. CoRR Anton Korznikov, Andrey V. Galichin, Alexey Dontsov, Oleg Rogov, Elena Tutubalina, Ivan V. Oseledets
2025	J	`jnl`	The Rogue Scalpel: Activation Steering Compromises LLM Safety. CoRR Anton Korznikov, Andrey V. Galichin, Alexey Dontsov, Oleg Y. Rogov, Ivan V. Oseledets, Elena Tutubalina
2024	—	`conf`	AIRI NLP Team at EHRSQL 2024 Shared Task: T5 and Logistic Regression to the Rescue. ClinicalNLP@NAACL Oleg Somov, Alexey Dontsov, Elena Tutubalina
2024	J	`jnl`	CLEAR: Character Unlearning in Textual and Visual Modalities. CoRR Alexey Dontsov, Dmitrii Korzh, Alexey Zhavoronkin, Boris Mikheev, Denis Bobkov, Aibek Alanov, Oleg Y. Rogov, Ivan V. Oseledets, Elena Tutubalina