Anton Korznikov

According to our database, Anton Korznikov authored at least 5 papers between 2025 and 2026.

Collaborative distances:

• Dijkstra number of five.
• Erdős number of five.

2026

Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?

Anton Korznikov, Andrey V. Galichin, Alexey Dontsov, Oleg Rogov, Ivan V. Oseledets, Elena Tutubalina

CoRR, February, 2026

Feature Drift: How Fine-Tuning Repurposes Representations in LLMs.

Andrey V. Galichin, Anton Korznikov, Alexey Dontsov, Oleg Rogov, Elena Tutubalina, Ivan V. Oseledets

Proceedings of the Findings of the Association for Computational Linguistics: EACL 2026, 2026

Out of Distribution, Out of Luck: Process Rewards Misguide Reasoning Models.

Alexey Dontsov, Anton Korznikov, Andrey V. Galichin, Elena Tutubalina

Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics, 2026

2025

The Rogue Scalpel: Activation Steering Compromises LLM Safety.

Anton Korznikov, Andrey V. Galichin, Alexey Dontsov, Oleg Y. Rogov, Ivan V. Oseledets, Elena Tutubalina

CoRR, September, 2025

OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features.

Anton Korznikov, Andrey V. Galichin, Alexey Dontsov, Oleg Rogov, Elena Tutubalina, Ivan V. Oseledets

CoRR, September, 2025