arXiv subjects

Gustavo Mercier

Publications and source records attributed to Gustavo Mercier.

2 recordsLinked to original sources

arXiv2026

Localizing RL-Induced Tool Use to a Single Crosscoder Feature

Fine-tuning through RL reshapes the internal representations of language models to enable agentic behaviors such as tool use, yet the mechanistic basis of these changes remains poorly understood. While RL substantially improves structured tool-call generation, it is unclear which features emerge, which are preserved, and whether identified features can be leveraged for retraining-free behavioral control. In this work, we show that $\textit{Dedicated Feature Crosscoders (DFC)}$ isolate a compact set of RL-specific features that mediate tool-calling capability in $\texttt{Qwen2.5-3B}$. Across a $48$-crosscoder hyperparameter sweep, encode-decode reconstruction improves the RL model's tool correctness by $+31.1 \pm {9.7}$ pp and passively transfers tool-calling ability to the frozen base model by $+6.8 \pm 5.0$ pp which we call a $\textit{capability spillover}$. Our findings show that DFC partitioning concentrates RL-introduced capability into a minimal, steerable feature set that enables runtime behavioral control of agentic LLMs.

cs.LG↗

arXiv2025

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.

cs.LG↗