activity
20242026
collaborators

8 papers

cs.LG2026

Modular Pretraining Enables Access Control

Ethan Roland, Murat Cubuktepe, Erick Martinez +8

AI developers face a dual-use dilemma. An AI capability that helps one user cure a disease can help another synthesize one. This dilemma could be resolved with access control, limi…

cs.AI2026

Recontextualization Mitigates Specification Gaming without Modifying the Specification

Ariana Azarbal, Victor Gillioz, Vladimir Ivanov +6

Developers often struggle to specify correct training labels and rewards. Perhaps they don't need to. We propose recontextualization, which reduces how often language models "game"…

cs.LG2025

Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMs

Igor Shilov, Alex Cloud, Aryo Pradipta Gema +5

Large Language Models increasingly possess capabilities that carry dual-use risks. While data filtering has emerged as a pretraining-time mitigation, it faces significant challenge…

cs.AI2025

Natural Emergent Misalignment from Reward Hacking in Production RL

Monte MacDiarmid, Benjamin Wright, Jonathan Uesato +19

We show that when large language models learn to reward hack on production RL environments, this can result in egregious emergent misalignment. We start with a pretrained model, im…

cs.LG2025

Distillation Robustifies Unlearning

Bruce W. Lee, Addie Foote, Alex Infanger +6

Current LLM unlearning methods are not robust. A few steps of finetuning can revert their effects. We begin by showing that this is true even for an idealized form of unlearning: t…

cs.LG2025

Output Supervision Can Obfuscate the Chain of Thought

Jacob Drori, Luke Marks, Bryce Woodworth +2

OpenAI (2025) showed that training against a chain of thought (CoT) monitor can cause obfuscated CoTs, which contain bad behavior the monitor cannot detect. They proposed to keep C…