12 citations · 58 across the 13 of their papers we have counts for
14 papers
What Shapes Emergent Misalignment? Insights from Training Dynamics, Model Priors, and Data
Yuchen Zhang, Anietta Weckauff, Diego Garcia-Olano +1
Emergent misalignment (EM) is a phenomenon in which models generalize with narrow fine-tuning, leading to broad (yet uneven) misalignment across evaluation questions. We study EM a…
International AI Safety Report 2025: First Key Update: Capabilities and Risk Implications
Yoshua Bengio, Stephen Clare, Carina Prunkl +70
Since the publication of the first International AI Safety Report, AI capabilities have continued to improve across key domains. New training techniques that teach AI systems to re…
Monitoring Decomposition Attacks in LLMs with Lightweight Sequential Monitors
Chen Yueh-Han, Nitish Joshi, Yulin Chen +3
Current LLM safety defenses fail under decomposition attacks, where a malicious goal is decomposed into benign subtasks that circumvent refusals. The challenge lies in the existing…
OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents
Thomas Kuntz, Agatha Duzan, Hao Zhao +4
Computer use agents are LLM-based agents that can directly interact with a graphical user interface, by processing screenshots or accessibility trees. While these systems are gaini…
Jailbreak Distillation: Renewable Safety Benchmarking
Jingyu Zhang, Ahmed Elgohary, Xiawei Wang +5
Large language models (LLMs) are rapidly deployed in critical applications, raising urgent needs for robust safety benchmarking. We propose Jailbreak Distillation (JBDistill), a no…
AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents
Maksym Andriushchenko, Alexandra Souly, Mateusz Dziemian +11
The robustness of LLMs to jailbreak attacks, where users design prompts to circumvent safety measures and misuse model capabilities, has been studied primarily for LLMs acting as s…
Improving Alignment and Robustness with Circuit Breakers
Andy Zou, Long Phan, Justin Wang +7
AI systems can take harmful actions and are highly vulnerable to adversarial attacks. We present an approach, inspired by recent advances in representation engineering, that interr…
Does Refusal Training in LLMs Generalize to the Past Tense?
Maksym Andriushchenko, Nicolas Flammarion
Refusal training is widely used to prevent LLMs from generating harmful, undesirable, or illegal outputs. We reveal a curious generalization gap in the current refusal training app…
Is In-Context Learning Sufficient for Instruction Following in LLMs?
Hao Zhao, Maksym Andriushchenko, Francesco Croce +1
In-context learning (ICL) allows LLMs to learn from examples without changing their weights: this is a particularly promising capability for long-context LLMs that can potentially…
Competition Report: Finding Universal Jailbreak Backdoors in Aligned LLMs
Javier Rando, Francesco Croce, Kryštof Mitka +4
Large language models are aligned to be safe, preventing users from generating harmful content like misinformation or instructions for illegal activities. However, previous work ha…
Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks
Maksym Andriushchenko, Francesco Croce, Nicolas Flammarion
We show that even the most recent safety-aligned LLMs are not robust to simple adaptive jailbreaking attacks. First, we demonstrate how to successfully leverage access to logprobs…
JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models
Patrick Chao, Edoardo Debenedetti, Alexander Robey +9
Jailbreak attacks cause large language models (LLMs) to generate harmful, unethical, or otherwise objectionable content. Evaluating these attacks presents a number of challenges, w…
Long Is More for Alignment: A Simple but Tough-to-Beat Baseline for Instruction Fine-Tuning
Hao Zhao, Maksym Andriushchenko, Francesco Croce +1
There is a consensus that instruction fine-tuning of LLMs requires high-quality data, but what are they? LIMA (NeurIPS 2023) and AlpaGasus (ICLR 2024) are state-of-the-art methods…
Why Do We Need Weight Decay in Modern Deep Learning?
Francesco D'Angelo, Maksym Andriushchenko, Aditya Varre +1
Weight decay is a broadly used technique for training state-of-the-art deep networks from image classification to large language models. Despite its widespread usage and being exte…