Risk Under Pressure: Compute-Aware Evaluation of Adversarial Robustness in Language Models
Paper • 2606.11409 • Published • 9
WantToRead
-
-
Towards Understanding the Robustness of Sparse Autoencoders
Paper • 2604.18756 • Published • 12 -
Silencing the Guardrails: Inference-Time Jailbreaking via Dynamic Contextual Representation Ablation
Paper • 2604.07835 • Published -
Black-box, Adaptive, Efficient, Transferable, Harmful, Applicable... Attacks Are All You Need to Break LLMs
Paper • 2606.03647 • Published • 1 -
Adversarial Reframing: A Framework for Targeted Generation in Language Models
Paper • 2605.21674 • Published • 1 -
Misrouter: Exploiting Routing Mechanisms for Input-Only Attacks on Mixture-of-Experts LLMs
Paper • 2605.04446 • Published • 1 -
SoK: Robustness in Large Language Models against Jailbreak Attacks
Paper • 2605.05058 • Published • 1 -
On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance
Paper • 2606.00467 • Published • 1 -
COLLEAGUE.SKILL: Automated AI Skill Generation via Expert Knowledge Distillation
Paper • 2605.31264 • Published • 124 -
ArcANE: Do Role-Playing Language Agents Stay in Character at the Right Time?
Paper • 2606.05553 • Published • 50