Alasdair Paren

BiasBusters: Uncovering and Mitigating Tool Selection Bias in Large Language Models

FORCE: Transferable Visual Jailbreaking Attacks via Feature Over-Reliance CorrEction

OMNI-LEAK: Orchestrator Multi-Agent Network Induced Data Leakage

ToolTweak: An Attack on Tool Selection in LLM-based Agents

Focus On This, Not That! Steering LLMs with Adaptive Feature Specification

Shh, don't say that! Domain Certification in LLMs

MIP against Agent: Malicious Image Patches Hijacking Multimodal OS Agents

Focus On This, Not That! Steering LLMs With Adaptive Feature Specification

Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders

Questioning Simplicity Bias Assumptions