SkillHarm: Lifecycle-Aware Skill-Based Attacks via Automated Construction Paper • 2606.02540 • Published Jun 1 • 10
QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks Paper • 2605.24218 • Published May 22 • 46
When Actions Go Off-Task: Detecting and Correcting Misaligned Actions in Computer-Use Agents Paper • 2602.08995 • Published Feb 9 • 2
Mind2Web 2: Evaluating Agentic Search with Agent-as-a-Judge Paper • 2506.21506 • Published Jun 26, 2025 • 52