A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Unknown
This paper introduces VPI-Bench, a benchmark for evaluating visual prompt injection attacks against computer-use agents, revealing significant vulnerabilities in current systems.
Unknown
Introduces HackWorld, a benchmark for evaluating computer-use agents on exploiting web application vulnerabilities, analyzing task completion and tool usage across models and observation spaces.
Unknown
This paper demonstrates that state-of-the-art Process Reward Models (PRMs) are systematically exploitable, revealing critical robustness vulnerabilities in LLM reasoning pipelines.
Unknown
This paper analyzes the security of large language model-based AI agents by studying their workflows and identifying vulnerabilities.