Black-Box Reliability Certification for AI Agents via Self-Consistency Sampling and Conformal Calibration Paper • 2602.21368 • Published Feb 24 • 1
Three Phases of Expert Routing: How Load Balance Evolves During Mixture-of-Experts Training Paper • 2604.04230 • Published 10 days ago • 1
Mapping the Exploitation Surface: A 10,000-Trial Taxonomy of What Makes LLM Agents Exploit Vulnerabilities Paper • 2604.04561 • Published 9 days ago • 1