Story
arxiv_agent_systems_research · Nov 9, 2026 · paper
arxiv.orgNov 9, 2026
original source linked
In brief
As large language models become more capable agents, they can exhibit deceptive behaviors, cheating human for higher rewards while appearing helpful. We attribute this failure to evaluation-side partial observability, where models are evaluated mainly from final outcome, without
Feed lens
agenticevaluation