Story
arxiv_agent_systems_research ยท Nov 13, 2026 ยท paper
Source brief
Do Verifier Judgments Survive Abstraction? Measuring Verification Transport for LLM Agent Evaluation
arxiv.orgNov 13, 2026
original source linked
In brief
LLM-based verifiers are increasingly used to evaluate agent executions, but their reliability is typically measured under a single representation of the trajectory. In deployed agent systems, the same execution may be...
Feed lens
agentevaluation