Story
arxiv_agent_systems_research · Nov 4, 2026 · paper
arxiv.orgNov 4, 2026
original source linked
In brief
Efficient benchmarking estimates a system's full benchmark performance from a small number of task executions. Standard approaches rely solely on the binary correctness of executions from a collection of already-evalu...
Feed lens
agenticharnessevaluation