Book
Open access
Jul 2026
AgentTelemetry: A Fault Detection Benchmark and Toolkit for LLM Agent Observability
AgentTelemetry, an open-source benchmark suite and toolkit for evaluating fault detection in agent systems, and an ablation study proves all nine span kinds are necessary: removing any one makes at least one fault type undetectable.
Krishna Chaitanya Balusu
· AIware · 2 citations