Agent Planning and Evaluation
Series · 1 posts
-
Agent Planning Benchmark: Diagnosing Planning and Failure in LLM Agents
Advanced Agent runtime, safety, and evaluationA close reading of APB’s 4,209 cases and five planning settings: how it separates long-horizon plans, feedback-conditioned actions, tool noise, tool failure, and refusal—and what constructed tasks, LLM judging, and controlled execution checks leave unresolved.
Understand it in 90 seconds
For speaking invitations, internal engineering sessions, or architecture exchange, see the topics and public work I can bring into the conversation.
Speaking & contact