execution

Definition

Actually running generated code or a generated plan. For agents, execution output becomes the observation that drives the next action — the feedback loop that lets a model recover from its own mistakes. For evaluation, execution-based scoring asks whether code runs and passes tests rather than whether it looks right, which is a far stricter standard. It presupposes a sandbox, since generated code is untrusted.

Explainers using this term

• Paper Walkthrough: AutoSaddler — Growing a Harness That Doesn't Break, from Agent Failure LogsAutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces
• Paper Explained: FreeToken — Treating Your Own PC as a Single Elastic Inference PlatformFreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution
• CPU Pipelines and Branch Prediction — The Factory Inside One Clock TickSpectre Attacks: Exploiting Speculative Execution