agents

Definition

Systems that pursue a goal by choosing actions, observing results, and adjusting — as opposed to answering a single prompt. In the LLM setting this usually means a loop around a model with tool calls, memory, and a stopping condition. Multi-agent work uses the plural for several such systems interacting, which raises coordination and evaluation problems a single agent does not have.

Explainers using this term

• Paper Deep Dive: From Evidence to Action: How Tool-Using Agents Fail — Reading the Failure Diagnostics and InterventionsFrom Evidence to Action: How Tool-Using Agents Fail
• SafeActBench explained (arXiv:2610.07753) — tool agents fail before executionFrom Evidence to Action: How Tool-Using Agents Fail
• Paper Explained: False Frontiers — Diagnosing and Breaking Co-Cheating in Self-Evolving Search AgentsFalse Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents