Computation, Not Memory: Chain-of-Thought and State Tracking in LLM Agents

Key points

• We ask how an agent's actions depend on its reasoning when the correct action depends on a changing state.1
• In a small file-system task, background events announce which file's lock has been toggled but not its new value.1
• With Qwen3-8B and Qwen3.5-9B, we separate two roles of reasoning: generating it before each action, and retaining earlier reasoning in the prompt.1
• When we give the agent reasoning written for a history that differs in one lock value, the agent acts on that reasoning even against its own history.1

Sources (1)

• [1]Computation, Not Memory: Chain-of-Thought and State Tracking in LLM Agents
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 04:18 AM

  Language-model agents often reason before they act, and the value of monitoring that reasoning depends on how it relates to their actions.

  We ask how an agent's actions depend on its reasoning when the correct action depends on a changing state.

Extractive summary: sentences quoted from the sources.