CLEAT: Learning Value-Guided Clarification for Interactive Agents

Key points

• LLM agents can execute increasingly complex tasks, but effective assistance also requires aligning their actions with goals and constraints that users may leave unstated.1
• We study this problem as interactive intent alignment through clarification.1
• As training changes how the agent handles incomplete information and uses additional answers, the same clarification can become more or less useful.1
• We formulate residual query value to measure the task utility gained by asking rather than proceeding under the current policy, and estimate it through paired rollouts from the same interaction history.1

Sources (1)

• [1]CLEAT: Learning Value-Guided Clarification for Interactive Agents
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 07:22 PM

  We introduce CLEAT (CLarification through valuE-guided Agent co-Training), a framework that co-trains a clarification-value controller and the task policy from shared interaction outcomes.

  LLM agents can execute increasingly complex tasks, but effective assistance also requires aligning their actions with goals and constraints that users may leave unstated.

Extractive summary: sentences quoted from the sources.