analiza aiqbench
This research introduces REVOIR, a novel framework that optimizes agent clarification by calculating the value-of-information for questions rather than simply minimizing uncertainty. By balancing the cost of asking versus acting and leveraging potential user corrections, the method significantly improves task success rates while reducing unnecessary queries in assistive applications. The approach demonstrates that adaptive reasoning can outperform standard prompting, fine-tuning, or information-gain strategies without requiring additional training data.
Rational Clarification by Assistive Agents via Value-of-Information Reasoning
arXiv:2609.37588v1 Announce Type: cross Abstract: Users of language-based assistive agents often make ambiguous requests. In response, an assistant can either directly act on its interpretation of the request --- risking misalignment with the user --- or ask a clarifying question. Which option is the most safe and helpful? A common approach is to ask questions that minimize uncertainty about the user's intent until a threshold is reached. However, this neglects the impact of uncertainty reduction on downstream performance, the costs of asking versus acting immediately, and the possibility that users may provide corrections without being asked. To navigate these trade-offs, we introduce Rational Enquiry via Value-of-Information Reasoning (REVOIR). REVOIR makes clarification decisions via inference-time reasoning about the value-of-information of a question, which captures the expected improvement in task reward due to the answer received. In two assistive tasks --- ambiguous question answering (CondAmbigQA) and preference-aligned household task planning (ADAPT) --- we show that REVOIR achieves greater success with fewer questions than approaches based on prompting, chain-of-thought, fine-tuning, or information gain, improving preference satisfaction on ADAPT by 13-15% over a fine-tuned clarification policy while requiring no training and asking five times fewer questions. Furthermore, when the assistant can receive cheap user corrections after acting, REVOIR naturally infers that asking questions is not always efficient, demonstrating the adaptivity of our approach. In contrast, we find that vanilla reasoning agents fail to adaptively clarify user requests, and request fewer clarifications as reasoning effort increases.
Abstract:Users of language-based assistive agents often make ambiguous requests. In response, an assistant can either directly act on its interpretation of the request --- risking misalignment with the user --- or ask a clarifying question. Which option is the most safe and helpful? A common approach is to ask questions that minimize uncertainty about the user's intent until a threshold is reached. However, this neglects the impact of uncertainty reduction on downstream performance, the costs of asking versus acting immediately, and the possibility that users may provide corrections without being asked. To navigate these trade-offs, we introduce Rational Enquiry via Value-of-Information Reasoning (REVOIR). REVOIR makes clarification decisions via inference-time reasoning about the value-of-information of a question, which captures the expected improvement in task reward due to the answer received. In two assistive tasks --- ambiguous question answering (CondAmbigQA) and preference-aligned household task planning (ADAPT) --- we show that REVOIR achieves greater success with fewer questions than approaches based on prompting, chain-of-thought, fine-tuning, or information gain, improving preference satisfaction on ADAPT by 13-15% over a fine-tuned clarification policy while requiring no training and asking five times fewer questions. Furthermore, when the assistant can receive cheap user corrections after acting, REVOIR naturally infers that asking questions is not always efficient, demonstrating the adaptivity of our approach. In contrast, we find that vanilla reasoning agents fail to adaptively clarify user requests, and request fewer clarifications as reasoning effort increases.
54 pages, 11 figures. Under review
Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Multiagent Systems (cs.MA)
arXiv:2609.37588 [cs.AI]
(or arXiv:2609.37588v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.37588 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 346 cuvinte
Subiecte conexe
A Candid Abacus AI Review: The All-in-One AI Platform for Professionals & Enterprises
KDnuggets · 10 sept. 2026
Hint-based CoT faithfulness evals still mostly work on Claude
LessWrong · 30 iul. 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026
Game-Theoretic Disempowerment: Loss of Control to Agencyless AI
LessWrong · 27 sept. 2026