Value of Information

Follow this topic Ask about this topic

Value of information (VOI) asks whether an observation, question, experiment, or computation is likely to improve a decision enough to justify its cost. Bayesian decision theory measures expected utility gains from sample or perfect information, while related methods use entropy reduction, regret, or risk bounds to guide experimental design, POMDP exploration, diagnosis, and clarification. Approaches range from dynamic programming and expected value of sample information calculations to Monte Carlo estimation, learned question policies, and certified rules for allocating sensing or computation. Across settings such as robotics, health sensing, retrieval-augmented generation, and language-agent interaction, a central challenge is that reducing uncertainty is not automatically useful: information must change actions for the better, given its costs and risks.

Latest papers

37 papers8 papers in the last four weeks, level with the four weeks before. 0.0% of all new papers.

Not All Information Is Worth Asking For: Decision-Informed Information Seeking with LLMs
Oct 23, 2026Seunghan Lee, Jun Seo, Junhyeok Kang +9LLM Decision-Making Value of Information
LG AI Research
When Is a Memory Safe Enough to Act On? Action-Conditioned Verification for Long-Lived Agents
Oct 22, 2026Yiqi Wang, Yifei Wang, Jiaqi Zhang +4LLM Agent Verification LLM Agent Safety
University of Southern Queensland, Australia · Southern University of Science and Technology, Shenzhen, China · Jiangsu University, China +2
CrashGym: Measuring the Decision Value of Experimentation for LLM Agents in Counterfactual Market Crashes
Oct 21, 2026Haocheng Lu, Haoxuan Cheng, Mingyang Chen +1Value of Information Autonomous Experimentation
Columbia University · Cornell University · Hithink RoyalFlush Information Network Co., Ltd. +1
When to Ask for Help: A Decision-Theoretic Account and Measurement Standard for Slow-Module Scheduling in Dual-System Vision-Language-Action Models
Oct 15, 2026Shilin ZhangVision-Language-Action Models Efficient VLA Model Inference
North China University of Technology
On Open-Ended Information Seeking for Information Elicitation Agents
Oct 5, 2026Victor De Lima, Grace Hui YangValue of Information Output Elicitation
InfoSense Lab, Georgetown University
Learning to Clarify Underspecified Intents Under Limited Interaction
Oct 3, 2026Pranav M R, Manuel Cherep, Pattie Maes +1Human-AI Interaction Clarification Question Generation
BITS Pilani & MIT · MIT · Dartmouth College
Should I stay or should I show? Learning to selectively disclose information
Sep 30, 2026Carlotta Giacchetta, Alessando Bogani, Cesare Barbera +4Value of Information Human-AI Decision Making
University of Trento, Trento, Italy · University of Pisa, Pisa, Italy · ETH AI Center & ETH, Zürich, Switzerland +1
Rational Clarification by Assistive Agents via Value-of-Information Reasoning
Sep 29, 2026T. Duy Nguyen-Hien, Yee Whye Teh, Wee Sun Lee +1Value of Information Clarification Question Generation
Department of Computer Science, National University of Singapore · Department of Statistics, University of Oxford · Agency for Science, Technology and Research (A*STAR)
CruxBench: A Benchmark of Information Discovery
Sep 26, 2026Hui Dai, Lina Piao, Nick Merrill +3LLM Evaluation Value of Information
The University of Chicago · Forecasting Research Institute · Federal Reserve Bank of Chicago
LAVOIR: Teaching a Single-Pass Decision Encoder When and What to Ask with Amortized Value of Information
Sep 25, 2026Furkan Yilmaz, Habibe Aleyna Tasdemir, Muhammed Faruk GozayLLM Decision-Making Active Information Gathering
Knowing When to Yield: Grounded Arbitration of User Corrections in Text-Based Embodied Agents
Sep 25, 2026Yezhou Cheng, Runjia Du, Zeming Liu +3Embodied AI Human-AI Agreement
University of Wisconsin - Madison · Independent Researcher · Brown University +1
When Does Test-Time Physical Diagnosis Pay?
Sep 14, 2026Zhengshu ZhangTest-Time Adaptation Robot Policy Adaptation
University of Southern California
Quantifying the Value of Privileged Information Using a PAC-Bayesian Approach
Sep 11, 2026Vasily Bokov, Sebastian Schmitt, Vedran Dunjko +1Value of Information Learning Using Privileged Information
aQa, Leiden University, The Netherlands · LIACS, Leiden University, Leiden, The Netherlands · Honda Research Institute Europe GmbH, Offenbach, Germany +1
A Mathematical Theory of Pragmatic Information
Sep 10, 2026Kai Niu, Ping ZhangRate-Distortion Theory Value of Information
Key Laboratory of Universal Wireless Communications, Ministry of Education · State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, 100876, China
When Information is Worth the Risk: Behavioral Valuation for Hazardous Robotic Exploration
Sep 9, 2026Alkesh K. Srivastava, Aamodh Suresh, Carlos Nieto-Granda +1Value of Information Informative Path Planning
Department of Mechanical Engineering, Temple University, Philadelphia, PA 19122, USA · U.S. DEVCOM Army Research Laboratory, Adelphi, MD 20783, USA · Lobster Robotics, The Hague, South Holland, Netherlands
HB-PVI: A Hierarchical Bayesian Personalization and Value-of-Information Framework for Complex Activity Recognition
Sep 4, 2026Hammed A. OlayinkaCost-Aware Inference Human Activity Recognition
Department of Mathematical Sciences, Worcester Polytechnic Institute, Worcester, MA 01609 USA
Pandora's AI Model Routing Box: Efficient Allocation with Costly Value Estimation
Aug 20, 2026Adam Fisch, Shubhendu Trivedi, Fantine Huot +5LLM Routing Cost-Aware Inference
Google DeepMind
Thermodynamics of Learning: A Typed Four-Component Accounting of Memory, Fit, and Value
Aug 13, 2026Akihito SudoStatistical Physics of Learning Distribution Shift
ZeroStruct Inc.
Bayesian Expected Uncertainty Reduction (B-EUR) Model: A Computational Account of What Makes Design Options Worth Trying
Aug 6, 2026Shimon Honda, Takuma Miyaguchi, Koji Koizumi +3Computational Cognitive Modeling Value of Information
Department of Mechanical Engineering, Graduate School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan · Laboratoire Conception de Produits et Innovation (LCPI), Arts et M´etiers ParisTech, 151 Boulevard de l’Hˆopital, 75013 Paris, France
Uncertainty Is Not Enough: Value-of-Information Routing for Mixtures of LoRA Experts
Aug 3, 2026Tom Saliencro, Rohan Desai, Priya Nair +2Adaptive Model Routing Selective Prediction
University of California, Irvine · University of Washington
Expected Free Energy as Belief-Dependent Utility for rho-POMDPs
Jul 18, 2026Patrick Cooper, Alvaro VelasquezFree Energy Calculation Active Inference
Department of Computer Science, University of Colorado Boulder, Boulder, CO, USA
Active Noise Floor Estimation for Reliability-Optimal POMDPs: A Value-of-Noise-Information Approach
Jul 13, 2026Hyung-Jin YoonValue of Information Active Sensing
Department of Mechanical and Nuclear Engineering, Tennessee Technological University, Cookeville, TN 37135, USA
Information Limits and Attractor Dynamics in Economies of Frontier LLM Agents: A Pre-Registered Test
Jul 7, 2026Cheng QianMulti-Agent LLM Systems Value of Information
Independent Researcher
How Much Due Diligence Before You Bid? Learning in Intractable Takeover Auctions
Jun 28, 2026Zain NaboulsiSelf-Play RL Imperfect-Information Games
Principal AI Engineer, Sparq
Value of Information under Imprecise Probabilities: Decision-Rule-Specific Values and Fixed-Measure Envelopes on a Credal Set
Jun 26, 2026Rowan IskandarValue of Information Imprecise Probability
Medtronic Trading Sàrl, Tolochenaz, VD, Switzerland · Center for Evidence Synthesis in Health, Brown University School of Public Health, Providence, RI, USA
Information-Directed Offline-to-Online Reinforcement Learning
May 28, 2026Keru ChenRegret Minimization in RL Bayesian RL
School of Electrical, Computer and Energy Engineering, Arizona State University
Delightful Exploration
May 13, 2026Ian OsbandMulti-Armed Bandits Value of Information
Google DeepMind
NH-CROP: Robust Pricing for Governed Language Data Assets under Cost Uncertainty
May 3, 2026Xu Zheng, Feiyu Wu, Zhuocheng Wang +2Value of Information Decision-Making under Uncertainty
School of Cyber Engineering, Xidian University
Extrapolating Volition with Recursive Information Markets
Apr 8, 2026Abhimanyu Pallavi Sudhir, Long Tran-ThanhValue of Information
University of Warwick, Coventry, United Kingdom
All Substitution Is Local
Apr 1, 2026Nidhish Shah, Shaurjya Mandal, Asfandyar AzharValue of Information Bayesian Decision Theory
Harvard University · University of Oxford

Oct 23, 2026J/K move · Enter open · S save

Not All Information Is Worth Asking For: Decision-Informed Information Seeking with LLMs

Seunghan Lee, Jun Seo, Junhyeok Kang, Jaehoon Lee+8

LG AI Research

LLM Decision-Making Value of Information

Open paper AskPDF arXiv

Large language models (LLMs) are increasingly used to support decisions under incomplete information, where deciding what to look up and when to stop matters as much as the final answer. However, LLMs fall into an information trap: They pay for checks, i.e., costly pieces of evidence, that cannot change their decision or are not worth their cost, and they keep asking when the evidence is already sufficient. To this end, we propose Decision-Informed Value-of-information Acquisition (DIVA), a simple yet effective procedure for LLM information seeking based on the decision value of evidence. DIVA 1) lets the LLM specify a small decision model, i.e., how likely a bad outcome is and how each check would shift this probability, through short numeric questions and 2) values each candidate check by how much its imagined outcomes improve the decision net of its cost, stopping when no check is worth its cost. Its variant, DIVA-U, further accounts for the LLM's uncertain estimates of how strongly each check shifts this probability. DIVA is training-free and applies to any LLM. Across four domains and 20 LLMs, DIVA-U has lower regret, i.e., value lost relative to the best possible decision, than every LLM baseline on real-world data, mainly by stopping once evidence suffices, and DIVA cuts the regret of the best LLM baseline by 75% in synthetic scenarios where every check's reliability is given. Code: github.com/seunghan96/diva.