Towards Agents That Know When They Don't Know: Uncertainty as a Control Signal for Structured Reasoning

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Stoisser, Josefa Lia, Martell, Marc Boubnovski, Phillips, Lawrence, Mazzoni, Gianluca, Harder, Lea Mørch, Torr, Philip, Ferkinghoff-Borg, Jesper, Martens, Kaspar, Fauqueur, Julien
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908515649454080
author Stoisser, Josefa Lia
Martell, Marc Boubnovski
Phillips, Lawrence
Mazzoni, Gianluca
Harder, Lea Mørch
Torr, Philip
Ferkinghoff-Borg, Jesper
Martens, Kaspar
Fauqueur, Julien
author_facet Stoisser, Josefa Lia
Martell, Marc Boubnovski
Phillips, Lawrence
Mazzoni, Gianluca
Harder, Lea Mørch
Torr, Philip
Ferkinghoff-Borg, Jesper
Martens, Kaspar
Fauqueur, Julien
contents Large language model (LLM) agents are increasingly deployed in structured biomedical data environments, yet they often produce fluent but overconfident outputs when reasoning over complex multi-table data. We introduce an uncertainty-aware agent for query-conditioned multi-table summarization that leverages two complementary signals: (i) retrieval uncertainty--entropy over multiple table-selection rollouts--and (ii) summary uncertainty--combining self-consistency and perplexity. Summary uncertainty is incorporated into reinforcement learning (RL) with Group Relative Policy Optimization (GRPO), while both retrieval and summary uncertainty guide inference-time filtering and support the construction of higher-quality synthetic datasets. On multi-omics benchmarks, our approach improves factuality and calibration, nearly tripling correct and useful claims per summary (3.0\(\rightarrow\)8.4 internal; 3.6\(\rightarrow\)9.9 cancer multi-omics) and substantially improving downstream survival prediction (C-index 0.32\(\rightarrow\)0.63). These results demonstrate that uncertainty can serve as a control signal--enabling agents to abstain, communicate confidence, and become more reliable tools for complex structured-data environments.
format Preprint
id arxiv_https___arxiv_org_abs_2509_02401
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards Agents That Know When They Don't Know: Uncertainty as a Control Signal for Structured Reasoning
Stoisser, Josefa Lia
Martell, Marc Boubnovski
Phillips, Lawrence
Mazzoni, Gianluca
Harder, Lea Mørch
Torr, Philip
Ferkinghoff-Borg, Jesper
Martens, Kaspar
Fauqueur, Julien
Artificial Intelligence
Large language model (LLM) agents are increasingly deployed in structured biomedical data environments, yet they often produce fluent but overconfident outputs when reasoning over complex multi-table data. We introduce an uncertainty-aware agent for query-conditioned multi-table summarization that leverages two complementary signals: (i) retrieval uncertainty--entropy over multiple table-selection rollouts--and (ii) summary uncertainty--combining self-consistency and perplexity. Summary uncertainty is incorporated into reinforcement learning (RL) with Group Relative Policy Optimization (GRPO), while both retrieval and summary uncertainty guide inference-time filtering and support the construction of higher-quality synthetic datasets. On multi-omics benchmarks, our approach improves factuality and calibration, nearly tripling correct and useful claims per summary (3.0\(\rightarrow\)8.4 internal; 3.6\(\rightarrow\)9.9 cancer multi-omics) and substantially improving downstream survival prediction (C-index 0.32\(\rightarrow\)0.63). These results demonstrate that uncertainty can serve as a control signal--enabling agents to abstain, communicate confidence, and become more reliable tools for complex structured-data environments.
title Towards Agents That Know When They Don't Know: Uncertainty as a Control Signal for Structured Reasoning
topic Artificial Intelligence
url https://arxiv.org/abs/2509.02401