StatefulDiscovery: Evidence-Calibrated Claim Formation in Open-Ended Scientific Discovery

2026-06-10

AI Summary

This work addresses the challenge of balancing phenomenon identification and evidential sufficiency in open-ended scientific discovery, where agents must avoid prematurely asserting claims lacking adequate support. To this end, the authors propose StatefulDiscovery, a novel framework that introduces an explicit discovery state mechanism for the first time. Through a state-externalization architecture, it cohesively integrates structured hypothesis representation, local claim adjudication, and frontier-guided exploration strategies, enabling dynamic coupling between exploration and assertion while ensuring rigorous evidence calibration. Evaluated on 40 real-world data-driven discovery tasks, the method significantly outperforms existing baselines, consistently generating a greater number of claims that exhibit both high scientific value and strong empirical support.

Abstract

Open-ended scientific discovery asks agents to move beyond executing analyses for predefined questions. Across multiple rounds of exploration, a discovery agent must decide which phenomena warrant investigation while avoiding overinterpretation, where emerging claims exceed the evidential scope of the analyses supporting them. This creates an evidence-calibration problem: the exploration trajectory must be coupled with claim status so that evidence can guide both what to investigate next and what can be claimed. We introduce StatefulDiscovery, a discovery framework that externalizes investigation state and uses it to coordinate frontier selection, evidence acquisition, and claim adjudication. We evaluate StatefulDiscovery across 40 real-data discovery tasks. Compared with several baselines, StatefulDiscovery produces more claims overall judged to be both well-supported and high-value. Ablations indicate that structured hypotheses, local adjudication, and frontier control contribute to performance. Together, these results suggest that explicit discovery state can couple exploration with evidence-calibrated claim formation.

Problem

• open-ended scientific discovery
• evidence calibration
• claim formation
• overinterpretation
• exploration trajectory

Innovation

• StatefulDiscovery
• evidence-calibrated claim formation
• open-ended scientific discovery
• investigation state
• claim adjudication