Evidence-Informed LLM Beliefs for Continual Scientific Discovery

• URL: http://arxiv.org/abs/2606.29182v1
• Date: Sun, 28 Jun 2026 04:06:12 GMT
• Title: Evidence-Informed LLM Beliefs for Continual Scientific Discovery
• Abstract summary: We introduce two complementary changes to the original search procedure: belief-update filtering and diversity.<n>We find that embedding-based retrieval-augmented generation over prior discoveries best anticipates eventual posteriors, identifying 37.5% of static surprisals as spurious.<n>We then modify search to avoid these spurious rewards and prioritize hypotheses that remain surprising under non-stationary beliefs.
• Score: 66.36520571709482
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Open-ended scientific discovery with large language models (LLMs) increasingly operates as a long-horizon loop of hypothesis search and verification, where a reward signal guides which hypotheses to test next. A notable recent example is AutoDiscovery, which uses "Bayesian surprise" - the belief shift an LLM undergoes after observing evidence for a hypothesis - as both a discovery metric and a reward for search. We first observe that AutoDiscovery treats surprisal as a static quantity, while surprisal in human reasoning is non-stationary - it is defined relative to beliefs that evolve with experience, a prerequisite for continual scientific discovery. We address this mismatch with evidence-informed LLM beliefs: priors updated with evidence from previous hypotheses to compute non-stationary surprisal for new hypotheses. We compare in-context belief-updating mechanisms and find that embedding-based retrieval-augmented generation over prior discoveries best anticipates eventual posteriors, identifying 37.5% of static surprisals as spurious. We then modify search to avoid these spurious rewards and prioritize hypotheses that remain surprising under non-stationary beliefs. Concretely, we introduce two complementary changes to the original search procedure: belief-update filtering and diversity maximization. Across five discovery domains, our method increases accumulated non-stationary surprisal by 30.62% on average compared to the original search procedure, demonstrating that continual scientific discovery with LLMs requires not only better belief measurement but also search procedures that avoid redundancy and encourage diversity.

Related papers

• RAP: Research Attention Prediction Reveals Target-Conditioned Evidence Acquisition Biases [58.92094935632493]
  Large language models (LLMs) increasingly act as research agents, yet their ability to track shifts in research attention is difficult to evaluate.<n>We introduce Research Attention Prediction (RAP), a rolling benchmark covering 278 AI/ML fields and 1,390 episodes.<n>At each cut-off, an LLM agent searches a temporally restricted arXiv corpus and predicts the next six months' paper shares across eight frozen research directions.
  arXiv  Detail & Related papers  (2026-09-09T12:15:14Z)
• Autonomous Scientific Discovery via Iterative Meta-Reflection [44.46230654454998]
  We introduce DiscoPER, a framework that conducts open-ended research by dynamically generating and executing code to explore datasets without pre-specified research objectives.<n>By treating prior discoveries as empirical data, DiscoPER actively redirecting hypothesis exploration toward uncharted regions of the search space.<n> evaluated on iNatDisco, a new multimodal ecological knowledge benchmark with pattern-level ground truth obtained from peer-reviewed literature.
  arXiv  Detail & Related papers  (2026-07-01T16:16:07Z)
• BayesEvolve: Explicit Belief States for Autonomous Scientific Discovery [10.04049356690421]
  BayesEvolve is a belief-guided discovery framework that converts experimental evidence into a predictive belief state.<n>We show that BayesEvolve exhibits productive late-stage concentration rather than unfocused exploration.
  arXiv  Detail & Related papers  (2026-06-29T14:14:33Z)
• Towards Diverse Scientific Hypothesis Search with Large Language Models [46.83325596046927]
  We formulate hypothesis search as a sampling problem, where the objective is to efficiently produce diverse, high-quality hypotheses under a fixed validation budget.<n>Across domains including molecular discovery, equation discovery, and algorithm discovery, our approach consistently improves both hypothesis quality and diversity under the same validation budget.
  arXiv  Detail & Related papers  (2026-06-09T08:52:49Z)
• Failing to Falsify: Evaluating and Mitigating Confirmation Bias in Language Models [38.162539966977484]
  Large language models (LLMs) exhibit confirmation bias by adapting the rule-discovery study from human psychology.<n>We find that LLMs exhibit confirmation bias, often proposing triples to confirm their hypothesis rather than trying to falsify it.
  arXiv  Detail & Related papers  (2026-04-02T19:39:52Z)
• Accelerating Social Science Research via Agentic Hypothesization and Experimentation [33.55093074029515]
  EXPERIGEN is a framework that operationalizes end-to-end discovery through a Bayesian optimization inspired two-phase search.<n>It consistently discovers 2-4x more statistically significant hypotheses that are 7-17 percent more predictive than prior approaches.<n>We conduct the first A/B test of LLM-generated hypotheses, observing statistically significant results with p less than 1e-6 and a large effect size of 344 percent.
  arXiv  Detail & Related papers  (2026-02-08T14:20:56Z)
• Bayes-Entropy Collaborative Driven Agents for Research Hypotheses Generation and Optimization [4.469102316542763]
  This paper proposes a multi-agent collaborative framework called HypoAgents.<n>It generates hypotheses through diversity sampling and establishes prior beliefs.<n>It then employs etrieval-augmented generation (RAG) to gather external literature evidence.<n>It identifies high-uncertainty hypotheses using information entropy $H = - sum p_ilog p_i$ and actively refines them.
  arXiv  Detail & Related papers  (2025-08-03T13:05:32Z)
• Open-ended Scientific Discovery via Bayesian Surprise [63.26412847240136]
  AutoDS is a method for open-ended scientific discovery that instead drives scientific exploration using Bayesian surprise.<n>We evaluate AutoDS in the setting of data-driven discovery across 21 real-world datasets spanning domains such as biology, economics, finance, and behavioral science.
  arXiv  Detail & Related papers  (2025-06-30T22:53:59Z)
• MOOSE-Chem2: Exploring LLM Limits in Fine-Grained Scientific Hypothesis Discovery via Hierarchical Search [102.11776494401705]
  Large language models (LLMs) have shown promise in automating scientific hypothesis generation.<n>Existing approaches primarily yield coarse-grained hypotheses lacking critical methodological and experimental details.<n>We introduce and formally define the new task of fine-grained scientific hypothesis discovery.
  arXiv  Detail & Related papers  (2025-05-25T16:13:46Z)
• ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Task Decomposition [67.26124739345332]
  Large language models (LLMs) have demonstrated potential in assisting scientific research, yet their ability to discover high-quality research hypotheses remains unexamined.<n>We introduce the first large-scale benchmark for evaluating LLMs with a near-sufficient set of sub-tasks of scientific discovery.<n>We develop an automated framework that extracts critical components - research questions, background surveys, inspirations, and hypotheses - from scientific papers.
  arXiv  Detail & Related papers  (2025-03-27T08:09:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.