English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Contrastive Discovery: Open-Ended Scientific Discovery over Competing Explanations with LLM Agents

Forum topic · 小凯 · 2026-11-12

Summary

A new paper (arXiv:2611.14514, Ziang Liu, James J. Kim, Yijia Dai, Jennifer J. Sun) introduces Contrastive Discovery, a framework that improves open-ended, data-driven scientific discovery with LLM agents. Most agents evaluate hypotheses pointwise, testing a claim h only against its complement ¬h; when evidence rejects h, the agent does not learn which alternative explanation is supported. Contrastive Discovery replaces the single hypothesis with a contrast set—a scientific question paired with mutually exclusive, collectively exhaustive competing explanations. Experiments are scored by how evidence redistributes belief across the set, and a question is resolved once belief concentrates on one explanation. Since only the proposed and scored unit changes, existing agents can adopt contrast sets to re-evaluate collected evidence or guide search. Across six domains, 89% of pointwise candidates with the largest belief change were rejections, yet 73% of these resolved to specific alternatives under contrast sets. With contrast-guided search, agents solved 67% more questions than pointwise surprise under equal experiment budgets. In blinded evaluation, 38 human experts agreed on the resolved explanation in 78% of solved questions. Results show explicitly competing explanations help discovery agents extract more from the same evidence and budget.

Overview

Field: NLP Authors: Ziang Liu, James J. Kim, Yijia Dai, Jennifer J. Sun Published: 2026-11-10 arXiv: 2611.14514

Abstract

LLM agents are increasingly used for open-ended, data-driven scientific discovery, yet most of them evaluate hypotheses pointwise, testing an empirical claim \(h\) only against its complement \(\neg h\). When evidence rejects \(h\), the agent learns that one explanation is unlikely but not which alternative the evidence supports.

To recover that alternative, the authors introduce Contrastive Discovery, which replaces the single hypothesis with a contrast set — a scientific question paired with mutually exclusive and collectively exhaustive competing explanations. Each experiment is scored by how its evidence redistributes belief across the set, and the question is resolved when belief concentrates on one explanation. Because contrast sets change only the unit an agent proposes and scores, an existing agent can use them to re-evaluate already collected evidence or to guide search.

Key Results

  • Across six domains, 89% of pointwise candidates with the largest belief change were rejections; with the same evidence, contrast sets resolved 73% of those to a specific alternative.
  • Contrast-guided search solved 67% more questions than pointwise surprise under the same experiment budget.
  • In blinded evaluation, 38 human experts selected the same explanation in 78% of solved questions, and 82% on questions where a pointwise agent rejected and stopped.
  • Explicitly competing explanations enable discovery agents to solve more scientific questions from the same evidence and budget.
---

*Auto-collected on 2026-11-12.*

Tags

#llm-agents#scientific-discovery#contrastive-discovery#hypothesis-testing#nlp#arxiv-paper

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178637064