Large Discovery Models: Empirically-grounded Model-Based Open-Ended Search

  • 类型：arxiv
  • 标识：2608.15669
  • 链接：https://arxiv.org/abs/2608.15669
  • 主分类：evaluation
  • 形态：method
  • 被引：0
  • 被引来源：Semantic Scholar
  • S2被引：0
  • 影响力被引：0
  • TLDR：This work introduces the Large Discovery Model (LDM), an empirically grounded recurrent architecture that couples a generative model with a Bayesian non-parametric reward surrogate model, yielding an uncertainty-aware value that guides candidate generation, refinement, and selection.
  • 待LLM分类：否
  • 标题中文：Large Discovery Models：基于经验、依托模型驱动的开放式搜索
  • TLDR中文：提出 Large Discovery Model (LDM)，一种经验驱动的循环架构，将生成模型与贝叶斯非参数奖励代理模型耦合，产生一种感知不确定性的价值，用于引导候选的生成、精炼与选择。
  • 来源文件：
  • /inbox/tom/_candidates/2026-08-18-agent-rag-longcontext-candidates.json
  • [S2 enrich]