跳到正文
arXiv cs.AI· Kaiwen Shi, Zheyuan Zhang, Weixiang Sun, Han Bao, Colby Nelson, Yanfang Ye·· 1 小时前AI 评分32

LProp:在 LLM 智能体中通过证据条件信念修正传播潜在不确定性

LProp: Propagating Latent Uncertainty through Evidence-Conditioned Belief Revision in LLM Agents

AI 导读

LProp(Latent Uncertainty Propagation)提出潜在空间不确定性估计方法 Latent Uncertainty,用潜在表示和新证据修正继承的答案分布,无需额外轨迹采样。该方法在选择题、是非题、开放式问答和数学推理任务上多数指标优于基线,并在全部 8 个传播设置中实现最低的平均推理 token 用量。消融实验显示继承的潜在内容比仅用答案概率或模型容量更能改善错误恢复。

正文 · 原文

View PDF HTML (experimental)

Abstract:When LLM agents pass intermediate judgments downstream without uncertainty, later decisions may treat tentative assumptions as facts. Effective uncertainty propagation must preserve information for correction and update inherited beliefs as new evidence arrives. Yet scalar scores omit information needed for revision, while detailed descriptions and reflection incur additional costs. We introduce Latent Uncertainty, a latent-space estimator that preserves uncertainty-relevant information without lengthy descriptions. Building on this, we propose LProp (Latent Uncertainty Propagation), which revises inherited answer distributions using latent representations and new evidence without additional trajectory sampling. We evaluate both methods across multiple-choice, yes/no, open-ended question answering, and mathematical reasoning tasks. Our approach outperforms baselines on most metrics, while LProp achieves the lowest mean inference-token usage in all eight propagation settings. Matched feature ablations further demonstrate that inherited latent content improves error recovery beyond answer probabilities or model capacity alone.
Subjects: Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.23564 [cs.AI]
  (or arXiv:2611.23564v1 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2611.23564

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Kaiwen Shi [view email]
[v1] Tue, 17 Nov 2026 22:12:54 UTC (1,111 KB)

来源:arXiv cs.AI · arxiv.org