让大模型主动提问，通过贝叶斯推理更高效地解决复杂任务。
BALAR : A Bayesian Agentic Loop for Active Reasoning
- 基于贝叶斯框架动态构建信念模型，自动判断该问什么。
- 在三类任务上平均提升30%以上准确率，最高达38.5%。
- 无需微调，通用性强，适合需要多轮交互的智能助手场景。
大型语言模型越来越多地在需要与用户多次交互的场景中执行任务。然而，当前多数系统仅被动响应，缺乏对缺失信息和下一步提问策略的系统性推理机制。本文提出BALAR（贝叶斯主动推理代理循环），一种无需微调的任务无关外层算法，可实现大模型代理与用户之间的结构化多轮交互。BALAR维护对潜在状态的结构化信念，通过最大化期望互信息来选择澄清问题，并在当前状态表示不足时动态扩展。我们在三个不同基准测试上评估了BALAR：AR-Bench-DC（侦探案例）、AR-Bench-SP（思维谜题）和iCraft-MD（临床诊断）。BALAR在所有三个基准上均显著优于所有基线方法，在AR-Bench-DC上准确率高出14.6%，在AR-Bench-SP上高出38.5%，在iCraft-MD上高出30.5%。
原文摘要 · Abstract (English)
Large language models increasingly operate in interactive settings where solving a task requires multiple rounds of information exchange with a user. However, most current systems treat dialogue reactively and lack a principled mechanism to reason about what information is missing and which question should be asked next. We propose BALAR (Bayesian Agentic Loop for Active Reasoning), a task-agnostic outer-loop algorithm that requires no fine-tuning and enables structured multi-turn interaction between an LLM agent and a user. BALAR maintains a structured belief over latent states, selects clarifying questions by maximizing expected mutual information, and dynamically expands its state representation when the current one proves insufficient. We evaluate BALAR on three diverse benchmarks: AR-Bench-DC (detective cases), AR-Bench-SP (thinking puzzles), and iCraft-MD (clinical diagnosis). BALAR significantly outperforms all baselines across all three benchmarks, with $14.6\%$ higher accuracy on AR-Bench-DC, $38.5\%$ on AR-Bench-SP, and $30.5\%$ on iCraft-MD.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。