让AI生成更真实：用逻辑约束减少幻觉，不降速度。
Truth-Aware Decoding: A Program-Logic Approach to Factual Language Generation
- 用程序逻辑构建实时校验机制，生成时自动核对知识。
- 实验证明可降低幻觉率，且处理速度不下降。
- 适合需要高准确性的对话系统和内容生成场景。
本文提出真相感知解码（Truth-Aware Decoding, TAD），一种基于知识库的验证型解码方案，将神经语言生成与知识库对齐。TAD 在现代指令微调系统中引入一个语义守卫的格结构，于解码阶段运行。主要贡献包括：(i) 构建基于约束的语义，将理想过滤视为程序逻辑判断；(ii) 证明在健全且完备的守卫下，贪心选择具有局部似然优势（定理2.7）；(iii) 提出一种熵风格不变量，通过知识感知安全质量量化事实风险；(iv) 设计多智能体操作演算，并以经验证的Lean形式化工具认证实现行为。数值与算法案例表明，该机制可有效减少幻觉而不牺牲吞吐量，为大规模经验模型与形式化验证之间搭建了实用桥梁。
原文摘要 · Abstract (English)
This paper introduces Truth-Aware Decoding (TAD), a verification-oriented decoding scheme that aligns neural language generation with knowledge bases. Situated in the tradition of probabilistic program semantics for sequence models, TAD augments modern instruction-tuned systems with a lattice of semantic guards that operate at decode time. Our contributions are fourfold: (i) a constraint-based semantics that renders oracle filtering as a program-logic judgment, (ii) a proof that greedy selection enjoys local likelihood dominance under sound and complete guards (Theorem 2.7), (iii) an entropy-style invariant that quantifies factual risk via knowledge-aware safe mass, and (iv) a multi-agent operational calculus with verified Lean artefacts to certify implementation behaviour. Numerical and algorithmic case studies confirm that the resulting guardrails reduce hallucinations without sacrificing throughput, yielding a pragmatic bridge between large-scale empirical models and formal verification.
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。