用确定性事实库和对抗检测器，实现金融推理零幻觉。
Neuro-Symbolic Financial Reasoning via Deterministic Fact Ledgers and Adversarial Low-Latency Hallucination Detector
- 构建严格类型的通用事实账本，通过数学约束确保数据确定性。
- 系统幻觉率降至1.2%，在金融场景下接近零误差。
- 轻量级哨兵模型可单令牌检测错误，适合高时效金融应用。
标准检索增强生成（RAG）架构在高风险金融领域表现不佳，主要因大语言模型存在固有算术缺陷，以及密集向量检索引发语义混淆（如将‘净利润’误映射为‘净销售额’）。在确定性领域，99%准确率仍无法建立操作信任。为实现零幻觉金融推理，本文提出可验证数值推理代理（VeNRA）。VeNRA 将 RAG 范式从概率性文本检索转向确定性变量检索，通过严格类型化的通用事实账本（UFL）实现，并以新颖的双锁接地算法进行数学约束。结合确定性 Python 执行，该神经符号路由将系统幻觉率压缩至近乎零的 1.2%。针对上游解析异常不可避免的问题，引入 VeNRA Sentinel——一个 30 亿参数的小型语言模型，仅用单令牌推理预算即可前瞻性审计候选结果，支持可选后验推理。为训练哨兵，采用对抗模拟替代传统幻觉数据集，程序化破坏财务记录以模拟生态错误。紧凑的哨兵模型在错误检测上超越 700 亿参数前沿模型。通过逆向链式思维训练中的损失稀释现象，提出新型微块损失算法，在极端惩罚下稳定梯度，实现 28 倍延迟加速且不牺牲审计严谨性。
原文摘要 · Abstract (English)
Standard Retrieval-Augmented Generation (RAG) architectures fail in high-stakes financial domains due to two fundamental limitations: the inherent arithmetic incompetence of Large Language Models (LLMs) and the distributional semantic conflation of dense vector retrieval (e.g., mapping "Net Income" to "Net Sales" due to contextual proximity). In deterministic domains, a 99% accuracy rate yields 0% operational trust. To achieve zero-hallucination financial reasoning, we introduce the Verifiable Numerical Reasoning Agent (VeNRA). VeNRA shifts the RAG paradigm from retrieving probabilistic text to retrieving deterministic variables via a strictly typed Universal Fact Ledger (UFL). We mathematically bound this ledger using a novel Double-Lock Grounding algorithm. Coupled with deterministic Python execution, this neuro-symbolic routing compresses systemic hallucination rates to a near-zero 1.2%. Recognising that upstream parsing anomalies inevitably occur, we introduce the VeNRA Sentinel: a 3-billion parameter SLM trained to forensically audit candidate using a single-token inference budget with optional post-hoc reasoning. To train the Sentinel, we steer away from traditional hallucination datasets in favour of Adversarial Simulation, programmatically sabotaging financial records to simulate Ecological Errors. The compact Sentinel consequently outperforms 70B+ frontier models in error detection. Through Loss Dilution phenomenon in Reverse-CoT training, we present a novel Micro-Chunking loss algorithm to stabilise gradients under extreme verdict penalisation, yielding a 28x latency speedup without sacrificing forensic rigor.
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。