---
title: "Theoria：非正式推理状态下的重写-接受验证架构"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.01223v1"
permalink: "https://feed.astralor.com/items/itm_324f64724240c32eda6292a5"
published: "2026-07-01T17:56:42.000Z"
score: 59
channel: "ai"
tags: ["AI安全","Theoria","可审计","对抗性检测","验证架构"]
---

# Theoria：非正式推理状态下的重写-接受验证架构

## 速览

Theoria 是一种新型验证架构，通过将AI系统的答案重写为类型化状态转换序列，每个转换都附带明确的引用、计算或问题给定事实作为理由，从而实现每一步的独立审计。在HLE-Verified Gold基准测试中，Theoria以91.4%的严格精度认证了105个问题，并生成人类可读的证明轨迹。与整体式LLM评判器相比，Theoria在对抗性中毒证明检测中表现出94.7%的捕获率，尤其在隐蔽前提和伪造引用上优势显著。该架构为提升LLM输出的可信度提供了可审计、可复现的解决路径。

## 关键点

- Theoria 将答案转化为带理由的步骤序列，所有步骤均可独立审查。
- 在185个专家级问题上达到91.4%严格精度，生成人类可读证明轨迹。
- 对抗性攻击下检测率94.7%，隐蔽前提检测优势达28个百分点。
- 与整体评判器互补，共同提升LLM输出可信度。

## 全文

When should an AI system's answer be trusted? Formal proof assistants offer certainty but cannot reach most of the problem distribution; scalar LLM judges offer coverage but produce opaque scores that cannot be audited after the fact and are subject to the same coherence issues as any LLM. We present Theoria, a verification architecture that closes this gap. A candidate solution is rewritten into a sequence of typed state transitions, each licensed by an explicit justification, whether that be a citation, computation, or problem-given fact, and every transition is independently auditable. The foundational invariant is completeness of change: every difference between consecutive proof states must be accounted for, so hidden premises surface as unlicensed mutations rather than passing silently. On HLE-Verified Gold (185 text-only expert problems), Theoria certifies 105 at 91.4% strict precision (Wilson 95% CI [84.5%, 95.4%]). Every certification produces a human readable proof trace in which each step can be independently challenged. Holistic LLM judges achieve comparable precision at matched coverage but fail on different problems (Jaccard 0.14-0.36), making the approaches complementary. On 95 adversarial poisoned proofs across 15 domains, structured judges catch 94.7% versus 83.2% for holistic judging (p= 0.0017). The overall 11.5 pp gap concentrates in hidden premises (90.6% vs. 62.5%, a 28 pp difference) and fabricated citations (100% vs. 90%), the error classes where the formal analysis predicts an advantage; performance is identical on arithmetic and theorem-misapplication errors, where no advantage is predicted. On GPQA Diamond (n= 65), certified precision is 97.1% (Wilson CI [85.1%, 99.5%]).

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_324f64724240c32eda6292a5