---
id: 20261104-T0-05
title: "Circuit Oracle：用自然语言提问自动分析归因图，加速模型可解释性研究"
title_en: "Circuit Oracle Automates Attribution Graph Analysis via Natural-Language Queries"
url: https://ai.daily.yangsir.net/daily/20261104-T0-05
issue_date: 2026-11-04
publish_date: 2026-11-03T05:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2611.00420
---

# Circuit Oracle：用自然语言提问自动分析归因图，加速模型可解释性研究

研究者提出Circuit Oracle，用自然语言查询自动完成归因图分析。归因图是机械可解释性领域的新工具，借助转码器把语言模型计算拆解为稀疏可解释特征，并用因果边连接。但把一张图转化为安全相关的洞见仍需大量人工。Circuit Oracle把这一环节自动化，让研究者直接用自然语言提问就能得到分析结果，降低可解释性研究的操作门槛。

## English Version

**Circuit Oracle Automates Attribution Graph Analysis via Natural-Language Queries**

Researchers introduced Circuit Oracle, a system that automates attribution graph analysis through natural-language queries. Attribution graphs, an emerging tool in mechanistic interpretability, use transcoders to decompose language model computations into sparse interpretable features linked by causal edges. Converting such a graph into safety-relevant insight has required substantial manual effort. Circuit Oracle automates that step, letting researchers query graphs in plain language and lowering the operational barrier to interpretability work.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2611.00420)

**详情页**：https://ai.daily.yangsir.net/daily/20261104-T0-05

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*