---
title: "The Geometry of Ignorance: Detecting Suppressed Knowledge in LLMs through Activation Fragmentation"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.05272v1"
permalink: "https://feed.astralor.com/items/itm_a41a0cea68f7544e0ecbb090"
published: "2026-11-03T17:55:57.000Z"
score: 59
channel: "tech"
tags: ["LLM","RAG","The Geometry","Ignorance","Detecting Suppressed Knowledge","LLMs","Activation Fragmentation Evaluating","Evaluating"]
---

# The Geometry of Ignorance: Detecting Suppressed Knowledge in LLMs through Activation Fragmentation

## 速览

Evaluating whether a large language model (LLM) genuinely knows a concept remains a fundamental problem, since behavioral metrics can be fooled by suppression: a model that refuses to answer may still harbor recoverable…

## 关键点

- Evaluating whether a large language model (LLM) genuinely knows a concept remains a fundamental problem, since behavioral metrics can be fooled by suppression: a model that refuses to answer may still harbor recoverable…
- The framework measures the geometric coherence of \textit{anon-vectors}, the differences in hidden states between a concept-specific query and its anonymized paraphrase, and shows that genuine knowledge forms a tight, l…
- Across ten instruction-tuned LLMs from three model families, all fragmentation metrics reliably separate known from unknown concepts, and post-unlearning, fragmentation geometry correlates with adversarial recovery acro…

## 全文

Evaluating whether a large language model (LLM) genuinely knows a concept remains a fundamental problem, since behavioral metrics can be fooled by suppression: a model that refuses to answer may still harbor recoverable knowledge. The framework measures the geometric coherence of \textit{anon-vectors}, the differences in hidden states between a concept-specific query and its anonymized paraphrase, and shows that genuine knowledge forms a tight, low-entropy cluster while ignorance forces geometric fragmentation. Across ten instruction-tuned LLMs from three model families, all fragmentation metrics reliably separate known from unknown concepts, and post-unlearning, fragmentation geometry correlates with adversarial recovery across four attack types and five unlearning algorithms, though the direction of that relationship is shown to depend on attack methodology. Furthermore, a layer-resolved analysis reveals that different adversarial mechanisms exploit this geometry at varying depths, with high attack-specific correlations that are invisible to behavioral evaluation. Together, these results constitute a single-phase, attack-free proxy for unlearning robustness that captures aspects of parametric knowledge inaccessible to output-level metrics. A reference implementation is available at https://github.com/AdarPeleg/geometry-of-ignorance

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_a41a0cea68f7544e0ecbb090