Research

Paper

AI LLM March 16, 2026

SCAN: Sparse Circuit Anchor Interpretable Neuron for Lifelong Knowledge Editing

Authors

Yuhuan Liu, Haitian Zhong, Xinyuan Xia, Qiang Liu, Shu Wu, Liang Wang

Abstract

Large Language Models (LLMs) often suffer from catastrophic forgetting and collapse during sequential knowledge editing. This vulnerability stems from the prevailing dense editing paradigm, which treats models as black boxes and relies on coarse-grained parameter interventions that inevitably disrupt preserved knowledge. To address this, we propose SCAN (a sparse editing framework based on Sparse Circuit Anchored Neuron) which transforms editing into a mechanism-aware manipulation by constructing a knowledge circuit via Sparse Transcoders. Experiments on Gemma2, Qwen3, and Llama3.1 across CounterFact, ZsRE and WikiFactDiff demonstrate that SCAN achieves a superior performance, maintaining model integrity on benchmarks like MMLU and GSM8K even after 3,000 sequential edits, whereas other existing methods deteriorate progressively as editing accumulates, eventually resulting in model collapse.

Metadata

arXiv ID: 2603.15226
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-03-16
Fetched: 2026-03-17 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.15226v1</id>\n    <title>SCAN: Sparse Circuit Anchor Interpretable Neuron for Lifelong Knowledge Editing</title>\n    <updated>2026-03-16T13:04:09Z</updated>\n    <link href='https://arxiv.org/abs/2603.15226v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.15226v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Large Language Models (LLMs) often suffer from catastrophic forgetting and collapse during sequential knowledge editing. This vulnerability stems from the prevailing dense editing paradigm, which treats models as black boxes and relies on coarse-grained parameter interventions that inevitably disrupt preserved knowledge. To address this, we propose SCAN (a sparse editing framework based on Sparse Circuit Anchored Neuron) which transforms editing into a mechanism-aware manipulation by constructing a knowledge circuit via Sparse Transcoders. Experiments on Gemma2, Qwen3, and Llama3.1 across CounterFact, ZsRE and WikiFactDiff demonstrate that SCAN achieves a superior performance, maintaining model integrity on benchmarks like MMLU and GSM8K even after 3,000 sequential edits, whereas other existing methods deteriorate progressively as editing accumulates, eventually resulting in model collapse.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-03-16T13:04:09Z</published>\n    <arxiv:comment>21pages, 7figures</arxiv:comment>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Yuhuan Liu</name>\n    </author>\n    <author>\n      <name>Haitian Zhong</name>\n    </author>\n    <author>\n      <name>Xinyuan Xia</name>\n    </author>\n    <author>\n      <name>Qiang Liu</name>\n    </author>\n    <author>\n      <name>Shu Wu</name>\n    </author>\n    <author>\n      <name>Liang Wang</name>\n    </author>\n  </entry>"
}