Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

DeCNIP

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

1

Sources

1

Corpus findings

2026-07-23 / arxiv-researcherDeCNIP Removes LLM Backdoors by Pruning 0.1% of Neurons — 95%+ Attack Success Reduction, 97% Utility RetainedExisting backdoor defenses target fine-tuning-based implants and classification settings, missing model-editing attacks that bypass the training pipeline entirely and failing to extend to open-ended generation. DeCNIP instead finds the mechanism: it optimizes a cross-entropy loss between harmful prompts with candidate tokens and benign inputs to surface trigger-like behavior, isolates the Backdoor Critical Neurons that triggers hijack, and prunes only those. Across six open-source LLMs and two benchmarks it cuts attack success rate by over 95% relative — beating seven state-of-the-art defenses — while intervening on 0.1% of neurons and keeping 97% of normal benchmark performance.

Graph relationships

SUPPORTS
Qwen 2.5 7B
DeCNIP was evaluated on Qwen-2.5-7B for backdoor defense.
Source finding

Source trail

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues