cs.LGNov 14, 2026

A Backdoor Trigger Is Legible in Fine-Tune Weights

Authors: Snigdha Chandan Khilar

Organizations: Independent Researcher

Abstract

A backdoored language model behaves normally until a trigger phrase appears in its input. Every published method for recovering such a trigger runs the model. We show the trigger is legible in the weights alone: a static statistic reads a fine-tuned checkpoint with no forward pass of the suspect, no data, no labelled population of poisoned models and no prior knowledge that a backdoor exists, and names one of the tokens of the trigger phrase as the single highest-ranked candidate in the vocabulary.

Figures & tables

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Language-Switching Triggers Take a Latent Detour Through Language Models

    May 18, 2026Francis Kulumba, Wissam Antoun, Théo Lasnier +2LLM Backdoor AttacksMechanistic Interpretability

  2. ToxScreen: Detecting Whether an LLM Has Been Poisoned

    Jul 29, 2026Anthony Hughes, Nicole Xing, Collin Francel +2LLM Backdoor AttacksBackdoor Detection