Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs

Abstract

Inference optimization is a vital technique for deploying LLMs at scale. Compilation is the most widely adopted optimization technique for LLMs. While it assumes semantic equivalence between the original and compiled graphs, we first uncover its numerical side effects can be maliciously exploited to implant stealthy backdoors in LLMs. We propose a unified optimization-triggered attack framework comprising two complementary strategies. Without any modification to the compiler or hardware, one str

Categories

Framework mappings

Suggested from the entry's categories.

@misc{wang2026trusted,
  title = {{Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs}},
  author = {Yifei Wang and Tianlin Li and Xiaohan Zhang and Yida Yang and Xiaoyu Zhang and Li Pan},
  year = {2026},
  month = may,
  eprint = {2605.20641},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.20641}
}