paperSeptember 2026Unreviewed

Membership Inference in Fine-tuned Diffusion Language Models via Token-level Memorization Asymmetry

Shengfang Zhai, Leo Marchyok, Yuling Shi, Huanran Chen, Yinpeng Dong, Jiaheng Zhang, Sanghyun Hong

Abstract

Diffusion language models (DLMs) have recently emerged as an alternative modeling paradigm to autoregressive LMs, offering advantages such as parallel generation and bidirectional context modeling. Despite growing interest in their generative capabilities, the privacy risks of DLMs remain underexplored. We identify a phenomenon termed token-level memorization asymmetry through theoretical analysis of diffusion training dynamics. Building on this finding, we propose Q-Skew, a quantile-weighted sk

Framework mappings

OWASP Top 10 for LLM Applications

LLM02Sensitive Information Disclosure

MITRE ATLAS

AML.T0024.000Infer Training Data Membership

Suggested from the entry's categories.

Cite

@misc{zhai2026membership,
  title = {{Membership Inference in Fine-tuned Diffusion Language Models via Token-level Memorization Asymmetry}},
  author = {Shengfang Zhai and Leo Marchyok and Yuling Shi and Huanran Chen and Yinpeng Dong and Jiaheng Zhang and Sanghyun Hong},
  year = {2026},
  month = sep,
  eprint = {2609.00873},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.00873}
}