Metraipolis › archive › dc_ef64e71e4a23

Hidden Threat in Synthetic Data: Covert Targeted Bias Injection through Benign Text — arXiv:2608.30619

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.30619
License (as archived)
CC0-1.0
Archived
2026-10-17T07:52:49.415Z
SHA-256 of the archived text
c5f142d215ba70c77f1e930f387ad82dbc5ec821ef18cd60c64f68200e75b393
Stable URL of this page
https://metraipolis.com/documento/dc_ef64e71e4a23

How to cite

Hidden Threat in Synthetic Data: Covert Targeted Bias Injection through Benign Text — arXiv:2608.30619 — arXiv (en), https://arxiv.org/abs/2608.30619 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ef64e71e4a23 (Metraipolis archive dc_ef64e71e4a23, sha256 c5f142d215ba70c77f1e930f387ad82dbc5ec821ef18cd60c64f68200e75b393)
BibTeX
@misc{dcef64e71e4a23,
  title = {Hidden Threat in Synthetic Data: Covert Targeted Bias Injection through Benign Text — arXiv:2608.30619},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.30619},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ef64e71e4a23; sha256 c5f142d215ba70c77f1e930f387ad82dbc5ec821ef18cd60c64f68200e75b393},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30619. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Synthetic data is increasingly used to train large language models (LLMs), yet its security implications remain poorly understood. Prior work on subliminal learning suggests that models can inherit behavioral traits from seemingly unrelated training data. In this work, we investigate whether such mechanisms can be exploited to inject targeted social biases into aligned models through semantically benign synthetic data. We construct a pipeline in which a misaligned teacher model generates filtered synthetic datasets across domains such as creative writing and code generation, which are then used to fine-tune aligned student models. Our experiments show that benign-looking synthetic data can act as a covert channel for transmitting targeted biases while largely preserving the student model's general task capabilities. These results reveal a previously underexplored security risk in synthetic data-driven LLM training pipelines and highlight the need for improved safeguards. As one possible step toward this goal, we suggest that log-linearity-based scoring may provide a useful signal for screening seemingly benign synthetic data.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.30619 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T07:28:26.889Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).