Metraipolis › archive › dc_dceeb3206ccf

Conflict-Aware Logit Adapters for Utility-Preserving Anti-Distillation — arXiv:2610.31046

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.31046
License (as archived): CC0-1.0
Archived: 2026-10-25T20:53:13.512Z
SHA-256 of the archived text: 4ec878522b2736ae13d7a69798bcc5147d5a20e2fa2b1c91c71d59424bf49064
Stable URL of this page: https://metraipolis.com/documento/dc_dceeb3206ccf

How to cite

Conflict-Aware Logit Adapters for Utility-Preserving Anti-Distillation — arXiv:2610.31046 — arXiv (en), https://arxiv.org/abs/2610.31046 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dceeb3206ccf (Metraipolis archive dc_dceeb3206ccf, sha256 4ec878522b2736ae13d7a69798bcc5147d5a20e2fa2b1c91c71d59424bf49064)

BibTeX

@misc{dcdceeb3206ccf,
  title = {Conflict-Aware Logit Adapters for Utility-Preserving Anti-Distillation — arXiv:2610.31046},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.31046},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dceeb3206ccf; sha256 4ec878522b2736ae13d7a69798bcc5147d5a20e2fa2b1c91c71d59424bf49064},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.31046. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

While reasoning traces improve model capability, they also provide supervision signals for downstream distillation. Initial efforts to mitigate this issue rely on decoding-time interventions, but they incur high inference overhead and can harm the teacher model's natural fluency. We propose Conflict-Aware Logit Adapters (CALA), a lightweight training-time defense. CALA attaches a small residual adapter to a frozen teacher model and trains only the adapter. A proxy student identifies positions where imitation is most sensitive, and the adapter perturbs the output distribution at these positions while remaining within high-probability regions. A conflict-aware optimization strategy balances anti-distillation and utility objectives, while regularization preserves the teacher's performance and fluency. At inference time, CALA incurs modest overhead and requires no auxiliary model. Experiments on reasoning benchmarks show that CALA substantially reduces successful student imitation while maintaining near-original teacher performance and generation quality. These results establish CALA as a practical and efficient alternative to decoding-time approaches.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.31046 — 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 2026-10-25T20:47:56.898Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).