Metraipolis › archive › dc_6455a64ddbaf

No Free Efficiency: Revisiting the Trade-off Between Training Efficiency and Model Vulnerability — arXiv:2609.33898

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33898
License (as archived)
CC0-1.0
Archived
2026-10-07T22:35:04.974Z
SHA-256 of the archived text
1101a6751f56d1eea30945af572745a3d48cb18489a2ee44b1e5f479e4585cc7
Stable URL of this page
https://metraipolis.com/documento/dc_6455a64ddbaf

How to cite

No Free Efficiency: Revisiting the Trade-off Between Training Efficiency and Model Vulnerability — arXiv:2609.33898 — arXiv (en), https://arxiv.org/abs/2609.33898 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6455a64ddbaf (Metraipolis archive dc_6455a64ddbaf, sha256 1101a6751f56d1eea30945af572745a3d48cb18489a2ee44b1e5f479e4585cc7)
BibTeX
@misc{dc6455a64ddbaf,
  title = {No Free Efficiency: Revisiting the Trade-off Between Training Efficiency and Model Vulnerability — arXiv:2609.33898},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33898},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6455a64ddbaf; sha256 1101a6751f56d1eea30945af572745a3d48cb18489a2ee44b1e5f479e4585cc7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33898. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Training efficiency has become the central driver of recent progress in foundation models. To overcome the massive computational and data requirements of large-scale training, researchers increasingly adopt strategies such as selective data sampling, efficient pre-training, and simplified reinforcement learning pipelines. While these strategies drastically reduce overhead, they prompt a critical, yet neglected question: Is efficiency achieved at the expense of model robustness and security? To our knowledge, we present the first systematic cross-domain investigation of the efficiency-vulnerability trade-off. Across vision and language models, we show that efficiency-oriented training increases susceptibility to adversarial and privacy attacks. We characterize this vulnerability by analyzing the models' internal geometry and functional representations, demonstrating that the evaluated efficient variants consistently exhibit sharper loss geometry together with systematic changes in representational structure. We further extend our analysis to "zero RL training", finding that models trained using simplified RL recipes exhibit substantially greater susceptibility to catastrophic forgetting and more pronounced overconfidence than those trained through conventional alignment pipelines. Our findings suggest that training efficiency is rarely a "free lunch"; rather, the mechanisms that minimize computation can inadvertently compromise safety. We conclude by calling for a paradigm shift toward multi-objective training that jointly optimizes for performance, cost, and security.

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CR).
Categorie arXiv: Machine Learning (cs.LG) ; Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2609.33898 — 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 2026-10-07T22:13:03.781Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).