Metraipolis › archive › dc_5912d6ce0aaa

When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic — arXiv:2609.39704

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39704
License (as archived)
CC0-1.0
Archived
2026-10-07T12:45:43.196Z
SHA-256 of the archived text
6c78d64f86d08c5a1981ab72f18f6a119e516eaf19023fde209aee45e4390ea7
Stable URL of this page
https://metraipolis.com/documento/dc_5912d6ce0aaa

How to cite

When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic — arXiv:2609.39704 — arXiv (en), https://arxiv.org/abs/2609.39704 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5912d6ce0aaa (Metraipolis archive dc_5912d6ce0aaa, sha256 6c78d64f86d08c5a1981ab72f18f6a119e516eaf19023fde209aee45e4390ea7)
BibTeX
@misc{dc5912d6ce0aaa,
  title = {When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic — arXiv:2609.39704},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39704},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5912d6ce0aaa; sha256 6c78d64f86d08c5a1981ab72f18f6a119e516eaf19023fde209aee45e4390ea7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39704. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

This paper demonstrate that whether masking-based token pruning helps or hurts worst-group robustness can be predicted before deployment, without labels or fine-tuning. A systematic study of semantic masking across 8 spurious-correlation benchmarks shows its effect on worst-group accuracy is highly unstable: it improves accuracy by up to 82.5\% relative on some datasets and degrades it by up to 100\% on others. We trace this instability to spurious inversion: background patches receive higher CLIP text-similarity than the true object when the spurious attribute is background-separable, inverting the assumption every text- and attention-guided pruning method relies on. We introduce the Spurious Inversion Metric (SIM), a label-free, pre-deployment diagnostic whose sign predicts this effect with statistical significance (binomial $p=0.035$) across all 8 datasets, and remains dependable across 6 CLIP architectures with a clean foreground/background split. Naive masking is itself a major source of risk: it causes the largest average-accuracy loss of any method we evaluate, and its own per-image segmentation step is a significant runtime bottleneck. To address this, we design a batched, synchronization-free GPU segmentation routine that cuts this overhead from 3.5$\times$ to 1.75$\times$ baseline. Gating deployment by SIM's sign recovers masking's benefits while avoiding its worst failures, matching or exceeding a strong pruning baseline on 7 of 8 datasets.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.39704 — 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 2026-10-07T12:14:14.141Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).