Metraipolis › archive › dc_fa16527cb753
SAGG: Sample-Adaptive Gradient Gating for Robust Multimodal Learning under Heterogeneous Corruption — arXiv:2609.20302
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.20302
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-10T04:09:00.675Z
- SHA-256 of the archived text
33e6412a066ec92bd51d430cf3b0487a6670f4f545507dcf469366ce8a7e525c- Stable URL of this page
https://metraipolis.com/documento/dc_fa16527cb753
How to cite
SAGG: Sample-Adaptive Gradient Gating for Robust Multimodal Learning under Heterogeneous Corruption — arXiv:2609.20302 — arXiv (en), https://arxiv.org/abs/2609.20302 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fa16527cb753 (Metraipolis archive dc_fa16527cb753, sha256 33e6412a066ec92bd51d430cf3b0487a6670f4f545507dcf469366ce8a7e525c)
BibTeX
@misc{dcfa16527cb753,
title = {SAGG: Sample-Adaptive Gradient Gating for Robust Multimodal Learning under Heterogeneous Corruption — arXiv:2609.20302},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.20302},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fa16527cb753; sha256 33e6412a066ec92bd51d430cf3b0487a6670f4f545507dcf469366ce8a7e525c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20302. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal gradient balancing methods modulate encoder gradients with a shared scalar per modality, implicitly assuming that corruption is uniform across the training batch. In practice, corruption is sample-heterogeneous: within a single mini-batch, different samples may have different modalities corrupted. We prove that under this heterogeneous corruption model, any batch-level sample-agnostic linear estimator with a shared modulation parameter incurs an irreducible bias with respect to the clean-data gradient, and that sample-level all-or-nothing gating is the unique unbiased strategy within a natural distribution-free estimator class. Motivated by this result, we propose Sample-Adaptive Gradient Gating (SAGG), which makes a binary retain-or-discard decision per sample via an online feature-norm quality test and incorporates a truncation mechanism for variance control. We prove that SAGG-based SGD converges at the standard O(1/sqrt(T)) rate to stationary points of the clean loss without a corruption-dependent error floor, and derive a certified robustness radius for the independent-encoder architecture that connects per-modality Lipschitz constants to the classification margin. Experiments on Kinetics-Sounds and UCF-101 under Gaussian noise injection, partial modality missing, and natural contribution imbalance show that SAGG consistently outperforms ten existing methods, with the largest gains in high-corruption regimes where batch-level bias is most severe. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.LG). Categorie arXiv: Machine Learning (cs.LG) ; Multimedia (cs.MM). Pagina del lavoro: https://arxiv.org/abs/2609.20302 — 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 2026-10-10T03:48:18.353Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).