Metraipolis › archive › dc_63bc7dde8f70
Representation Editing for Multimodal Test-Time Adaptation — arXiv:2609.32263
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.32263
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T22:36:22.706Z
- SHA-256 of the archived text
4bd88a279dbddd1346a4e04bd6a9330d7246ddccabdcbefbc3f3069db6df40ac- Stable URL of this page
https://metraipolis.com/documento/dc_63bc7dde8f70
How to cite
Representation Editing for Multimodal Test-Time Adaptation — arXiv:2609.32263 — arXiv (en), https://arxiv.org/abs/2609.32263 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_63bc7dde8f70 (Metraipolis archive dc_63bc7dde8f70, sha256 4bd88a279dbddd1346a4e04bd6a9330d7246ddccabdcbefbc3f3069db6df40ac)
BibTeX
@misc{dc63bc7dde8f70,
title = {Representation Editing for Multimodal Test-Time Adaptation — arXiv:2609.32263},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.32263},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_63bc7dde8f70; sha256 4bd88a279dbddd1346a4e04bd6a9330d7246ddccabdcbefbc3f3069db6df40ac},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32263. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal test-time adaptation (TTA) aims to adapt a pretrained multimodal model online to distribution shift across modalities using unlabeled test data, showing broad potential in real-world applications. However, existing methods primarily focus on adjusting fused features to bridge the source-target gap, lacking explicit control over intermediate representation misalignment, which is a key driver of performance drop under distribution shift. In this work, we tackle this challenge from the perspective of representation engineering. Unlike previous TTA methods that update fusion weights in place, we propose FourIer Representation Editor (FIRE), a novel multimodal TTA approach that directly edits semantically rich intermediate representations. Specifically, we first adopt representation editors into each intermediate layer of the unimodal encoders, enabling layer-wise calibration of unimodal representations. To further enhance the diversity and stability of the low-rank editing subspaces, each representation editor performs frequency domain mixing via the fast Fourier transform to construct structured bases. Moreover, we introduce multi-level adaptation objectives to optimize these editors, jointly promoting cross-modal semantic alignment, source-target statistical alignment, and asymmetric prediction consistency. In this way, FIRE yields aligned unimodal representations for fusion and further improves prediction reliability. Extensive experiments on two widely used multimodal benchmarks under various corruption types demonstrate the superiority of FIRE over existing multimodal TTA methods. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.32263 — 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 2026-10-07T22:14:27.623Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).