RiVaT-Fuse: Reliability-Calibrated Variational Tensor Fusion for Multimodal Prediction under Modality Uncertainty
This story is from 2026-09-11. It is preserved in the archive; the latest stories are on the live feed.
arXiv:2609.10798v1 Announce Type: new Abstract: Image-metadata prediction requires fusing heterogeneous evidence whose reliability can vary across samples and latent factors. Existing representation-level fusion methods typically choose an aggregation architecture, such as concatenation, gating, co…
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-09-11 04:00 · arXiv cs.LG
RiVaT-Fuse: Reliability-Calibrated Variational Tensor Fusion for Multimodal Prediction under Modality Uncertainty