Universal multimodal retrieval requires compact embeddings that preserve task-relevant semantic information across diverse modalities. Prior works have incorporated latent reasoning into multimodal embedding learning to refine this information before embedding extraction. However, most existing approaches remain confin...
Pei-Xi Wu, Ming-Zhou Jiang, Fei-Peng Ma et al.· 0 citations
Multimodal classifiers can converge to modality-dominant solutions in which one modality dominates the joint prediction, suppressing the learning of others. Existing balancing methods mainly adjust losses, gradients, or modality contributions, largely treating modality imbalance as an optimization problem while implici...
Ze-Chang Xiong, Da Li, Rong Yin et al.· 0 citations
Embeddings have emerged as a standard representational interface linking foundation models with downstream systems. Most embedding benchmarks assess representations through discriminative tasks or geometric criteria centered on separability in embedding space. However, strong performance on such evaluations does not es...
This work introduces Retrieval-Aware Self-Distillation (RASD), which extracts privileged guidance from input-supported facts and evidence distinguishing the positive from hard negatives, and Retrieval-Adaptive Inference (RAI), which learns a retrieval-aware stopping criterion from prefix-level retrieval feedback.
Ming-Zhou Jiang, Pei-Xi Wu, Hang Cheng et al.· 0 citations
MPAR-Bench is introduced, a bilingual English-Chinese benchmark that isolates reasoning breadth through multi-point associative reasoning, and case-level analysis shows that extended reasoning can overturn an initially correct hypothesis.
Si'an Xie, Jiaxu Liu, Biao Yang et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
