An image that is hard to classify on its own may have a helpful neighbor in the same unlabeled batch. Zero-shot vision-language models (VLMs) score images independently and leave this evidence unused. We study small unlabeled query batches with no labeled examples. Visual similarity can reflect pose or background, so i...
Jun-Xi Huang, Wen-Shuo Chen, Xiu-Cheng Wang et al.· 0 citations
BeamRMX is proposed, which is the first dedicated framework to treat the spatial radiation pattern as the primary BeamRM query and learn how scene geometry transforms it into the received power field, and shows consistent gains over deterministic and diffusion baselines.
Yue Zhang, Xiu-Cheng Wang, Wen-Shuo Chen et al.· 0 citations
ABLE (Attribution-Based Large-model Embedding), a framework that leverages the interpretability space to construct model representations by aggregating gradient-based feature attributions via a tokenizer-agnostic word-level alignment, captures model-specific input-sensitivity patterns rather than only surface-level out...