Visual Tokens
2 stories tagged “Visual Tokens”.
-
RenderRank: Learning Text Reranking from Compressed Visual Tokens
RenderRank reranks documents using compressed visual tokens, reducing input length 16.5-35.5% while achieving …
-
VisionWeave Creates Elastic Visual Representations as a Native Capability of Multimodal Large Language Models
VisionWeave uses elastic visual representations, reducing visual tokens 43.0% on Qwen3.8-27B while retaining 9…