SigLIP-HD by Fine-to-Coarse Supervision
2026-07-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors studied how to get detailed visual information from images without needing bigger, higher-resolution pictures that require more computing power. They created a method called SigLIP-HD that teaches a lower-resolution image to imitate the detailed features of a higher-resolution one. This approach helps the model understand images better without extra computational cost. Their method was tested on several tasks and showed improved results, especially in reading text from images.
multimodal LLMsvisual representationhigh-resolution imagesfine-grained featuresimage processingSigLIP 2supervision designOCRinference budget
Authors
Lihe Yang, Zhen Zhao, Hengshuang Zhao
Abstract
High-quality visual representation is a long-standing pursuit in computer vision. In the context of multimodal LLMs (MLLMs), feeding higher-resolution images can produce more fine-grained visual tokens. However, it introduces additional computational and design complexity, due to multiple forward passes and post-processing of increased tokens. Before simply adopting a higher resolution, have we truly unlocked the model's full perception capability at a standard resolution? Therefore, we study an interesting problem: how to achieve fine visual perception under lower cost without larger images. We present SigLIP-HD in this work. The core is a highly simple fine-to-coarse supervision design. We enforce the coarse feature of a mid-resolution image to mimic the fine-grained feature of its high-resolution version. We build this framework on the advanced SigLIP 2 model. Our final model produces better visual tokens at exactly the same inference budget. It is validated on extensive MLLM benchmarks and consistently delivers stronger results than our baseline model, especially on OCR-related tasks.