SLICEChat: Progressive In-Encoder Token Pruning for Whole-Slide Pathology Language Models
SLICEChat prunes pathology slide tokens inside a hybrid Mamba-Transformer encoder before multimodal fusion.
SLICEChat is a slide-level multimodal model that progressively prunes tokens inside a hybrid Mamba-Transformer encoder for gigapixel whole-slide pathology images. Language-supervised, region-aware pruning removes spatially coherent low-utility regions on a keep-rate schedule before multimodal fusion. On SlideBench VQA it reaches 79.84% accuracy on TCGA and 59.09% on BCNB, outperforming prior slide-level pathology MLLMs, and leads overall WSI-Bench metrics with competitive memory use and latency.
- In-encoder pruning shortens slide tokens before multimodal fusion.
- Hybrid Mamba-Transformer encoder combines long-range and global layers.
- SlideBench VQA: 79.84% on TCGA and 59.09% on BCNB.
- Highest overall WSI-Bench metrics with competitive memory and latency.
Full article149 words · extracted from arxiv.org · click to collapse
Whole-slide pathology images (WSIs) contain gigapixel-scale visual content, creating a major scalability challenge for slide-level multimodal large language models (MLLMs). Existing approaches process thousands of patch tokens and typically apply compression only after slide encoding, leaving multimodal attention computationally expensive. We introduce SLICEChat, a slide-level MLLM that integrates progressive token pruning within a hybrid Mamba--Transformer slide encoder. Mamba layers enable efficient long-range propagation, while Transformer layers preserve global interactions as the sequence is progressively shortened. Between stages, language-supervised, region-aware pruning removes spatially coherent low-utility regions under a controlled keep-rate schedule, producing compact slide representations before multimodal fusion. On SlideBench VQA, SLICEChat achieves 79.84% accuracy on TCGA and 59.09% on BCNB cohorts, outperforming prior slide-level pathology MLLMs, and achieves the highest overall WSI-Bench metrics. It also provides competitive memory usage and the inference latency among the evaluated models. These results demonstrate accurate and computationally efficient multimodal reasoning over gigapixel WSIs.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.24894