Saved in:
Bibliographic Details
Main Authors: Wang, Lingfeng, Lin, Hualing, Chen, Senda, Wang, Tao, Cheng, Changxu, Zhong, Yangyang, Zheng, Dong, Zhao, Wuyue
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.16495
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908628844281856
author Wang, Lingfeng
Lin, Hualing
Chen, Senda
Wang, Tao
Cheng, Changxu
Zhong, Yangyang
Zheng, Dong
Zhao, Wuyue
author_facet Wang, Lingfeng
Lin, Hualing
Chen, Senda
Wang, Tao
Cheng, Changxu
Zhong, Yangyang
Zheng, Dong
Zhao, Wuyue
contents While humans effortlessly draw visual objects and shapes by adaptively allocating attention based on their complexity, existing multimodal large language models (MLLMs) remain constrained by rigid token representations. Bridging this gap, we propose ALTo, an adaptive length tokenizer for autoregressive mask generation. To achieve this, a novel token length predictor is designed, along with a length regularization term and a differentiable token chunking strategy. We further build ALToLLM that seamlessly integrates ALTo into MLLM. Preferences on the trade-offs between mask quality and efficiency is implemented by group relative policy optimization (GRPO). Experiments demonstrate that ALToLLM achieves state-of-the-art performance with adaptive token cost on popular segmentation benchmarks. Code and models are released at https://github.com/yayafengzi/ALToLLM.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16495
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ALTo: Adaptive-Length Tokenizer for Autoregressive Mask Generation
Wang, Lingfeng
Lin, Hualing
Chen, Senda
Wang, Tao
Cheng, Changxu
Zhong, Yangyang
Zheng, Dong
Zhao, Wuyue
Computer Vision and Pattern Recognition
While humans effortlessly draw visual objects and shapes by adaptively allocating attention based on their complexity, existing multimodal large language models (MLLMs) remain constrained by rigid token representations. Bridging this gap, we propose ALTo, an adaptive length tokenizer for autoregressive mask generation. To achieve this, a novel token length predictor is designed, along with a length regularization term and a differentiable token chunking strategy. We further build ALToLLM that seamlessly integrates ALTo into MLLM. Preferences on the trade-offs between mask quality and efficiency is implemented by group relative policy optimization (GRPO). Experiments demonstrate that ALToLLM achieves state-of-the-art performance with adaptive token cost on popular segmentation benchmarks. Code and models are released at https://github.com/yayafengzi/ALToLLM.
title ALTo: Adaptive-Length Tokenizer for Autoregressive Mask Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.16495