Jiacong Wang

MuRA: Multi-Rank Adaptation for Efficient and Effective Test-Time Vision-Language Generalization

Gengyuan Liu, Nanzhou Wang, Chang Liu, Qinwen Wu et autres

Vision-language models exhibit remarkable zero-shot capabilities but suffer significant performance degradation under distribution shifts. While test-time adaptation (TTA) via Low-Rank Adaptation offers a parameter-efficient solution, we identify a fundamental bottleneck in current methods: the reliance on static rank configurations. Because visual inputs …

cn (code pays fourni par la source)

The Scalability of Simplicity: Empirical Analysis of Vision-Language Learning with a Single Transformer

Weixian Lei, Jiacong Wang, Haochen Wang, Xiangtai Li et autres

This paper introduces SAIL, a single transformer unified multimodal large language model (MLLM) that integrates raw pixel encoding and language decoding within a singular architecture. Unlike existing modular MLLMs, which rely on a pre-trained vision transformer (ViT), SAIL eliminates the need for …

pl (code pays fourni par la source)

UVE: Are MLLMs Unified Evaluators for AI-Generated Videos?

Yuanxin Liu, Rui Zhu, Shuhuai Ren, Jiacong Wang et autres

With the rapid growth of video generative models (VGMs), it is essential to develop reliable and comprehensive automatic metrics for AI-generated videos (AIGVs). Existing methods either use off-the-shelf models optimized for other tasks or rely on human assessment data to train specialized …

cn, gb (code pays fourni par la source)

Benchmarking and Improving Detail Image Caption

Hongyuan Dong, Jiawen Li, Bohong Wu, Jiacong Wang et autres

Image captioning has long been regarded as a fundamental task in visual understanding. Recently, however, few large vision-language model (LVLM) research discusses model's image captioning performance because of the outdated short-caption benchmarks and unreliable evaluation metrics. In this work, we propose to …

Seeing the Image: Prioritizing Visual Correlation by Contrastive Alignment

Xin Xiao, Bohong Wu, Jiacong Wang, Chunyuan Li et autres

Existing image-text modality alignment in Vision Language Models (VLMs) treats each text token equally in an autoregressive manner. Despite being simple and effective, this method results in sub-optimal cross-modal alignment by over-emphasizing the text tokens that are less correlated with or even …

Unveiling the Tapestry of Consistency in Large Vision-Language Models

Yuan Zhang, Fei Xiao, Tao Huang, Chun-Kai Fan et autres

Large vision-language models (LVLMs) have recently achieved rapid progress, exhibiting great perception and reasoning abilities concerning visual information. However, when faced with prompts in different sizes of solution spaces, LVLMs fail to always give consistent answers regarding the same knowledge point. This …

Unveiling the Tapestry of Consistency in Large Vision-Language Models

Fei Xiao, Tao Huang, Chun-Kai Fan, Hongyuan Dong et autres

Seeing the Image: Prioritizing Visual Correlation by Contrastive Alignment

Xin Xiao, Bohong Wu, Jiacong Wang, Xun Zhou et autres

Poisson-based image editing for semi-supervised vitiligo lesion segmentation with limited annotations

Jiacong Wang, Xiaolan Ding, Jun Xiao

cn (code pays fourni par la source)

Towards Real-time Analysis of Marine Phytoplankton Images Sampled at High Frame Rate by a YOLOX-based Object Detection Algorithm

Jiacong Wang, Cheng Tang, Jianping Li

Rapid and quantitative analysis of phytoplankton cells in natural seawater is of great need for marine ecological science research and harmful algae bloom monitoring applications. In this paper, we propose a YOLOX network-based object detection algorithm exclusively for high-throughput real-time analysis of …

cn (code pays fourni par la source)