Chart Question Answering (CQA) evaluates Multimodal Large Language Models (MLLMs) on visual understanding and reasoning over chart data. However, existing benchmarks mostly test surface-level parsing, such as reading labels and legends, while overlooking deeper scientific reasoning. We propose DomainCQA, a framework for constructing domain-specific CQA benchmarks that emphasize both visual comprehension and knowledge-intensive reasoning. It integrates complexity-aware chart selection, multitier QA generation, and expert validation. Applied to astronomy, DomainCQA yields AstroChart, a benchmark of 1,690 QA pairs over 482 charts, exposing persistent weaknesses in fine-grained perception, numerical reasoning, and domain knowledge integration across 21 MLLMs. Fine-tuning on AstroChart improves performance across fundamental and advanced tasks. Pilot QA sets in biochemistry, economics, medicine, and social science further demonstrate DomainCQA’s generality. Together, our results establish DomainCQA as a unified pipeline for constructing and augmenting domain-specific chart reasoning benchmarks.

Visual Prompt Tuning (VPT) is a parameter-efficient fune-tuning technique that adapts a pre-trained vision Transformer (ViT) by learning a small set of parameters in the input space, known as prompts. In VPT, we uncover ``burstiness'' in the values arising from the interaction of image patch embeddings, and the key and query projectors within Transformer's self-attention module. Furthermore, the values of patch embeddings and the key and query projectors exhibit Laplacian and hyper-Laplacian distribution, respectively. Intuitively, these non-Gaussian distributions pose challenges for learning prompts. To address this, we propose whitening these data, de-correlating them and equalizing their variance towards more Gaussian before learning prompts. We derive the whitening matrix over random image patch embeddings and ViT's key and query projectors, and multiply it with the prompt to be learned in a bilinear manner. Surprisingly, this method significantly accelerates prompt tuning and boosts accuracy, e.g., $>$25 accuracy points on the CUB dataset; interestingly, it learns ``bursty prompts''. Extending the bilinear model which is known to introduce burstiness, we present a compact, low-rank version by learning two smaller matrices whose multiplication yields the final prompts. We call the proposed methods Bilinear Prompt Tuning (BPT). Extensive experiments across multiple benchmark datasets demonstrate that BPT methods not only outperform various VPT methods but also reduce parameter count and computation overhead.

Large language models (LLMs) demonstrate extraordinary abilities in a wide range of natural language processing (NLP) tasks. In this paper, we show that, beyond text understanding capability, LLMs are capable of processing text layouts that are denoted by spatial markers. They are able to answer questions that require explicit spatial perceiving and reasoning, while a drastic performance drop is observed when the spatial markers from the original data are excluded. We perform a series of experiments with the GPT-3.5/4, Baichuan2, Llama2, and ChatGLM3 models on various types of layout-sensitive datasets for further analysis. The experimental results reveal that the layout understanding ability of LLMs is mainly introduced by the coding data for pre-training, which is further enhanced at the instruction-tuning stage. In addition, layout understanding can be enhanced by integrating low-cost, auto-generated data approached by a novel text game. Finally, we show that layout understanding ability is beneficial for building efficient visual question-answering (VQA) systems.

Visual prompt tuning (VPT) is a promising solution incorporating learnable prompt tokens to customize pre-trained models for downstream tasks. However, VPT and its variants often encounter challenges like prompt initialization, prompt length, and subpar performance in self-supervised pretraining, hindering successful contextual adaptation. This study commences by exploring the correlation evolvement between prompts and patch tokens during proficient training. Inspired by the observation that the prompt tokens tend to share high mutual information with patch tokens, we propose initializing prompts with downstream token prototypes. The strategic initialization, a stand-in for the previous initialization, substantially improves performance. To refine further, we optimize token construction with a streamlined pipeline that maintains excellent performance with almost no increase in computational expenses compared to VPT. Exhaustive experiments show our proposed approach outperforms existing methods by a remarkable margin. For instance, after MAE pre-training, our method improves accuracy by up to 10%$\sim$30% compared to VPT, and outperforms Full fine-tuning 19 out of 24 cases while using less than 0.4% of learnable parameters. Besides, the experimental results demonstrate the proposed SPT is robust to prompt lengths and scales well with model capacity and training data size. We finally provide an insightful exploration into the amount of target data facilitating the adaptation of pre-trained models to downstream tasks. The code is available at https://github.com/WangYZ1608/Self-Prompt-Tuning.

Interactive segmentation allows users to extract target masks by making positive/negative clicks. Although explored by many previous works, there is still a gap between academic approaches and industrial needs: first, existing models are not efficient enough to work on low-power devices; second, they perform poorly when used to refine preexisting masks as they could not avoid destroying the correct part. FocalClick solves both issues at once by predicting and updating the mask in localized areas. For higher efficiency, we decompose the slow prediction on the entire image into two fast inferences on small crops: a coarse segmentation on the Target Crop, and a local refinement on the Focus Crop. To make the model work with preexisting masks, we formulate a sub-task termed Inter-active Mask Correction, and propose Progressive Merge as the solution. Progressive Merge exploits morphological information to decide where to preserve and where to update, enabling users to refine any preexisting mask effectively. FocalClick achieves competitive results against SOTA methods with significantly smaller FLOPs. It also shows significant superiority when making corrections on preexisting masks. Code and data will be released at github.com/XavierCHEN34/ClickSEG

In click-based interactive segmentation, the mask extraction process is dictated by positive/negative user clicks; however, most existing methods do not fully exploit the user cues, requiring excessive numbers of clicks for satisfactory results. We propose Conditional Diffusion Network (CDNet), which propagates labeled representations from clicks to conditioned destinations with two levels of affinities: Feature Diffusion Module (FDM) spreads features from clicks to potential target regions with global similarity; Pixel Diffusion Module (PDM) diffuses the predicted logits of clicks within locally connected regions. Thus, the information inferred by user clicks could be generalized to proper destinations. In addition, we put forward Diversified Training (DT), which reduces the optimization ambiguity caused by click simulation. With FDM,PDM and DT, CDNet could better understand user’s intentions and make better predictions with limited interactions. CDNet achieves state-of-the-art performance on several benchmarks.

Geographical context is very important for images. Millions of images on the Web have been already assigned latitude and longitude information. Due to the rapid proliferation of such images with geographical context, it is still difficult to effectively search and browse them, since we do not have ways to decide their relevance. In this paper, we focus on the geographical relevance of images, which is defined as to what extent the main objects in an image match landmarks at the location where the image was taken. Recently, researchers have proposed to use game based approaches to label large scale data such as Web images. However, previous works have not shown the quality of collected game logs in detail and how the logs can improve existing applications. To answer these questions, we design and implement a Web-based and multi-player game to collect human knowledge while people are enjoying the game. Then we thoroughly analyze the game logs obtained during a three week study with 147 participants and propose methods to determine the image geographical relevance. In addition, we conduct an experiment to compare our methods with a commercial search engine. Experimental results show that our methods dramatically improve image search relevance. Furthermore, we show that we can derive geographically relevant objects and their salient portion in images, which is valuable for a number of applications such as image location recognition.

Recently, there has been growing interest in mining co-location visual patterns from a collection of images. To find a proper usage of visual patterns in near-duplicate image retrieval systems, we study a TF-IDF weighting function for visual patterns. We show usage of TF and IDF respectively in this weighting function. Experiments demonstrate that 1) visual patterns and words should be weighted separately; 2) visual patterns consist of more frequent visual words would be more important; 3) visual patterns should be given lower weight values than visual words in near-duplicate image retrieval tasks.