Multimodal Large Language Models (MLLM) have enabled a wide range of advanced vision-language applications, including fine-grained object recognition and contextual understanding. When querying specific regions or objects in an image, human users naturally use "Visual Prompts" (VP) like bounding boxes to provide reference. However, no existing benchmark systematically evaluates the ability of MLLMs to interpret such VPs. This gap raises uncertainty about whether current MLLMs can effectively recognize VPs, an intuitive prompting method for humans, and utilize them to solve problems. To address this limitation, we introduce VP-Bench, aiming to assess MLLMs’ capability in VP perception and utilization. VP-Bench employs a two-stage evaluation framework: Stage 1 examines models’ ability to perceive VPs in natural scenes, utilizing 100K visualized prompts spanning 8 shapes and 355 attribute combinations. Stage 2 investigates the impact of VPs on downstream tasks, measuring their effectiveness in real-world problem-solving scenarios. Using VP-Bench, we evaluate 21 MLLMs, including proprietary systems (e.g., GPT-4o) and open-source models (e.g., InternVL-2.5 and Qwen2.5-VL). In addition, we conduct a comprehensive analysis of the factors influencing VP understanding, such as attribute variations and model scale. VP-Bench establishes a new reference framework for studying MLLMs’ ability to comprehend and resolve grounded referring questions.

Controversial contents largely inundate the Internet, infringing various cultural norms and child protection standards. Traditional Image Content Moderation (ICM) models fall short in producing precise moderation decisions for diverse standards, while recent multimodal large language models (MLLMs), when adopted to general rule-based ICM, often produce classification and explanation results that are inconsistent with human moderators. Aiming at flexible, explainable, and accurate ICM, we design a novel rule-based dataset generation pipeline, decomposing concise human-defined rules and leveraging well-designed multi-stage prompts to enrich short explicit image annotations. Our ICM-Instruct dataset includes detailed moderation explanation and moderation Q-A pairs. Built upon it, we create our ICM-Assistant model in the framework of rule-based ICM, making it readily applicable in real practice. Our ICM-Assistant model demonstrates exceptional performance and flexibility. Specifically, it significantly outperforms existing approaches on various sources, improving both the moderation classification (36.8% on average) and moderation explanation quality (26.6% on average) consistently over existing MLLMs. Caution: Content includes offensive language or images.

Ziyi Guan, Jason Chun Lok Li, Zhijian Hou, Pingping Zhang, Donglai Xu, Yuzhi Zhao, Mengyang Wu, Jinpeng Chen, Thanh-Toan Nguyen, Pengfei Xian, Wenao Ma, Shengchao Qin, Graziano Chesi, Ngai Wong. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

We present PanoGPS, a new approach for generating indoor panoramas, capable of following a room layout sketch and description text to generate panoramas with geometrically-plausible room layouts and semantically-consistent contents. Overall, our idea is to inform the generative model collectively with the two inputs, to inspire it to implicitly generate and refine a unified latent code for panorama generation. Specifically, we first propose using the semantic layout map to encode the room geometry to condition the generative model with plausible geometric information. Second, we establish a pipeline to integrate features from the two inputs and generate a single unified latent code that describes both the room layout and room contents of the target panorama. Third, we progressively refine the latent code to produce a coarse panorama to enforce seamless left-right border connections, while iteratively upsizing the coarse panorama with details. Our method outperforms existing approaches by generating high-quality panoramas with geometrically plausible structures and semantically meaningful content. Code is available at: https://github.com/wumengyangok/PanoGPS

Scene Graph Generation (SGG) converts visual scenes into structured graph representations, providing deeper scene understanding for complex vision tasks. However, existing SGG models often overlook essential spatial relationships and struggle with generalization in open-vocabulary contexts. To address these limitations, we propose LLaVA-SpaceSGG, a multimodal large language model (MLLM) designed for open-vocabulary SGG with enhanced spatial relation modeling. To train it, we collect the SGG instruction-tuning dataset, named SpaceSGG. This dataset is constructed by combining publicly available datasets and synthesizing data using open-source models within our data construction pipeline. It combines object locations, object relations, and depth information, resulting in three data formats: spatial SGG description, question-answering, and conversation. To enhance the transfer of MLLMs' inherent capabilities to the SGG task, we introduce a two-stage training paradigm. Experiments show that LLaVA-SpaceSGG outperforms other open-vocabulary SGG methods, boosting recall by 8.6% and mean recall by 28.4% compared to the baseline. Our codebase, dataset, and trained models are publicly accessible on GitHub at the following URL: https://github.com/Endlinc/LLaVA-SpaceSGG.

The ability to recognize the position and order of the floor-level lines that divide adjacent building floors can benefit many applications, for example, urban augmented reality (AR). This work tackles the problem of locating floor-level lines in street-view images, using a supervised deep learning approach. Unfortunately, very little data is available for training such a network - current street-view datasets contain either semantic annotations that lack geometric attributes, or rectified facades without perspective priors. To address this issue, we first compile a new dataset and develop a new data augmentation scheme to synthesize training samples by harassing (i) the rich semantics of existing rectified facades and (ii) perspective priors of buildings in diverse street views. Next, we design FloorLevel-Net, a multi-task learning network that associates explicit features of building facades and implicit floor-level lines, along with a height-attention mechanism to help enforce a vertical ordering of floor-level lines. The generated segmentations are then passed to a second-stage geometry post-processing to exploit self-constrained geometric priors for plausible and consistent reconstruction of floor-level lines. Quantitative and qualitative evaluations conducted on assorted facades in existing datasets and street views from Google demonstrate the effectiveness of our approach. Also, we present context-aware image overlay results and show the potentials of our approach in enriching AR-related applications. Project website: https://wumengyangok.github.io/Project/FloorLevelNet.

This paper presents a new approach to recognizing vanishing-point-constrained building planes from a single image of street view. We first design a novel convolutional neural network (CNN) architecture that generates geometric segmentation of per-pixel orientations from a single street-view image. The network combines two-stream features of general visual cues and surface normals in gated convolution layers, and employs a deeply supervised loss that encapsulates multi-scale convolutional features. Our experiments on a new benchmark with fine-grained plane segmentations of real-world street views show that our network outperforms state-of-the-arts methods of both semantic and geometric segmentation. The pixel-wise segmentation exhibits coarse boundaries and discontinuities. We then propose to rectify the pixel-wise segmentation into perspectively-projected quads based on spatial proximity between the segmentation masks and exterior line segments detected through an image processing. We demonstrate how the results can be utilized to perspectively overlay images and icons on building planes in input photos, and provide visual cues for various applications.