Scientific Figure Analysis (SFA) aims to derive analytical insights from figures while incorporating background instructions. Unlike conventional tasks such as figure captioning or description generation, which focus on extracting surface-level information from the sole visual modality, SFA requires an intelligent system to summarize key patterns, infer implications, and contextualize scientific findings from visual and textual inputs. It demands not only visual recognition but also the integration of scientific knowledge, multimodal understanding, and contextual reasoning. In this work, we introduce an SFA dataset, AnaFig, comprising 2,000 high-quality samples across 56 domains. All samples are evaluated by using human-aligned five-dimensional scoring criteria, resulting 10,000 human-annotated score labels. The AnaFig dataset facilitates the assessment of three critical capabilities of multimodal large language models (MLLMs): adherence to complex instructions, multimodal perception, and analytical summarization. By building a new benchmark with widely used MLLMs, this study contributes to scientific knowledge discovery and reasoning, fostering the alignment of MLLMs and human experts in scientific analysis.
论文检索
输入标题、作者或关键词，从 12,319 篇学术成果中精准定位
Environmental, Social, and Governance (ESG) reports are essential for assessing sustainability, regulatory compliance, and financial transparency. However, these documents are typically long, multimodal, and structurally complex, combining dense text, tables, figures, and layout-sensitive semantics. Existing AI systems often struggle to perform reliable document-level reasoning in such settings, and no dedicated benchmark currently exists in ESG domain. To fill the gap, we introduce MMESGBench, a first-of-its-kind benchmark dataset targeted to evaluate multimodal understanding and reasoning across multi-source ESG documents. This dataset is constructed via a human-AI collaborative, multi-stage pipeline. First, a multimodal LLM generates candidate question-answer (QA) pairs by jointly interpreting textual, tabular, and visual information from layout-aware document pages. Second, an LLM verifies the semantic accuracy, completeness, and reasoning complexity of each QA pair. This automated process is followed by an expert-in-the-loop validation, where domain specialists validate and calibrate QA pairs to ensure quality, relevance, and diversity. MMESGBench comprises 933 validated QA pairs derived from 45 ESG documents, spanning across seven distinct document types and three major ESG source categories. Questions are categorized as single-page, cross-page, or unanswerable, with each accompanied by fine-grained multimodal evidence. Initial experiments validate that multimodal and retrieval-augmented models substantially outperform text-only baselines. MMESGBench is publicly available as an open-source dataset at https://github.com/Zhanglei1103/MMESGBench.
Vision-Language Models (VLMs) have recently emerged, demonstrating remarkable vision-understanding capabilities. However, training these models requires large-scale datasets, which brings challenges related to efficiency, effectiveness, and quality of web data. In this paper, we introduce SynthVLM, a new data synthesis and curation method for generating image-caption pairs. Unlike traditional methods, where captions are generated from images, SynthVLM utilizes advanced diffusion models and high-quality captions to synthesize and select images from text captions, thereby creating precisely aligned image-text pairs. We further introduce SynthVLM-100K, a high-quality dataset consisting of 100K curated and synthesized image-caption pairs. In both model and human evaluations, SynthVLM-100K outperforms traditional real-world datasets. Leveraging this dataset, we develop a new family of multimodal large language models (MLLMs), SynthVLM-7B and SynthVLM-13B, which achieve state-of-the-art (SOTA) performance on various vision question-answering (VQA) tasks. Notably, our models outperform LLaVA across most metrics with only 18% pretrain data. Furthermore, SynthVLM-7B and SynthVLM-13B attain SOTA performance on the MMLU benchmark, demonstrating that the high-quality SynthVLM-100K dataset preserves language abilities. Our dataset and the complete data generating and curating methods can be found in https://github.com/starriver030515/SynthVLM.
Visual parsing of images and videos is critical for a wide range of real-world applications. However, progress in this field is constrained by limitations of existing datasets: (1) limited annotation diversity, which limits the support for diverse vision tasks within a unified dataset; (2) insufficient coverage of domains, particularly a lack of datasets tailored for educational scenarios; and (3) a lack of explicit procedural guidance, with weak logical rules and insufficient representation of a structured task process. To address these gaps, we introduce PhysLab, the first dataset that captures students conducting complex physics experiments. The dataset includes four representative experiments that feature diverse scientific instruments and rich human-object interaction (HOI) patterns. PhysLab comprises 620 long-form videos and provides multi-granularity annotations that support a variety of vision tasks, including action recognition, object detection, HOI analysis, etc. We establish baselines and perform extensive evaluations to highlight key challenges in the parsing of procedural educational videos. We expect PhysLab to serve as a valuable resource for advancing comprehensive visual parsing, facilitating intelligent classroom systems, and fostering closer integration among computer vision, multimedia, and educational technologies. The dataset and the evaluation toolkit are publicly available at https://github.com/ZMH-SDUST/PhysLab.
The development of multi-modal Unmanned Aerial Vehicles (UAVs) environment perception systems is hindered by three critical gaps in existing datasets: (1) insufficient modalities and pixel misalignment, (2) noisy labels, and (3) limited task types. To address these gaps, we propose an automatic data construction approach and construct a multi-modal UAV-based environment perception dataset, UEMM-Air. Its synthetic nature ensures scalability, reproducibility, and rare-event coverage, making it suitable for large-scale model pre-training. Benefiting from our automated data collection and annotation pipeline, UEMM-Air encompasses 120k data pairs across 6 aligned modalities and supports 4 perception tasks, significantly exceeding existing datasets (max 60k data, 3 modalities, 2 tasks). Compared to existing synthetic datasets like SynDrone, UEMM-Air provides more accurate annotations by avoiding noisy labels from direct coordinate computation. Notably, models pre-trained on UEMM-Air achieve a 5.8% accuracy improvement compared to those utilizing other synthetic datasets, while requiring less than half the data. This benchmark establishes performance evaluation of UAV multi-modal environmental perception models, and hopefully encourages more research efforts towards enabling UAVs to undertake more multi-modal tasks. The dataset and its generation engine are openly accessible under a permissive license at https://github.com/1e12Leon/UEMM-Air.
Spatial reasoning is a fundamental capability of multimodal large language models (MLLMs), yet their performance in open aerial environments remains underexplored. In this work, we present Open3D-VQA, a novel benchmark for evaluating MLLMs' ability to reason about complex spatial relationships from an aerial perspective. The benchmark comprises 73k QA pairs across seven general spatial reasoning tasks, offered in multiple-choice, true/false, and short-answer formats, and supports both visual and point cloud modalities. The questions are automatically generated from spatial relations extracted from both real-world and simulated aerial scenes. Evaluation on 13 popular MLLMs reveals that: 1) Models are generally better at answering questions about relative spatial relations than absolute distances, 2) 3D LLMs fail to demonstrate significant advantages over 2D LLMs, and 3) Fine-tuning solely on the simulated dataset can significantly improve the model's spatial reasoning performance in real-world scenarios. The benchmark, generation pipeline, and evaluation toolkit are released on this page.
Fine-grained analysis of complex and high-speed sports like badminton presents a significant challenge for Multimodal Large Language Models (MLLMs), despite their notable advancements in general video understanding. This difficulty arises primarily from the scarcity of datasets with sufficiently rich and domain-specific annotations. To bridge this gap, we introduce FineBadminton, a novel and large-scale dataset featuring a unique multi-level semantic annotation hierarchy (Foundational Actions, Tactical Semantics, and Decision Evaluation) for comprehensive badminton understanding. The construction of FineBadminton is powered by an innovative annotation pipeline that synergistically combines MLLM-generated proposals with human refinement. We also present FBBench, a challenging benchmark derived from FineBadminton, to rigorously evaluate MLLMs on nuanced spatio-temporal reasoning and tactical comprehension. Together, FineBadminton and FBBench provide a crucial ecosystem to catalyze research in fine-grained video understanding and advance the development of MLLMs in sports intelligence. Furthermore, we propose an optimized baseline approach incorporating Hit-Centric Keyframe Selection to focus on pivotal moments and Coordinate-Guided Condensation to distill salient visual information. The results on FBBench reveal that while current MLLMs still face significant challenges in deep sports video analysis, our proposed strategies nonetheless achieve substantial performance gains. The project homepage is available at https://finebadminton.github.io/FineBadminton/.
In an NBA game scenario, consider the challenge of locating and analyzing the 3D poses of players performing a user-specified action, such as attempting a shot. Traditional 3D human pose estimation (3DHPE) methods often fall short in such complex, multi-person scenes due to their lack of semantic integration and reliance on isolated pose data. To address these limitations, we introduce Language-Driven 3D Human Pose Estimation (L3DHPE), a novel approach that extends 3DHPE to general multi-person contexts by incorporating detailed language descriptions. We present Panoptic-L3D, the first dataset designed for L3DHPE, featuring 3,838 linguistic annotations for 1,476 individuals across 588 videos, with 6,035 masks and 91k frame-level 3D skeleton annotations. Additionally, we propose Cascaded Pose Perception (CPP), a benchmarking method that simultaneously performs language-driven mask segmentation and 3D pose estimation within a unified model. CPP first learns 2D pose information, utilizes a body fusion module to aid in mask segmentation, and employs a mask fusion module to mitigate mask noise before outputting 3D poses. Extensive evaluation of CPP and existing benchmarks on Panoptic-L3D demonstrates the necessity of this novel task and dataset for advancing 3DHPE. Our dataset is available at https://languagedriven3dposeestimation.github.io/.
Accurate and robust 3D hand pose estimation (HPE) plays a crucial role in human-computer interaction. Existing 3D HPE solutions predominantly rely on vision-based or inertial measurement units (IMUs)-based methods. Vision-based methods benefit from rich appearance information for high-accuracy HPE but are sensitive to field of view (FoV), occlusion, motion blur and lighting. IMU-based methods can operate immune to optical sensitivity and FoV constraints but remain vulnerable to cumulative integration errors and drift. Given their complementary strengths, combining dual modalities offers a promising direction for HPE in complex environments. However, the lack of large-scale visual-inertial datasets has limited progress in this area. In this paper, we construct VIHand, the first large-scale glove-worn dataset for visual-inertial HPE, comprising over 1.4 million synchronized RGB-D and IMU frames from 15 subjects. It enables comprehensive research in HPE tasks, such as multimodal fusion and cross-modal knowledge transfer. Building on VIHand, we propose visual-inertial fusion network (VIFNet) for dual-modalities estimation, and its distilled student model (VIFNet-S) for IMU-only evaluation. Experimental results reveal that integrating visual and inertial modalities significantly improves the accuracy and robustness of 3D HPE, particularly under occlusion and motion blur. In IMU-only inference even sparse IMU configurations, models distilled from visual-inertial supervision achieve substantial performance gains, enabling robust HPE for challenging optical sensitive scenarios. Our dataset and supplementary materials are available on the project website: https://shirley0118.github.io/VIHand.
Video content comprehension is essential for various applications, ranging from video analysis to interactive systems. Despite advancements in large-scale vision-language models (VLMs), these models often struggle to capture the nuanced, spatiotemporal details essential for thorough video analysis. To address this gap, we introduce Video-CoT, a groundbreaking dataset designed to enhance spatiotemporal understanding using Chain-of-Thought(CoT) methodologies. Video-CoT contains 192,000 fine-grained spatiotemporal question-answer pairs and 23,000 high-quality CoT-annotated samples, providing a solid foundation for evaluating spatiotemporal understanding in video comprehension. Addition- ally, we provide a comprehensive benchmark for assessing these tasks, with each task featuring 750 images and tailored evaluation metrics. Our extensive experiments reveal that current VLMs face significant challenges in achieving satisfactory performance, high- lighting the difficulties of effective spatiotemporal understanding. Overall, the Video-CoT dataset and benchmark open new avenues for research in multimedia understanding and support future innovations in intelligent systems requiring advanced video analysis capabilities. By making these resources publicly available, we aim to encourage further exploration in this critical area. Project website: https://video-cot.github.io/ .
AI-based image enhancement techniques have been widely adopted in various visual applications, significantly improving the perceptual quality of user-generated content (UGC). However, the lack of specialized quality assessment models has become a significant limiting factor in this field, limiting user experience and hindering the advancement of enhancement methods. While perceptual quality assessment methods have shown strong performance on UGC and AIGC individually, their effectiveness on AI-enhanced UGC (AI-UGC) which blends features from both-remains largely unexplored. To address this gap, we construct AU-IQA, a benchmark dataset comprising 4,800 AI-UGC images produced by three representative enhancement types which include super-resolution, low-light enhancement, and denoising. On this dataset, we further evaluate a range of existing quality assessment models, including traditional IQA methods and large multimodal models. Finally, we provide a comprehensive analysis of how well current approaches perform in assessing the perceptual quality of AI-UGC. The access link to the AU-IQA is https://github.com/WNNGGU/AU-IQA-Dataset.
With the increasing integration of Multimodal Large Language Models (MLLMs) into the medical field, comprehensive evaluation of their performance in various medical domains becomes critical. However, existing benchmarks primarily assess general medical tasks, inadequately capturing performance in nuanced areas like the spine, which relies heavily on visual input. To address this, we introduce SpineBench, a comprehensive Visual Question Answering (VQA) benchmark designed for fine-grained analysis and evaluation of MLLMs in the spinal domain. SpineBench comprises 64,878 QA pairs from 40,263 spine images, covering 11 spinal diseases through two critical clinical tasks: spinal disease diagnosis and spinal lesion localization, both in multiple-choice format. SpineBench is built by integrating and standardizing image-label pairs from open-source spinal disease datasets, and samples challenging hard negative options for each VQA pair based on visual similarity (similar but not the same disease), simulating real-world challenging scenarios. We evaluate 12 leading MLLMs on SpineBench. The results reveal that these models exhibit poor performance in spinal tasks, highlighting limitations of current MLLM in the spine domain and guiding future improvements in spinal medicine applications. SpineBench is publicly available at https://zhangchenghanyu.github.io/SpineBench.github.io/.
Early-stage fire scenes (0-15 minutes after ignition) represent a crucial temporal window for emergency interventions. During this stage, the smoke produced by combustion significantly reduces the visibility of surveillance systems, severely impairing situational awareness and hindering effective emergency response and rescue operations. Consequently, there is an urgent need to remove smoke from images to obtain clear scene information. However, the development of smoke removal algorithms remains limited due to the lack of large-scale, real-world datasets comprising paired smoke-free and smoke-degraded images. To address these limitations, we present a real-world surveillance image desmoking benchmark dataset named SmokeBench, which contains image pairs captured under diverse scenes setup and smoke concentration. The curated dataset provides precisely aligned degraded and clean images, enabling supervised learning and rigorous evaluation. We conduct comprehensive experiments by benchmarking a variety of desmoking methods on our dataset. Our dataset provides a valuable foundation for advancing robust and practical image desmoking in real-world fire scenes. This dataset has been released to the public and can be downloaded from https://github.com/ncfjd/SmokeBench.
Singing accent research is underexplored compared to speech accent studies, primarily due to the scarcity of suitable datasets. Existing singing datasets often suffer from detail loss, frequently resulting from the vocal-instrumental separation process. Additionally, they often lack regional accent annotations. To address this, we introduce the Multi-Accent Mandarin Dry-Vocal Singing Dataset (MADVSD). MADVSD comprises over 670 hours of dry vocal recordings from 4,026 native Mandarin speakers across nine distinct Chinese regions. In addition to each participant recording audio of three popular songs in their native accent, they also recorded phonetic exercises covering all Mandarin vowels and a full octave range. We validated MADVSD through benchmark experiments in singing accent recognition, demonstrating its utility for evaluating state-of-the-art speech models in singing contexts. Furthermore, we explored dialectal influences on singing accent and analyzed the role of vowels in accentual variations, leveraging MADVSD's unique phonetic exercises.
Robot manipulation is a fundamental capability of embodied intelligence, enabling effective robot interactions with the physical world. In robotic manipulation tasks, predicting precise grasping positions and object placement is essential. Achieving this requires object recognition to localize target object, predicting object affordances for interaction and spatial affordances for optimal arrangement. While Vision-Language Models (VLMs) provide insights for high-level task planning and scene understanding, they often struggle to predict precise action positions, such as functional grasp points and spatial placements. This limitation stems from the lack of annotations for object and spatial affordance data in their training datasets. To address this gap, we introduce RoboAfford , a novel large-scale dataset designed to enhance object and spatial affordance learning in robot manipulation. Our dataset comprises 819,987 images paired with 1.9 million question answering (QA) annotations, covering three critical tasks: object affordance recognition to identify objects based on attributes and spatial relationships, object affordance prediction to pinpoint functional grasping parts, and spatial affordance localization to identify free space for placement. Complementing this dataset, we propose RoboAfford-Eval , a comprehensive benchmark for assessing affordance-aware prediction in real-world scenarios, featuring 338 meticulously annotated samples across the same three tasks. Extensive experimental results reveal the deficiencies of existing VLMs in affordance learning, while fine-tuning on the RoboAfford dataset significantly enhances their affordance prediction in robot manipulation, validating the dataset's effectiveness. The dataset, benchmark and evaluation code will be made publicly available to facilitate future research. Project website: https://roboafford-dataset.github.io/.
With the rapid development of MLLMs, evaluating their visual capabilities has become increasingly crucial. Current benchmarks primarily fall into two main types: basic perception benchmarks,which focus on local details but lack deep reasoning (e.g., ''what is in the image?''), and mainstream reasoning benchmarks, which concentrate on prominent image elements but may fail to assess subtle clues requiring intricate analysis. However, profound visual understanding and complex reasoning depend more on interpreting subtle, inconspicuous local details than on perceiving salient, macro-level objects. These details, though occupying minimal image area, often contain richer, more critical information for robust analysis. To bridge this gap, we introduce the VER-Bench, a novel framework to evaluate MLLMs' ability to: 1) identify fine-grained visual clues, often occupying, on average, just 0.25% of the image area; 2) integrate these clues with world knowledge for complex reasoning. Comprising 374 carefully designed questions across Geospatial, Temporal, Situational, Intent, System State, and Symbolic reasoning, each question in VER-Bench is accompanied by structured evidence: visual clues and question-related reasoning derived from them. VER-Bench reveals current models' limitations in extracting subtle visual evidence and constructing evidence-based reasoning chains, highlighting the need to enhance models' capabilities in fine-grained visual evidence extraction, integration, and reasoning for genuine visual understanding and human-like analysis. The dataset is available at https://github.com/verbta/ACMMM-25-Materials.
The rapid evolution of Multi-modality Large Language Models (MLLMs) is driving significant advancements in visual understanding and generation. Nevertheless, a comprehensive assessment of their capabilities, concerning the fine-grained physical principles especially in geometric optics, remains underexplored. To address this gap, we introduce GOBench, the first benchmark to systematically evaluate MLLMs' ability across two tasks: 1) Generating Optically Authentic Imagery and 2) Understanding Underlying Optical Phenomena. We curate high-quality prompts of geometric optical scenarios and use MLLMs to construct the GOBench-Gen-1k dataset. We then organize subjective experiments to assess the generated imagery based on Optical Authenticity, Aesthetic Quality, and Instruction Fidelity, revealing MLLMs' generation flaws that violate optical principles. For the understanding task, we apply crafted evaluation instructions to test the optical understanding ability of eleven prominent MLLMs. The experimental results demonstrate that current models face significant challenges in both optical generation and understanding. The top-performing generative model, GPT-4o-Image, cannot perfectly complete all generation tasks, and the best-performing MLLM model, Gemini-2.5Pro, attains a mere 37.35% accuracy in optical understanding. Database and codes are publicly available at: https://github.com/aiben-ch/GOBench.
3D Gaussian Splatting (3DGS) has emerged as a promising approach for novel view synthesis, offering real-time rendering with high visual fidelity. However, its substantial storage requirements present significant challenges for practical applications. While recent state-of-the-art (SOTA) 3DGS methods increasingly incorporate dedicated compression modules, there is a lack of a comprehensive framework to evaluate their perceptual impact. Therefore we present 3DGS-IEval-15K, the first large-scale image quality assessment (IQA) dataset specifically designed for compressed 3DGS representations. Our dataset encompasses 15,200 images rendered from 10 real-world scenes through 6 representative 3DGS algorithms at 20 strategically selected viewpoints, with different compression levels leading to various distortion effects. Through controlled subjective experiments, we collect human perception data from 60 viewers. We validate dataset quality through scene diversity and MOS distribution analysis, and establish a comprehensive benchmark with 30 representative IQA metrics covering diverse types. As the largest-scale 3DGS quality assessment dataset to date, our work provides a foundation for developing 3DGS specialized IQA metrics, and offers essential data for investigating view-dependent quality distribution patterns unique to 3DGS. The database is publicly available at https://github.com/YukeXing/3DGS-IEval-15K.
With the rapid advancement of generative models, the realism of AI-generated images has significantly improved, posing critical challenges for verifying digital content authenticity. Current deepfake detection methods often depend on datasets with limited generation models and content diversity that fail to keep pace with the evolving complexity and increasing realism of the AI-generated content. Large multimodal models (LMMs), widely adopted in various vision tasks, have demonstrated strong zero-shot capabilities, yet their potential in deepfake detection remains largely unexplored. To bridge this gap, we present DFBench, a large-scale DeepFake Benchmark featuring (i) broad diversity, including 540,000 images across real, AI-edited, and AI-generated content, (ii) latest model, the fake images are generated by 12 state-of-the-art generation models, and (iii) bidirectional benchmarking and evaluating for both the detection accuracy of deepfake detectors and the evasion capability of generative models. Based on DFBench, we propose MoA-DF, Mixture of Agents for DeepFake detection, leveraging a combined probability strategy from multiple LMMs. MoA-DF achieves state-of-the-art performance, further proving the effectiveness of leveraging LMMs for deepfake detection. Database and codes are publicly available at https://github.com/IntMeGroup/DFBench.
Creating recipe images is a key challenge in food computing, with applications in culinary education and multimodal recipe assistants. However, existing datasets lack fine-grained alignment between recipe goals, step-wise instructions, and visual content. We present RecipeGen, the first large-scale, real-world benchmark for recipe-based Text-to-Image (T2I), Image-to-Video (I2V), and Text-to-Video (T2V) generation. RecipeGen contains 26,435 recipes, 196,724 images, and 4,491 videos, covering diverse ingredients, cooking procedures, styles, and dish types. We further propose domain-specific evaluation metrics to assess ingredient fidelity and interaction modeling, benchmark representative T2I, I2V, and T2V models, and provide insights for future recipe generation models. Project page is available at https://wenbin08.github.io/RecipeGen.