Ruohong Zhang, Bowen Zhang, Yanghao Li, Haotian Zhang, Zhiqing Sun, Zhe Gan, Yinfei Yang, Ruoming Pang, Yiming Yang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent advancements in multimodal models highlight the value of rewritten captions for improving performance, yet key challenges remain. For example, while synthetic captions often provide superior quality and image-text alignment, it is not clear whether they can fully replace AltTexts: the role of synthetic captions and their interaction with original web-crawled AltTexts in pre-training is still not well understood. Moreover, different multimodal foundation models may have unique preferences for specific caption formats, but efforts to identify the optimal captions for each model remain limited. In this work, we propose a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models. By examining short synthetic captions (SSC) and descriptive synthetic captions (DSC) as case studies, we systematically explore their effects and interactions with AltTexts across models such as CLIP, multimodal LLMs, and diffusion models. Our findings reveal that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone, improving both alignment and performance, with each model demonstrating preferences for particular caption formats. This comprehensive analysis provides valuable insights into optimizing captioning strategies, thereby advancing the pre-training of multimodal foundation models.

Building a generalist model for user interface (UI) understanding is challenging due to various foundational issues, such as platform diversity, resolution variation, and data limitation. In this paper, we introduce Ferret-UI 2, a multimodal large language model (MLLM) designed for universal UI understanding across a wide range of platforms, including iPhone, Android, iPad, Webpage, and AppleTV. Building on the foundation of Ferret-UI, Ferret-UI 2 introduces three key innovations: support for multiple platform types, high-resolution perception through adaptive scaling, and advanced task training data generation powered by GPT-4o with set-of-mark visual prompting. These advancements enable Ferret-UI 2 to perform complex, user-centered interactions, making it highly versatile and adaptable for the expanding diversity of platform ecosystems. Extensive empirical experiments on referring, grounding, user-centric advanced tasks (comprising 9 subtasks $\times$ 5 platforms), GUIDE next-action prediction dataset, and GUI-World multi-platform benchmark demonstrate that Ferret-UI 2 significantly outperforms Ferret-UI, and also shows strong cross-platform transfer capabilities.

This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we automatically generate 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long in Ego4D based on human-annotated data. This is one of the largest egocentric QA datasets. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel ``Memory Pointer Prompting" mechanism. This design includes a global glimpse step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding.

We present MM1.5, a new family of multimodal large language models (MLLMs) designed to enhance capabilities in text-rich image understanding, visual referring and grounding, and multi-image reasoning. Building upon the MM1 architecture, MM1.5 adopts a data-centric approach to model training, systematically exploring the impact of diverse data mixtures across the entire model training lifecycle. This includes high-quality OCR data and synthetic captions for continual pre-training, as well as an optimized visual instruction-tuning data mixture for supervised fine-tuning. Our models range from 1B to 30B parameters, encompassing both dense and mixture-of-experts (MoE) variants, and demonstrate that careful data curation and training strategies can yield strong performance even at small scales (1B and 3B). Additionally, we introduce two specialized variants: MM1.5-Video, designed for video understanding, and MM1.5-UI, tailored for mobile UI understanding. Through extensive empirical studies and ablations, we provide detailed insights into the training processes and decisions that inform our final designs, offering valuable guidance for future research in MLLM development.

CLIP has been a celebrated method for training vision encoders to generate image/text representations facilitating various applications. Recently, it has been widely adopted as the vision backbone of multimodal large language models (MLLMs). The success of CLIP relies on aligning web-crawled noisy text annotations at image levels. However, such criteria may be insufficient for downstream tasks in need of fine-grained vision representations, especially when understanding region-level is demanding for MLLMs. We improve the localization capability of CLIP with several advances. Our proposed pre-training method, Contrastive Localized Language-Image Pre-training (CLOC), complements CLIP with region-text contrastive loss and modules. We formulate a new concept, promptable embeddings, of which the encoder produces image embeddings easy to transform into region representations given spatial hints. To support large-scale pre-training, we design a visually-enriched and spatially-localized captioning framework to effectively generate region-text labels. By scaling up to billions of annotated images, CLOC enables high-quality regional embeddings for recognition and retrieval tasks, and can be a drop-in replacement of CLIP to enhance MLLMs, especially on referring and grounding tasks.

We introduce Ferret, a new Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. To unify referring and grounding in the LLM paradigm, Ferret employs a novel and powerful hybrid region representation that integrates discrete coordinates and continuous features jointly to represent a region in the image. To extract the continuous features of versatile regions, we propose a spatial-aware visual sampler, adept at handling varying sparsity across different shapes. Consequently, Ferret can accept diverse region inputs, such as points, bounding boxes, and free-form shapes. To bolster the desired capability of Ferret, we curate GRIT, a comprehensive refer-and-ground instruction tuning dataset including 1.1M samples that contain rich hierarchical spatial knowledge, with an additional 130K hard negative data to promote model robustness. The resulting model not only achieves superior performance in classical referring and grounding tasks, but also greatly outperforms existing MLLMs in region-based and localization-demanded multimodal chatting. Our evaluations also reveal a significantly improved capability of describing image details and a remarkable alleviation in object hallucination.

Unsupervised Domain Adaptation (UDA) aims to leverage the labeled source domain to solve the tasks on the unlabeled target domain. Traditional UDA methods face the challenge of the tradeoff between domain alignment and semantic class discriminability, especially when a large domain gap exists between the source and target domains. The efforts of applying large-scale pre-training to bridge the domain gaps remain limited. In this work, we propose that Vision-Language Models (VLMs) can empower UDA tasks due to their training pattern with language alignment and their large-scale pre-trained datasets. For example, CLIP and GLIP have shown promising zero-shot generalization in classification and detection tasks. However, directly fine-tuning these VLMs into downstream tasks may be computationally expensive and not scalable if we have multiple domains that need to be adapted. Therefore, in this work, we first study an efficient adaption of VLMs to preserve the original knowledge while maximizing its flexibility for learning new knowledge. Then, we design a domain-aware pseudo-labeling scheme tailored to VLMs for domain disentanglement. We show the superiority of the proposed methods in four UDA-classification and two UDA-detection benchmarks, with a significant improvement (+9.9%) on DomainNet.

This paper presents a grounded language-image pretraining (GLIP) model for learning object-level, language-aware, and semantic-rich visual representations. GLIP unifies object detection and phrase grounding for pre-training. The unification brings two benefits: 1) it allows GLIP to learn from both detection and grounding data to improve both tasks and bootstrap a good grounding model; 2) GLIP can leverage massive image-text pairs by generating grounding boxes in a self-training fashion, making the learned representations semantic-rich. In our experiments, we pre-train GLIP on 27M grounding data, including 3M human-annotated and 24M web-crawled image-text pairs. The learned representations demonstrate strong zero-shot and few-shot transferability to various object-level recognition tasks. 1) When directly evaluated on COCO and LVIS (without seeing any images in COCO during pre-training), GLIP achieves 49.8 AP and 26.9 AP, respectively, surpassing many supervised baselines.11Supervised baselines on COCO object detection: Faster-RCNN w/ ResNet50 (40.2) or ResNet101 (42.0), and DyHead w/ Swin-Tiny (49.7). 2) After fine-tuned on COCO, GLIP achieves 60.8 AP on val and 61.5 AP on test-dev, surpassing prior SoTA. 3) When transferred to 13 downstream object detection tasks, a 1-shot GLIP rivals with a fully-supervised Dynamic Head. Code will be released at https://github.com/microsoft/GLIP.

We present GLIPv2, a grounded VL understanding model, that serves both localization tasks (e.g., object detection, instance segmentation) and Vision-Language (VL) understanding tasks (e.g., VQA, image captioning). GLIPv2 elegantly unifies localization pre-training and Vision-Language Pre-training (VLP) with three pre-training tasks: phrase grounding as a VL reformulation of the detection task, region-word contrastive learning as a novel region-word level contrastive learning task, and the masked language modeling. This unification not only simplifies the previous multi-stage VLP procedure but also achieves mutual benefits between localization and understanding tasks. Experimental results show that a single GLIPv2 model (all model weights are shared) achieves near SoTA performance on various localization and understanding tasks. The model also shows (1) strong zero-shot and few-shot adaption performance on open-vocabulary object detection tasks and (2) superior grounding capability on VL understanding tasks.

The Radar Object Detection 2021 (ROD2021) Challenge, held in the ACM International Conference on Multimedia Retrieval (ICMR) 2021, has been introduced to detect and classify objects purely using an FMCW radar for autonomous driving applications. As a robust sensor to all-weather conditions, radar has rich information hidden in the radio frequencies, which can potentially achieve object detection and classification. This insight will provide a new object perception solution for an autonomous vehicle even in adverse driving scenarios. The ROD2021 Challenge is the first public benchmark focusing on this topic, which attracts great attention and participation. There are more than 260 participants among 37 teams from more than 10 countries with different academic and industrial affiliations, contributing about 300 submissions in the first phase and 400 submissions in the second phase. The final performance is evaluated by average precision (AP). Results add strong value and a better understanding of the radar object detection task for the autonomous vehicle community.

Monocular visual odometry (VO), which is a subset of simultaneous localization and mapping (SLAM) used to determine the position and orientation of a moving object by analyzing the associated monocular camera image sequences, is a critical part in the vision system of autonomous driving. However, based on the frame-by-frame pose estimation, drift error can be incrementally accumulated. Bundle adjustment (BA) is thus introduced to deal with the error-drift problem through correlating several image frames together to optimize camera poses and extracted 3D map points simultaneously. In this paper, we propose a joint BA framework which takes into account additional constraints from the detected road traffic signs. This framework can be effectively integrated into existing VO systems, as evidenced by the improved vehicular localization accuracy in experimental performance when compared with the state-of-the-art baseline VO method.

Multi-object tracking (MOT) is an important topic and critical task related to both static and moving camera applications, such as traffic flow analysis, autonomous driving and robotic vision. However, due to unreliable detection, occlusion and fast camera motion, tracked targets can be easily lost, which makes MOT very challenging. Most recent works exploit spatial and temporal information for MOT, but how to combine appearance and temporal features is still not well addressed. In this paper, we propose an innovative and effective tracking method called TrackletNet Tracker (TNT) that combines temporal and appearance information together as a unified framework. First, we define a graph model which treats each tracklet as a vertex. The tracklets are generated by associating detection results frame by frame with the help of the appearance similarity and the spatial consistency. To compensate camera movement, epipolar constraints are taken into consideration in the association. Then, for every pair of two tracklets, the similarity, called the connectivity in the paper, is measured by our designed multi-scale TrackletNet. Afterwards, the tracklets are clustered into groups and each group represents a unique object ID. Our proposed TNT has the ability to handle most of the challenges in MOT, and achieves promising results on MOT16 and MOT17 benchmark datasets compared with other state-of-the-art methods.

Drones, or general UAVs, equipped with a single camera have been widely deployed to a broad range of applications, such as aerial photography, fast goods delivery and most importantly, surveillance. Despite the great progress achieved in computer vision algorithms, these algorithms are not usually optimized for dealing with images or video sequences acquired by drones, due to various challenges such as occlusion, fast camera motion and pose variation. In this paper, a drone-based multi-object tracking and 3D localization scheme is proposed based on the deep learning based object detection. We first combine a multi-object tracking method called TrackletNet Tracker (TNT) which utilizes temporal and appearance information to track detected objects located on the ground for UAV applications. Then, we are also able to localize the tracked ground objects based on the group plane estimated from the Multi-View Stereo technique. The system deployed on the drone can not only detect and track the objects in a scene, but can also localize their 3D coordinates in meters with respect to the drone camera. The experiments have proved our tracker can reliably handle most of the detected objects captured by drones and achieve favorable 3D localization performance when compared with the state-of-the-art methods.