Juncheng Li

Qifan Yu, Juncheng Li, Yu Wu et al.

Scene Graph Generation (SGG) aims to extract <subject, predicate, object> relationships in images for vision understanding. Although recent works have made steady progress on SGG, they still suffer long-tail distribution issues that tail-predicates are more costly to train and hard to distinguish due to a small amount of annotated data compared to frequent predicates. Existing re-balancing strategies try to handle it via prior rules but are still confined to pre-defined conditions, which are not scalable for various models and datasets. In this paper, we propose a Cross-modal prediCate boosting (CaCao) framework, where a visually-prompted language model is learned to generate diverse fine-grained predicates in a low-resource way. The proposed CaCao can be applied in a plug-and-play fashion and automatically strengthen existing SGG to tackle the long-tailed problem. Based on that, we further introduce a novel Entangled cross-modal prompt approach for open-world predicate scene graph generation (Epic), where models can generalize to unseen predicates in a zero-shot manner. Comprehensive experiments on three benchmark datasets show that CaCao consistently boosts the performance of multiple scene graph generation models in a model-agnostic way. Moreover, our Epic achieves competitive performance on open-world predicate prediction. The data and code for this paper are publicly available.

Kaihang Pan, Zhaoyu Fan, Juncheng Li et al.

The swift advancement in Multimodal LLMs (MLLMs) also presents significant challenges for effective knowledge editing. Current methods, including intrinsic knowledge editing and external knowledge resorting, each possess strengths and weaknesses, struggling to balance the desired properties of reliability, generality, and locality when applied to MLLMs. In this paper, we propose UniKE, a novel multimodal editing method that establishes a unified perspective and paradigm for intrinsic knowledge editing and external knowledge resorting. Both types of knowledge are conceptualized as vectorized key-value memories, with the corresponding editing processes resembling the assimilation and accommodation phases of human cognition, conducted at the same semantic levels. Within such a unified framework, we further promote knowledge collaboration by disentangling the knowledge representations into the semantic and truthfulness spaces. Extensive experiments validate the effectiveness of our method, which ensures that the post-edit MLLM simultaneously maintains excellent reliability, generality, and locality. The code for UniKE is available at \url{https://github.com/beepkh/UniKE}.

Hongzhe Huang, Jiang Liu, Zhewen Yu et al.

Recent advances in Multi-modal Large Language Models (MLLMs), such as LLaVA-series models, are driven by massive machine-generated instruction-following data tuning. Such automatic instruction collection pipelines, however, inadvertently introduce significant variability in data quality. This paper introduces a novel instruction curation algorithm, derived from two unique perspectives, human and LLM preference alignment, to compress this vast corpus of machine-generated multimodal instructions to a compact and high-quality form: (i) For human preference alignment, we have collected a machine-generated multimodal instruction dataset and established a comprehensive set of both subjective and objective criteria to guide the data quality assessment critically from human experts. By doing so, a reward model was trained on the annotated dataset to internalize the nuanced human understanding of instruction alignment. (ii) For LLM preference alignment, given the instruction selected by the reward model, we propose leveraging the inner LLM used in MLLM to align the writing style of visual instructions with that of the inner LLM itself, resulting in LLM-aligned instruction improvement. Extensive experiments demonstrate that we can maintain or even improve model performance by compressing synthetic multimodal instructions by up to 90%. Impressively, by aggressively reducing the training instructions from 158k to 14k (9$\times$ smaller), our model consistently outperforms its full-size dataset counterpart across various MLLM benchmarks. Our project is available at https://github.com/DCDmllm/Align2LLaVA.

Wenqiao Zhang, Tianwei Lin, Jiang Liu et al.

Recent advancements indicate that scaling up Multimodal Large Language Models (MLLMs) effectively enhances performance on downstream multimodal tasks. The prevailing MLLM paradigm, \emph{e.g.}, LLaVA, transforms visual features into text-like tokens using a \emph{static} vision-language mapper, thereby enabling \emph{static} LLMs to develop the capability to comprehend visual information through visual instruction tuning. Although promising, the \emph{static} tuning strategy~\footnote{The static tuning refers to the trained model with static parameters.} that shares the same parameters may constrain performance across different downstream multimodal tasks. In light of this, we introduce HyperLLaVA, which involves adaptive tuning of the projector and LLM parameters, in conjunction with a dynamic visual expert and language expert, respectively. These experts are derived from HyperNetworks, which generates adaptive parameter shifts through visual and language guidance, enabling dynamic projector and LLM modeling in two-stage training. Our experiments demonstrate that our solution significantly surpasses LLaVA on existing MLLM benchmarks, including MME, MMBench, SEED-Bench, and LLaVA-Bench. ~\footnote{Our project is available on the link https://github.com/DCDmllm/HyperLLaVA}.

Dang Jisheng, Wu Xudong, Wang Bimei et al.

Existing video segmenter and grounder approaches, exemplified by Sa2VA, directly fuse features within segmentation models. This often results in an undesirable entanglement of dynamic visual information and static semantics, thereby degrading segmentation accuracy. To systematically mitigate this issue, we propose DeSa2VA, a decoupling-enhanced prompting scheme integrating text pre-training and a linear decoupling module to address the information processing limitations inherent in SAM-2. Specifically, first, we devise a pre-training paradigm that converts textual ground-truth labels into point-level prompts while generating corresponding text masks. These masks are refined through a hybrid loss function to strengthen the model's semantic grounding capabilities. Next, we employ linear projection to disentangle hidden states that generated by a large language model into distinct textual and visual feature subspaces. Finally, a dynamic mask fusion strategy synergistically combines these decoupled features through triple supervision from predicted text/visual masks and ground-truth annotations. Extensive experiments demonstrate state-of-the-art performance across diverse tasks, including image segmentation, image question answering, video segmentation, and video question answering. Our codes are available at https://github.com/longmalongma/DeSa2VA.

Qifan Yu, Juncheng Li, Wentao Ye et al.

Recent text-to-image generation models have shown promising results in generating high-fidelity photo-realistic images. In parallel, the problem of data scarcity has brought a growing interest in employing AIGC technology for high-quality data expansion. However, this paradigm requires well-designed prompt engineering that cost-less data expansion and labeling remain under-explored. Inspired by LLM's powerful capability in task guidance, we propose a new paradigm of annotated data expansion named as ChatGenImage. The core idea behind it is to leverage the complementary strengths of diverse models to establish a highly effective and user-friendly pipeline for interactive data augmentation. In this work, we extensively study how LLMs communicate with AIGC model to achieve more controllable image generation and make the first attempt to collaborate them for automatic data augmentation for a variety of downstream tasks. Finally, we present fascinating results obtained from our ChatGenImage framework and demonstrate the powerful potential of our synthetic data for systematic vision adaptation. Our codes are available at https://github.com/Yuqifan1117/Labal-Anything-Pipeline.

Wei Chow, Juncheng Li, Qifan Yu et al.

In recent times, Vision-Language Models (VLMs) have been trained under two predominant paradigms. Generative training has enabled Multimodal Large Language Models (MLLMs) to tackle various complex tasks, yet issues such as hallucinations and weak object discrimination persist. Discriminative training, exemplified by models like CLIP, excels in zero-shot image-text classification and retrieval, yet struggles with complex scenarios requiring fine-grained semantic differentiation. This paper addresses these challenges by proposing a unified approach that integrates the strengths of both paradigms. Considering interleaved image-text sequences as the general format of input samples, we introduce a structure-induced training strategy that imposes semantic relationships between input samples and the MLLM's hidden state. This approach enhances the MLLM's ability to capture global semantics and distinguish fine-grained semantics. By leveraging dynamic sequence alignment within the Dynamic Time Warping framework and integrating a novel kernel for fine-grained semantic differentiation, our method effectively balances generative and discriminative tasks. Extensive experiments demonstrate the effectiveness of our approach, achieving state-of-the-art results in multiple generative tasks, especially those requiring cognitive and discrimination abilities. Additionally, our method surpasses discriminative benchmarks in interleaved and fine-grained retrieval tasks. By employing a retrieval-augmented generation strategy, our approach further enhances performance in some generative tasks within one model, offering a promising direction for future research in vision-language modeling.

Leigang Qu, Haochuan Li, Wenjie Wang et al.

Large Multimodal Models (LMMs) have demonstrated impressive capabilities in multimodal understanding and generation, pushing forward advancements in text-to-image generation. However, achieving accurate text-image alignment for LMMs, particularly in compositional scenarios, remains challenging. Existing approaches, such as layout planning for multi-step generation and learning from human feedback or AI feedback, depend heavily on prompt engineering, costly human annotations, and continual upgrading, limiting flexibility and scalability. In this work, we introduce a model-agnostic iterative self-improvement framework (SILMM) that can enable LMMs to provide helpful and scalable self-feedback and optimize text-image alignment via Direct Preference Optimization (DPO). DPO can readily applied to LMMs that use discrete visual tokens as intermediate image representations; while it is less suitable for LMMs with continuous visual features, as obtaining generation probabilities is challenging. To adapt SILMM to LMMs with continuous features, we propose a diversity mechanism to obtain diverse representations and a kernel-based continuous DPO for alignment. Extensive experiments on three compositional text-to-image generation benchmarks validate the effectiveness and superiority of SILMM, showing improvements exceeding 30% on T2I-CompBench++ and around 20% on DPG-Bench.

Wei Chow, Yuan Gao, Linfeng Li et al.

Semantic retrieval is crucial for modern applications yet remains underexplored in current research. Existing datasets are limited to single languages, single images, or singular retrieval conditions, often failing to fully exploit the expressive capacity of visual information as evidenced by maintained performance when images are replaced with captions. However, practical retrieval scenarios frequently involve interleaved multi-condition queries with multiple images. Hence, this paper introduces MERIT, the first multilingual dataset for interleaved multi-condition semantic retrieval, comprising 320,000 queries with 135,000 products in 5 languages, covering 7 distinct product categories. Extensive experiments on MERIT identify existing models's limitation: focusing solely on global semantic information while neglecting specific conditional elements in queries. Consequently, we propose Coral, a novel fine-tuning framework that adapts pre-trained MLLMs by integrating embedding reconstruction to preserve fine-grained conditional elements and contrastive learning to extract comprehensive global semantics. Experiments demonstrate that Coral achieves a 45.9% performance improvement over conventional approaches on MERIT, with strong generalization capabilities validated across 8 established retrieval benchmarks. Collectively, our contributions - a novel dataset, identification of critical limitations in existing approaches, and an innovative fine-tuning framework - establish a foundation for future research in interleaved multi-condition semantic retrieval.

Juncheng Li, David J. Cappelleri

This paper presents Sim-Suction, a robust object-aware suction grasp policy for mobile manipulation platforms with dynamic camera viewpoints, designed to pick up unknown objects from cluttered environments. Suction grasp policies typically employ data-driven approaches, necessitating large-scale, accurately-annotated suction grasp datasets. However, the generation of suction grasp datasets in cluttered environments remains underexplored, leaving uncertainties about the relationship between the object of interest and its surroundings. To address this, we propose a benchmark synthetic dataset, Sim-Suction-Dataset, comprising 500 cluttered environments with 3.2 million annotated suction grasp poses. The efficient Sim-Suction-Dataset generation process provides novel insights by combining analytical models with dynamic physical simulations to create fast and accurate suction grasp pose annotations. We introduce Sim-Suction-Pointnet to generate robust 6D suction grasp poses by learning point-wise affordances from the Sim-Suction-Dataset, leveraging the synergy of zero-shot text-to-segmentation. Real-world experiments for picking up all objects demonstrate that Sim-Suction-Pointnet achieves success rates of 96.76%, 94.23%, and 92.39% on cluttered level 1 objects (prismatic shape), cluttered level 2 objects (more complex geometry), and cluttered mixed objects, respectively. The Sim-Suction policies outperform state-of-the-art benchmarks tested by approximately 21% in cluttered mixed scenes.

Juncheng Li, David J. Cappelleri

In this paper, we present Sim-MEES: a large-scale synthetic dataset that contains 1,550 objects with varying difficulty levels and physics properties, as well as 11 million grasp labels for mobile manipulators to plan grasps using different gripper modalities in cluttered environments. Our dataset generation process combines analytic models and dynamic simulations of the entire cluttered environment to provide accurate grasp labels. We provide a detailed study of our proposed labeling process for both parallel jaw grippers and suction cup grippers, comparing them with state-of-the-art methods to demonstrate how Sim-MEES can provide precise grasp labels in cluttered environments.

Juncheng Li, Maopeng Ran, Lihua Xie

Autonomous ground vehicles (AGVs) are receiving increasing attention, and the motion planning and control problem for these vehicles has become a hot research topic. In real applications such as material handling, an AGV is subject to large uncertainties and its motion planning and control become challenging. In this paper, we investigate this problem by proposing a hierarchical control scheme, which is integrated by a model predictive control (MPC) based path planning and trajectory tracking control at the high level, and a reduced-order extended state observer (RESO) based dynamic control at the low level. The control at the high level consists of an MPC-based improved path planner, a velocity planner, and an MPC-based tracking controller. Both the path planning and trajectory tracking control problems are formulated under an MPC framework. The control at the low level employs the idea of active disturbance rejection control (ADRC). The uncertainties are estimated via a RESO and then compensated in the control in real-time. We show that, for the first-order uncertain AGV dynamic model, the RESO-based control only needs to know the control direction. Finally, simulations and experiments on an AGV with different payloads are conducted. The results illustrate that the proposed hierarchical control scheme achieves satisfactory motion planning and control performance with large uncertainties.

Han Wang, Juncheng Li, Maopeng Ran et al.

Loop closure detection plays an important role in reducing localization drift in Simultaneous Localization And Mapping (SLAM). It aims to find repetitive scenes from historical data to reset localization. To tackle the loop closure problem, existing methods often leverage on the matching of visual features, which achieve good accuracy but require high computational resources. However, feature point based methods ignore the patterns of image, i.e., the shape of the objects as well as the distribution of objects in an image. It is believed that this information is usually unique for a scene and can be utilized to improve the performance of traditional loop closure detection methods. In this paper we leverage and compress the information into a binary image to accelerate an existing fast loop closure detection method via binary content. The proposed method can greatly reduce the computational cost without sacrificing recall rate. It consists of three parts: binary content construction, fast image retrieval and precise loop closure detection. No offline training is required. Our method is compared with the state-of-the-art loop closure detection methods and the results show that it outperforms the traditional methods at both recall rate and speed.

Juncheng Li, Xin Wang, Siliang Tang et al.

Visual navigation is a task of training an embodied agent by intelligently navigating to a target object (e.g., television) using only visual observations. A key challenge for current deep reinforcement learning models lies in the requirements for a large amount of training data. It is exceedingly expensive to construct sufficient 3D synthetic environments annotated with the target object information. In this paper, we focus on visual navigation in the low-resource setting, where we have only a few training environments annotated with object information. We propose a novel unsupervised reinforcement learning approach to learn transferable meta-skills (e.g., bypass obstacles, go straight) from unannotated environments without any supervisory signals. The agent can then fast adapt to visual navigation through learning a high-level master policy to combine these meta-skills, when the visual-navigation-specified reward is provided. Evaluation in the AI2-THOR environments shows that our method significantly outperforms the baseline by 53.34% relatively on SPL, and further qualitative analysis demonstrates that our method learns transferable motor primitives for visual navigation.