Zeqing Wang

Zeqing Wang, Danze Chen, Zhaohu Xing et al.

Current game world models simulate environments from a subjective, player-centric perspective. However, by treating the Non-Player Character (NPC) merely as background pixels, these models cannot capture interactions between the player and NPC. In that sense, they act as passive video renderers rather than real simulation engines, lacking the physical understanding needed to model action-induced NPC reactivities. We introduce ReactiveGWM, a reactive game world model that synthesizes dynamic interactions between the player and NPC. Instead of entangling all interaction dynamics, ReactiveGWM explicitly decouples player controls from NPC behaviors. Player actions are injected into the diffusion backbone via a lightweight additive bias, while high-level NPC responses (e.g., Offense, Control, Defense) are grounded through cross-attention modules. Crucially, these modules learn a game-agnostic representation of interactive logic. This enables zero-shot strategy transfer: our learned modules can be plugged directly into off-the-shelf, unannotated world models of different games. This instantly unlocks steerable NPC interactions without any domain-specific retraining. Evaluated on two Street Fighter games, ReactiveGWM maintains fine-grain player controllability while achieving robust, prompt-aligned NPC strategy adherence, paving the way for scalable, strategy-rich interaction with the NPC.

Wentao Wan, Nan Kang, Zeqing Wang et al.

Recently, Visual Programming (VProg) has emerged as a significant framework for visual reasoning (VR) tasks due to its interpretability and cross-task generality. However, even with invoking powerful pre-trained Vision-Language models (VLMs) as visual sub-modules, the performance of VProg on specific VR tasks is markedly inferior compared to well-trained task-specific networks. Although invoking task-specific models can further enhance the performance of VProg on specific VR tasks, it greatly diminishes the cross-task generalization ability of VProg. Besides, the non-differentiable nature of VProg prevents direct fine-tuning on specific VR tasks for further performance improvement. Attempt to address these issues, we propose SDVP, a Stepwise Distillation learning strategy for non-differentiable VPorg across various VR tasks. Specifically, our SDVP stepwise distills the capabilities of existing, well-trained small task-specific models for decomposed visual sub-tasks in VProg into the much larger VLMs invoked by corresponding visual sub-modules. Besides, distilling the knowledge of little-size task-specific models into pre-trained larger VLMs rather than replacing them helps keep the cross-task abilities of VProgs. Extensive and comprehensive experimental results on different VProg frameworks demonstrate that our SDVP obtains significant performance gains on specific VR benchmarks, i.e., GQA (+2.4\%) and NLVRv2 (+6.2\%) for VisProg and GQA (+6.5\%) and NLVRv2 (+4.0\%) for ViperGPT, and also maintains a promising performance for VProg on unseen and previous VR tasks.

Lirong Gao, Zeqing Wang, Yuyan Cai et al.

While Large Language Models (LLMs) have increasingly assisted in historical tasks such as text processing, their capacity for professional-level historical reasoning remains underexplored. Existing benchmarks primarily assess basic knowledge breadth or lexical understanding, failing to capture the higher-order skills, such as evidentiary reasoning,that are central to historical research. To fill this gap, we introduce ProHist-Bench, a novel benchmark anchored in the Chinese Imperial Examination (Keju) system, a comprehensive microcosm of East Asian political, social, and intellectual history spanning over 1,300 years. Developed through deep interdisciplinary collaboration, ProHist-Bench features 400 challenging, expert-curated questions across eight dynasties, accompanied by 10,891 fine-grained evaluation rubrics. Through a rigorous evaluation of 18 LLMs, we reveal a significant proficiency gap: even state-of-the-art LLMs struggle with complex historical research questions. We hope ProHist-Bench will facilitate the development of domain-specific reasoning LLMs, advance computational historical research, and further uncover the untapped potential of LLMs. We release ProHist-Bench at https://github.com/inclusionAI/ABench/tree/main/ProHist-Bench.

Zeqing Wang, Wentao Wan, Qiqing Lao et al.

Recently, to comprehensively improve Vision Language Models (VLMs) for Visual Question Answering (VQA), several methods have been proposed to further reinforce the inference capabilities of VLMs to independently tackle VQA tasks rather than some methods that only utilize VLMs as aids to Large Language Models (LLMs). However, these methods ignore the rich common-sense knowledge inside the given VQA image sampled from the real world. Thus, they cannot fully use the powerful VLM for the given VQA question to achieve optimal performance. Attempt to overcome this limitation and inspired by the human top-down reasoning process, i.e., systematically exploring relevant issues to derive a comprehensive answer, this work introduces a novel, explainable multi-agent collaboration framework by leveraging the expansive knowledge of Large Language Models (LLMs) to enhance the capabilities of VLMs themselves. Specifically, our framework comprises three agents, i.e., Responder, Seeker, and Integrator, to collaboratively answer the given VQA question by seeking its relevant issues and generating the final answer in such a top-down reasoning process. The VLM-based Responder agent generates the answer candidates for the question and responds to other relevant issues. The Seeker agent, primarily based on LLM, identifies relevant issues related to the question to inform the Responder agent and constructs a Multi-View Knowledge Base (MVKB) for the given visual scene by leveraging the build-in world knowledge of LLM. The Integrator agent combines knowledge from the Seeker agent and the Responder agent to produce the final VQA answer. Extensive and comprehensive evaluations on diverse VQA datasets with a variety of VLMs demonstrate the superior performance and interpretability of our framework over the baseline method in the zero-shot setting without extra training cost.

Zeqing Wang, Shiyuan Zhang, Chengpei Tang et al.

Reasoning about temporal causality, particularly irreversible transformations of objects governed by real-world knowledge (e.g., fruit decay and human aging), is a fundamental aspect of human visual understanding. Unlike temporal perception based on simple event sequences, this form of reasoning requires a deeper comprehension of how object states change over time. Although the current powerful Vision-Language Models (VLMs) have demonstrated impressive performance on a wide range of downstream tasks, their capacity to reason about temporal causality remains underexplored. To address this gap, we introduce \textbf{TimeCausality}, a novel benchmark specifically designed to evaluate the causal reasoning ability of VLMs in the temporal dimension. Based on our TimeCausality, we find that while the current SOTA open-source VLMs have achieved performance levels comparable to closed-source models like GPT-4o on various standard visual question answering tasks, they fall significantly behind on our benchmark compared with their closed-source competitors. Furthermore, even GPT-4o exhibits a marked drop in performance on TimeCausality compared to its results on other tasks. These findings underscore the critical need to incorporate temporal causality into the evaluation and development of VLMs, and they highlight an important challenge for the open-source VLM community moving forward. Code and Data are available at \href{https://github.com/Zeqing-Wang/TimeCausality }{TimeCausality}.

Zeqing Wang, Gongfan Fang, Xinyin Ma et al.

While diffusion language models (DLMs) offer a promising alternative to autoregressive models (ARs), existing open-source DLMs suffer from high inference latency. This bottleneck is mainly due to the attention's quadratic complexity with respect to context length in computing all query-key pairs. Intuitively, to reduce this complexity, a natural strategy is to restrict attention to sparse patterns that retain only the most relevant connections. Such approaches are well-established in ARs, where attention follows fixed and clearly defined sparse patterns. However, in DLMs, we observe distinct sparsity behaviors: (1) attention patterns vary across heads, (2) attention patterns in each head remain highly similar across denoising steps, and (3) early denoising steps are critical for generation. These findings render sparse attention methods designed for ARs largely incompatible with DLMs, as they fail to capture head-specific structures and risk degrading generation when applied in early denoising steps. To address these challenges, we propose SparseD, a novel sparse attention method for DLMs. Leveraging the observations, SparseD only requires pre-computing head-specific sparse patterns one time, and reuses them across all steps. This prevents recomputing sparse patterns at each denoising step. Meanwhile, SparseD uses full attention in the early steps, then switches to sparse attention later to maintain generation quality. Together, these establish SparseD as a practical and efficient solution for deploying DLMs in long-context applications. Experimental results demonstrate that SparseD achieves lossless acceleration, delivering up to $1.50\times$ speedup over FlashAttention at a 64k context length with 1,024 denoising steps.

Hui Fu, Zeqing Wang, Ke Gong et al.

Speech-driven 3D facial animation aims to synthesize vivid facial animations that accurately synchronize with speech and match the unique speaking style. However, existing works primarily focus on achieving precise lip synchronization while neglecting to model the subject-specific speaking style, often resulting in unrealistic facial animations. To the best of our knowledge, this work makes the first attempt to explore the coupled information between the speaking style and the semantic content in facial motions. Specifically, we introduce an innovative speaking style disentanglement method, which enables arbitrary-subject speaking style encoding and leads to a more realistic synthesis of speech-driven facial animations. Subsequently, we propose a novel framework called \textbf{Mimic} to learn disentangled representations of the speaking style and content from facial motions by building two latent spaces for style and content, respectively. Moreover, to facilitate disentangled representation learning, we introduce four well-designed constraints: an auxiliary style classifier, an auxiliary inverse classifier, a content contrastive loss, and a pair of latent cycle losses, which can effectively contribute to the construction of the identity-related style space and semantic-related content space. Extensive qualitative and quantitative experiments conducted on three publicly available datasets demonstrate that our approach outperforms state-of-the-art methods and is capable of capturing diverse speaking styles for speech-driven 3D facial animation. The source code and supplementary video are publicly available at: https://zeqing-wang.github.io/Mimic/

Zeqing Wang, Fei Cheng, Kangye Ji et al.

Continual learning (CL) is a technique that enables neural networks to constantly adapt to their dynamic surroundings. Despite being overlooked for a long time, this technology can considerably address the customized needs of users in edge devices. Actually, most CL methods require huge resource consumption by the training behavior to acquire generalizability among all tasks for delaying forgetting regardless of edge scenarios. Therefore, this paper proposes a compact algorithm called LightCL, which evaluates and compresses the redundancy of already generalized components in structures of the neural network. Specifically, we consider two factors of generalizability, learning plasticity and memory stability, and design metrics of both to quantitatively assess generalizability of neural networks during CL. This evaluation shows that generalizability of different layers in a neural network exhibits a significant variation. Thus, we $\textit{Maintain Generalizability}$ by freezing generalized parts without the resource-intensive training process and $\textit{Memorize Feature Patterns}$ by stabilizing feature extracting of previous tasks to enhance generalizability for less-generalized parts with a little extra memory, which is far less than the reduction by freezing. Experiments illustrate that LightCL outperforms other state-of-the-art methods and reduces at most $\textbf{6.16$\times$}$ memory footprint. We also verify the effectiveness of LightCL on the edge device.

Zeqing Wang, Kangye Ji, Di Wang et al.

Segment Anything Model (SAM) struggles with segmenting objects in the open world, especially across diverse and dynamic domains. Continual segmentation (CS) is a potential technique to solve this issue, but a significant obstacle is the intractable balance between previous domains (stability) and new domains (plasticity) during CS. Furthermore, how to utilize two kinds of features of SAM, images and prompts, in an efficient and effective CS manner remains a significant hurdle. In this work, we propose a novel CS method, termed SAMCL, to address these challenges. It is the first study to empower SAM with the CS ability across dynamic domains. SAMCL decouples stability and plasticity during CS by two components: $\textit{AugModule}$ and $\textit{Module Selector}$. Specifically, SAMCL leverages individual $\textit{AugModule}$ to effectively and efficiently learn new relationships between images and prompts in each domain. $\textit{Module Selector}$ selects the appropriate module during testing, based on the inherent ability of SAM to distinguish between different domains. These two components enable SAMCL to realize a task-agnostic method without any interference across different domains. Experimental results demonstrate that SAMCL outperforms state-of-the-art methods, achieving an exceptionally low average forgetting of just $0.5$%, along with at least a $2.5$% improvement in transferring to unseen domains. Moreover, the tunable parameter consumption in AugModule is about $0.236$MB, marking at least a $23.3$% reduction compared to other fine-tuning methods.

Zhenxiong Tan, Zeqing Wang, Xingyi Yang et al.

We introduce Vision Bridge Transformer (ViBT), a large-scale instantiation of Brownian Bridge Models designed for conditional generation. Unlike traditional diffusion models that transform noise into data, Bridge Models directly model the trajectory between inputs and outputs, creating an efficient data-to-data translation paradigm. By scaling these models to 20B and 1.3B parameters, we demonstrate their effectiveness for image and video translation tasks. To support this scale, we adopt a Transformer architecture and propose a variance-stabilized velocity-matching objective for robust training. Together, these advances highlight the power of scaling Bridge Models for instruction-based image editing and complex video translation.