Yuhang He

Zihan Li, Jialan Zheng, Ziyu Li et al.

Understanding the human brain requires access to its microscopic tissue architecture. Diffusion magnetic resonance imaging (MRI) provides the only noninvasive window into whole-brain microstructure in vivo, yet reliable quantitative mapping remains confined to specialized research settings requiring dense sampling and optimized acquisition protocols. To address this gap, we present a physics-informed generative microstructure network (PIGMENT) that learns a universal generative prior of human brain microstructure and adapts it zero-shot to each participant's measured data to recover subject-specific maps. Trained on 11375 scans spanning multiple sites, vendors, and field strengths, PIGMENT enabled reliable quantitative mapping for tensor, kurtosis, and NODDI models across external datasets from five independent centers. It remains effective where conventional fitting becomes unreliable, recovering meaningful maps from extremely sparse acquisitions while supporting downstream tractography and structural connectivity mapping. PIGMENT estimates demonstrated strong biological validity, preserving submillimeter cortical microarchitectural patterns and early-childhood white matter developmental trajectories from 10-fold accelerated scans. Furthermore, PIGMENT enables reliable quantitative tensor mapping on cost-efficient low-field systems and the extraction of tumor-related biomarkers using ultra-fast clinical protocols. Together, these results establish PIGMENT as a physics-informed foundation model that extends quantitative diffusion MRI into regimes traditionally too sparse, heterogeneous, or clinically constrained for reliable analysis.

Cong Wan, Yuhang He, Xiang Song et al.

Diffusion models have revolutionized customized text-to-image generation, allowing for efficient synthesis of photos from personal data with textual descriptions. However, these advancements bring forth risks including privacy breaches and unauthorized replication of artworks. Previous researches primarily center around using prompt-specific methods to generate adversarial examples to protect personal images, yet the effectiveness of existing methods is hindered by constrained adaptability to different prompts. In this paper, we introduce a Prompt-Agnostic Adversarial Perturbation (PAP) method for customized diffusion models. PAP first models the prompt distribution using a Laplace Approximation, and then produces prompt-agnostic perturbations by maximizing a disturbance expectation based on the modeled distribution. This approach effectively tackles the prompt-agnostic attacks, leading to improved defense stability. Extensive experiments in face privacy and artistic style protection, demonstrate the superior generalization of PAP in comparison to existing techniques. Our project page is available at https://github.com/vancyland/Prompt-Agnostic-Adversarial-Perturbation-for-Customized-Diffusion-Models.github.io.

Jizhou Han, Chenhao Ding, Yuhang He et al.

Generalized Category Discovery (GCD) seeks to uncover novel categories in unlabeled data while preserving recognition of known categories, yet prevailing visual-only pipelines and the loose coupling between supervised learning and discovery often yield brittle boundaries on fine-grained, look-alike categories. We introduce the Analogical Textual Concept Generator (ATCG), a plug-and-play module that analogizes from labeled knowledge to new observations, forming textual concepts for unlabeled samples. Fusing these analogical textual concepts with visual features turns discovery into a visual-textual reasoning process, transferring prior knowledge to novel data and sharpening category separation. ATCG attaches to both parametric and clustering style GCD pipelines and requires no changes to their overall design. Across six benchmarks, ATCG consistently improves overall, known-class, and novel-class performance, with the largest gains on fine-grained data. Our code is available at: https://github.com/zhou-9527/AnaLogical-GCD.

Yuhang He, Lin Chen, Junkun Xie et al.

We propose a novel framework to learn 3D point cloud semantics from 2D multi-view image observations containing pose error. On the one hand, directly learning from the massive, unstructured and unordered 3D point cloud is computationally and algorithmically more difficult than learning from compactly-organized and context-rich 2D RGB images. On the other hand, both LiDAR point cloud and RGB images are captured in standard automated-driving datasets. This motivates us to conduct a "task transfer" paradigm so that 3D semantic segmentation benefits from aggregating 2D semantic cues, albeit pose noises are contained in 2D image observations. Among all difficulties, pose noise and erroneous prediction from 2D semantic segmentation approaches are the main challenges for the task transfer. To alleviate the influence of those factor, we perceive each 3D point using multi-view images and for each single image a patch observation is associated. Moreover, the semantic labels of a block of neighboring 3D points are predicted simultaneously, enabling us to exploit the point structure prior to further improve the performance. A hierarchical full attention network~(HiFANet) is designed to sequentially aggregates patch, bag-of-frames and inter-point semantic cues, with hierarchical attention mechanism tailored for different level of semantic cues. Also, each preceding attention block largely reduces the feature size before feeding to the next attention block, making our framework slim. Experiment results on Semantic-KITTI show that the proposed framework outperforms existing 3D point cloud based methods significantly, it requires much less training data and exhibits tolerance to pose noise. The code is available at https://github.com/yuhanghe01/HiFANet.

Jia-Xing Zhong, Ta-Ying Cheng, Yuhang He et al.

A truly generalizable approach to rigid segmentation and motion estimation is fundamental to 3D understanding of articulated objects and moving scenes. In view of the closely intertwined relationship between segmentation and motion estimates, we present an SE(3) equivariant architecture and a training strategy to tackle this task in an unsupervised manner. Our architecture is composed of two interconnected, lightweight heads. These heads predict segmentation masks using point-level invariant features and estimate motion from SE(3) equivariant features, all without the need for category information. Our training strategy is unified and can be implemented online, which jointly optimizes the predicted segmentation and motion by leveraging the interrelationships among scene flow, segmentation mask, and rigid transformations. We conduct experiments on four datasets to demonstrate the superiority of our method. The results show that our method excels in both model performance and computational efficiency, with only 0.25M parameters and 0.92G FLOPs. To the best of our knowledge, this is the first work designed for category-agnostic part-level SE(3) equivariance in dynamic point clouds.

Xinyuan Gao, Songlin Dong, Yuhang He et al.

The problem of Rehearsal-Free Continual Learning (RFCL) aims to continually learn new knowledge while preventing forgetting of the old knowledge, without storing any old samples and prototypes. The latest methods leverage large-scale pre-trained models as the backbone and use key-query matching to generate trainable prompts to learn new knowledge. However, the domain gap between the pre-training dataset and the downstream datasets can easily lead to inaccuracies in key-query matching prompt selection when directly generating queries using the pre-trained model, which hampers learning new knowledge. Thus, in this paper, we propose a beyond prompt learning approach to the RFCL task, called Continual Adapter (C-ADA). It mainly comprises a parameter-extensible continual adapter layer (CAL) and a scaling and shifting (S&S) module in parallel with the pre-trained model. C-ADA flexibly extends specific weights in CAL to learn new knowledge for each task and freezes old weights to preserve prior knowledge, thereby avoiding matching errors and operational inefficiencies introduced by key-query matching. To reduce the gap, C-ADA employs an S&S module to transfer the feature space from pre-trained datasets to downstream datasets. Moreover, we propose an orthogonal loss to mitigate the interaction between old and new knowledge. Our approach achieves significantly improved performance and training speed, outperforming the current state-of-the-art (SOTA) method. Additionally, we conduct experiments on domain-incremental learning, surpassing the SOTA, and demonstrating the generality of our approach in different settings.

Songlin Dong, Haoyu Luo, Yuhang He et al.

Current class-incremental learning research mainly focuses on single-label classification tasks while multi-label class-incremental learning (MLCIL) with more practical application scenarios is rarely studied. Although there have been many anti-forgetting methods to solve the problem of catastrophic forgetting in class-incremental learning, these methods have difficulty in solving the MLCIL problem due to label absence and information dilution. In this paper, we propose a knowledge restore and transfer (KRT) framework for MLCIL, which includes a dynamic pseudo-label (DPL) module to restore the old class knowledge and an incremental cross-attention(ICA) module to save session-specific knowledge and transfer old class knowledge to the new model sufficiently. Besides, we propose a token loss to jointly optimize the incremental cross-attention module. Experimental results on MS-COCO and PASCAL VOC datasets demonstrate the effectiveness of our method for improving recognition performance and mitigating forgetting on multi-label class-incremental learning tasks.

Zeyang Zhao, Qilong Xue, Yuhang He et al.

This paper introduces the point-axis representation for oriented object detection, emphasizing its flexibility and geometrically intuitive nature with two key components: points and axes. 1) Points delineate the spatial extent and contours of objects, providing detailed shape descriptions. 2) Axes define the primary directionalities of objects, providing essential orientation cues crucial for precise detection. The point-axis representation decouples location and rotation, addressing the loss discontinuity issues commonly encountered in traditional bounding box-based approaches. For effective optimization without introducing additional annotations, we propose the max-projection loss to supervise point set learning and the cross-axis loss for robust axis representation learning. Further, leveraging this representation, we present the Oriented DETR model, seamlessly integrating the DETR framework for precise point-axis prediction and end-to-end detection. Experimental results demonstrate significant performance improvements in oriented object detection tasks.

Jing Wu, Daphne Barretto, Yiye Chen et al.

Long-horizon, repetitive workflows are common in professional settings, such as processing expense reports from receipts and entering student grades from exam papers. These tasks are often tedious for humans since they can extend to extreme lengths proportional to the size of the data to process. However, they are ideal for Computer-Use Agents (CUAs) due to their structured, recurring sub-workflows with logic that can be systematically learned. Identifying the absence of an evaluation benchmark as a primary bottleneck, we establish OS-Marathon, comprising 242 long-horizon, repetitive tasks across 2 domains to evaluate state-of-the-art (SOTA) agents. We then introduce a cost-effective method to construct a condensed demonstration using only few-shot examples to teach agents the underlying workflow logic, enabling them to execute similar workflows effectively on larger, unseen data collections. Extensive experiments demonstrate both the inherent challenges of these tasks and the effectiveness of our proposed method. Project website: https://os-marathon.github.io/.

Jiashuo Li, Shaokun Wang, Bo Qian et al.

Non-exemplar class Incremental Learning (NECIL) enables models to continuously acquire new classes without retraining from scratch and storing old task exemplars, addressing privacy and storage issues. However, the absence of data from earlier tasks exacerbates the challenge of catastrophic forgetting in NECIL. In this paper, we propose a novel framework called Dynamic Integration of task-specific Adapters (DIA), which comprises two key components: Task-Specific Adapter Integration (TSAI) and Patch-Level Model Alignment. TSAI boosts compositionality through a patch-level adapter integration strategy, which provides a more flexible compositional solution while maintaining low computation costs. Patch-Level Model Alignment maintains feature consistency and accurate decision boundaries via two specialized mechanisms: Patch-Level Distillation Loss (PDL) and Patch-Level Feature Reconstruction method (PFR). Specifically, the PDL preserves feature-level consistency between successive models by implementing a distillation loss based on the contributions of patch tokens to new class learning. The PFR facilitates accurate classifier alignment by reconstructing old class features from previous tasks that adapt to new task knowledge. Extensive experiments validate the effectiveness of our DIA, revealing significant improvements on benchmark datasets in the NECIL setting, maintaining an optimal balance between computational complexity and accuracy.

Yuhang He, Haodong Wu, Siyi Liu et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has substantially improved the reasoning ability of Large Language Models (LLMs). However, its sparse outcome-based rewards pose a fundamental credit assignment problem. We analyze this problem through the joint lens of reward polarity and token entropy. Our diagnostic tool, the Four Quadrant Decomposition, isolates token updates by polarity and entropy, and controlled ablations show that reasoning improvements concentrate in the high-entropy quadrants. To justify this observation theoretically, we adapt Conditional Mutual Information to the autoregressive RLVR setting and prove that the credit a token can carry is upper-bounded by its entropy. This view yields testable predictions that reasoning gains arise primarily from high-entropy tokens, with unique roles for positive and negative updates. A gradient analysis of GRPO further reveals how uniform reward broadcast dilutes signal at high-entropy positions while over-crediting deterministic tokens. Grounded in these insights, we propose Entropy-Aware Policy Optimization (EAPO) that modulates token-level learning signals accordingly. Extensive experiments demonstrate that EAPO outperforms strong baselines across two model families.

Jiayu Lin, Guanrong Chen, Bojun Jin et al.

In this paper we present the results of the AI-Debater 2023 Challenge held by the Chinese Conference on Affect Computing (CCAC 2023), and introduce the related datasets. We organize two tracks to handle the argumentative generation tasks in different scenarios, namely, Counter-Argument Generation (Track 1) and Claim-based Argument Generation (Track 2). Each track is equipped with its distinct dataset and baseline model respectively. In total, 32 competing teams register for the challenge, from which we received 11 successful submissions. In this paper, we will present the results of the challenge and a summary of the systems, highlighting commonalities and innovations among participating systems. Datasets and baseline models of the AI-Debater 2023 Challenge have been already released and can be accessed through the official website of the challenge.

Jizhou Han, Chenhao Ding, SongLin Dong et al.

Continual Generalized Category Discovery (C-GCD) requires identifying novel classes from unlabeled data while retaining knowledge of known classes over time. Existing methods typically update classifier weights dynamically, resulting in forgetting and inconsistent feature alignment. We propose GOAL, a unified framework that introduces a fixed Equiangular Tight Frame (ETF) classifier to impose a consistent geometric structure throughout learning. GOAL conducts supervised alignment for labeled samples and confidence-guided alignment for novel samples, enabling stable integration of new classes without disrupting old ones. Experiments on four benchmarks show that GOAL outperforms the prior method Happy, reducing forgetting by 16.1% and boosting novel class discovery by 3.2%, establishing a strong solution for long-horizon continual discovery.

Shaokun Wang, Yifan Yu, Yuhang He et al.

Recently, adapting pre-trained models to downstream tasks has attracted increasing interest. Previous Parameter-Efficient-Tuning (PET) methods regard the pre-trained model as an opaque Black Box model, relying purely on data-driven optimization and underutilizing their inherent prior knowledge. This oversight limits the models' potential for effective downstream task adaptation. To address these issues, we propose a novel black-whIte bOx prompT leArning framework (IOTA), which integrates a data-driven Black Box module with a knowledge-driven White Box module for downstream task adaptation. Specifically, the White Box module derives corrective knowledge by contrasting the wrong predictions with the right cognition. This knowledge is verbalized into interpretable human prompts and leveraged through a corrective knowledge-guided prompt selection strategy to guide the Black Box module toward more accurate predictions. By jointly leveraging knowledge- and data-driven learning signals, IOTA achieves effective downstream task adaptation. Experimental results on 12 image classification benchmarks under few-shot and easy-to-hard adaptation settings demonstrate the effectiveness of corrective knowledge and the superiority of our method over state-of-the-art methods.

Xianda Guo, Ruijun Zhang, Yiqun Duan et al.

Accurate spatial reasoning in outdoor environments - covering geometry, object pose, and inter-object relationships - is fundamental to downstream tasks such as mapping, motion forecasting, and high-level planning in autonomous driving. We introduce SURDS, a large-scale benchmark designed to systematically evaluate the spatial reasoning capabilities of vision language models (VLMs). Built on the nuScenes dataset, SURDS comprises 41,080 vision-question-answer training instances and 9,250 evaluation samples, spanning six spatial categories: orientation, depth estimation, pixel-level localization, pairwise distance, lateral ordering, and front-behind relations. We benchmark leading general-purpose VLMs, including GPT, Gemini, and Qwen, revealing persistent limitations in fine-grained spatial understanding. To address these deficiencies, we go beyond static evaluation and explore whether alignment techniques can improve spatial reasoning performance. Specifically, we propose a reinforcement learning-based alignment scheme leveraging spatially grounded reward signals - capturing both perception-level accuracy (location) and reasoning consistency (logic). We further incorporate final-answer correctness and output-format rewards to guide fine-grained policy adaptation. Our GRPO-aligned variant achieves an overall score of 40.80 in the SURDS benchmark. Notably, it outperforms proprietary systems such as GPT-4o (13.30) and Gemini-2.0-flash (35.71). To our best knowledge, this is the first study to demonstrate that reinforcement learning-based alignment can significantly and consistently enhance the spatial reasoning capabilities of VLMs in real-world driving contexts. We release the SURDS benchmark, evaluation toolkit, and GRPO alignment code through: https://github.com/XiandaGuo/Drive-MLLM.

Shiwei Lu, Yuhang He, Jiashuo Li et al.

Federated learning (FL) facilitates the secure utilization of decentralized images, advancing applications in medical image recognition and autonomous driving. However, conventional FL faces two critical challenges in real-world deployment: ineffective knowledge fusion caused by model updates biased toward majority-class features, and prohibitive communication overhead due to frequent transmissions of high-dimensional model parameters. Inspired by the human brain's efficiency in knowledge integration, we propose a novel Generative Federated Prototype Learning (GFPL) framework to address these issues. Within this framework, a prototype generation method based on Gaussian Mixture Model (GMM) captures the statistical information of class-wise features, while a prototype aggregation strategy using Bhattacharyya distance effectively fuses semantically similar knowledge across clients. In addition, these fused prototypes are leveraged to generate pseudo-features, thereby mitigating feature distribution imbalance across clients. To further enhance feature alignment during local training, we devise a dual-classifier architecture, optimized via a hybrid loss combining Dot Regression and Cross-Entropy. Extensive experiments on benchmarks show that GFPL improves model accuracy by 3.6% under imbalanced data settings while maintaining low communication cost.

Qiang Wang, Xiang Song, Yuhang He et al.

Deep neural networks (DNNs) often underperform in real-world, dynamic settings where data distributions change over time. Domain Incremental Learning (DIL) offers a solution by enabling continual model adaptation, with Parameter-Isolation DIL (PIDIL) emerging as a promising paradigm to reduce knowledge conflicts. However, existing PIDIL methods struggle with parameter selection accuracy, especially as the number of domains and corresponding classes grows. To address this, we propose SOYO, a lightweight framework that improves domain selection in PIDIL. SOYO introduces a Gaussian Mixture Compressor (GMC) and Domain Feature Resampler (DFR) to store and balance prior domain data efficiently, while a Multi-level Domain Feature Fusion Network (MDFN) enhances domain feature extraction. Our framework supports multiple Parameter-Efficient Fine-Tuning (PEFT) methods and is validated across tasks such as image classification, object detection, and speech enhancement. Experimental results on six benchmarks demonstrate SOYO's consistent superiority over existing baselines, showcasing its robustness and adaptability in complex, evolving environments. The codes will be released in https://github.com/qwangcv/SOYO.

Yuhang He, Yash Jain, Xubo Liu et al.

Despite significant advancements in Text-to-Audio (TTA) generation models achieving high-fidelity audio with fine-grained context understanding, they struggle to model the relations between audio events described in the input text. However, previous TTA methods have not systematically explored audio event relation modeling, nor have they proposed frameworks to enhance this capability. In this work, we systematically study audio event relation modeling in TTA generation models. We first establish a benchmark for this task by: 1. proposing a comprehensive relation corpus covering all potential relations in real-world scenarios; 2. introducing a new audio event corpus encompassing commonly heard audios; and 3. proposing new evaluation metrics to assess audio event relation modeling from various perspectives. Furthermore, we propose a finetuning framework to enhance existing TTA models ability to model audio events relation. Code is available at: https://github.com/yuhanghe01/RiTTA

Huoren Yang, Jianchao Zhao, Hu Yusong et al.

Vision-Language-Action (VLA) models have advanced rapidly with stronger backbones, broader pre-training, and larger demonstration datasets, yet their action heads remain largely homogeneous: most directly predict action commands in a fixed world coordinate frame. We propose \textbf{MCF-Proto}, a lightweight action head that equips VLA policies with a Motion-Centric Action Frame (MCF) and a prototype-based action parameterization. At each step, the policy predicts a rotation $R_t \in SO(3)$, composes actions in the transformed local frame from a set of prototypes, and maps them back to the world frame for end-to-end training, using only standard demonstrations without auxiliary supervision. This simple design induces stable emergent structure. Without explicit directional labels, the learned local frames develop a stable geometric structure whose axes are strongly compatible with demonstrated end-effector motion. Meanwhile, actions in the learned representation become substantially more compact, with variation captured by fewer dominant directions and more regularly organized by shared prototypes. These structural properties translate into improved robustness, especially under geometric perturbations. Our results suggest that adding lightweight geometric and compositional structure to the action head can materially improve how VLA policies organize and generalize robotic manipulation behavior. An anonymized code repository is provided in the supplementary material.

Amirhossein Abaskohi, Yuhang He, Peter West et al.

Computer-use agents~(CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed using a learned patch selector that compares patch representations across consecutive screenshots while preserving spatial structure required by the model. Across three benchmarks, OSWorld, WebTailBench, and AgentNetBench, when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by approximately 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens. With this improved efficiency, we revisit the role of history in CUAs and find that performance continues to improve as more past observations are incorporated when redundancy is removed. This suggests that the commonly observed saturation in visual history is not due to limited usefulness of past information, but rather a consequence of inefficient token representations.

Yuanxing Duan, Fangyin Wei, Qiyu Dai et al.

We consider the problem of novel-view synthesis (NVS) for dynamic scenes. Recent neural approaches have accomplished exceptional NVS results for static 3D scenes, but extensions to 4D time-varying scenes remain non-trivial. Prior efforts often encode dynamics by learning a canonical space plus implicit or explicit deformation fields, which struggle in challenging scenarios like sudden movements or generating high-fidelity renderings. In this paper, we introduce 4D Gaussian Splatting (4DRotorGS), a novel method that represents dynamic scenes with anisotropic 4D XYZT Gaussians, inspired by the success of 3D Gaussian Splatting in static scenes. We model dynamics at each timestamp by temporally slicing the 4D Gaussians, which naturally compose dynamic 3D Gaussians and can be seamlessly projected into images. As an explicit spatial-temporal representation, 4DRotorGS demonstrates powerful capabilities for modeling complicated dynamics and fine details--especially for scenes with abrupt motions. We further implement our temporal slicing and splatting techniques in a highly optimized CUDA acceleration framework, achieving real-time inference rendering speeds of up to 277 FPS on an RTX 3090 GPU and 583 FPS on an RTX 4090 GPU. Rigorous evaluations on scenes with diverse motions showcase the superior efficiency and effectiveness of 4DRotorGS, which consistently outperforms existing methods both quantitatively and qualitatively.

Xiangyu Bai, He Liang, Bishoy Galoaa et al.

While text-to-video (T2V) generation has achieved remarkable progress in photorealism, generating intent-aligned videos that faithfully obey physics principles remains a core challenge. In this work, we systematically study Newtonian motion-controlled text-to-video generation and evaluation, emphasizing physical precision and motion coherence. We introduce MoReGen, a motion-aware, physics-grounded T2V framework that integrates multi-agent LLMs, physics simulators, and renderers to generate reproducible, physically accurate videos from text prompts in the code domain. To quantitatively assess physical validity, we propose object-trajectory correspondence as a direct evaluation metric and present MoReSet, a benchmark of 1,275 human-annotated videos spanning nine classes of Newtonian phenomena with scene descriptions, spatiotemporal relations, and ground-truth trajectories. Using MoReSet, we conduct experiments on existing T2V models, evaluating their physical validity through both our MoRe metrics and existing physics-based evaluators. Our results reveal that state-of-the-art models struggle to maintain physical validity, while MoReGen establishes a principled direction toward physically coherent video synthesis.

Xin Liu, Yuhang He, Sichen Zhao et al.

Root cause localization in cloud native microservice systems requires modeling complex service dependencies, irregular temporal dynamics, and heterogeneous observability data. We present HyperODE RCA, a unified framework that combines hypergraph attention learning, latent ordinary differential equations, and multimodal cross attention fusion for fine grained root cause analysis. The method learns higher order service interactions through differentiable hyperedge construction, captures continuous anomaly evolution from irregular observations with an ODE RNN encoder, and adaptively fuses logs, traces, metrics, entities, and events using context aware modality routing. We further improve robustness with a variational information bottleneck, temporal causal regularization, and invariant risk constraints. Experiments on the Tianchi AIOps benchmark show clear gains over strong baselines in ranking and classification performance, while preserving interpretability through learned hypergraph attention.

Xiaonan Xu, Yichao Wu, Penghao Liang et al.

With the boom of e-commerce and web applications, recommender systems have become an important part of our daily lives, providing personalized recommendations based on the user's preferences. Although deep neural networks (DNNs) have made significant progress in improving recommendation systems by simulating the interaction between users and items and incorporating their textual information, these DNN-based approaches still have some limitations, such as the difficulty of effectively understanding users' interests and capturing textual information. It is not possible to generalize to different seen/unseen recommendation scenarios and reason about their predictions. At the same time, the emergence of large language models (LLMs), represented by ChatGPT and GPT-4, has revolutionized the fields of natural language processing (NLP) and artificial intelligence (AI) due to their superior capabilities in the basic tasks of language understanding and generation, and their impressive generalization and reasoning capabilities. As a result, recent research has sought to harness the power of LLM to improve recommendation systems. Given the rapid development of this research direction in the field of recommendation systems, there is an urgent need for a systematic review of existing LLM-driven recommendation systems for researchers and practitioners in related fields to gain insight into. More specifically, we first introduced a representative approach to learning user and item representations using LLM as a feature encoder. We then reviewed the latest advances in LLMs techniques for collaborative filtering enhanced recommendation systems from the three paradigms of pre-training, fine-tuning, and prompting. Finally, we had a comprehensive discussion on the future direction of this emerging field.

Xinyuan Gao, Songlin Dong, Yuhang He et al.

In real-world applications, dynamic scenarios require the models to possess the capability to learn new tasks continuously without forgetting the old knowledge. Experience-Replay methods store a subset of the old images for joint training. In the scenario of more strict privacy protection, storing the old images becomes infeasible, which leads to a more severe plasticity-stability dilemma and classifier bias. To meet the above challenges, we propose a new architecture, named continual expansion and absorption transformer~(CEAT). The model can learn the novel knowledge by extending the expanded-fusion layers in parallel with the frozen previous parameters. After the task ends, we losslessly absorb the extended parameters into the backbone to ensure that the number of parameters remains constant. To improve the learning ability of the model, we designed a novel prototype contrastive loss to reduce the overlap between old and new classes in the feature space. Besides, to address the classifier bias towards the new classes, we propose a novel approach to generate the pseudo-features to correct the classifier. We experiment with our methods on three standard Non-Exemplar Class-Incremental Learning~(NECIL) benchmarks. Extensive experiments demonstrate that our model gets a significant improvement compared with the previous works and achieves 5.38%, 5.20%, and 4.92% improvement on CIFAR-100, TinyImageNet, and ImageNet-Subset.

Qiang Wang, Yuhang He, SongLin Dong et al.

Domain-Incremental Learning (DIL) enables vision models to adapt to changing conditions in real-world environments while maintaining the knowledge acquired from previous domains. Given privacy concerns and training time, Rehearsal-Free DIL (RFDIL) is more practical. Inspired by the incremental cognitive process of the human brain, we design Dual-level Concept Prototypes (DualCP) for each class to address the conflict between learning new knowledge and retaining old knowledge in RFDIL. To construct DualCP, we propose a Concept Prototype Generator (CPG) that generates both coarse-grained and fine-grained prototypes for each class. Additionally, we introduce a Coarse-to-Fine calibrator (C2F) to align image features with DualCP. Finally, we propose a Dual Dot-Regression (DDR) loss function to optimize our C2F module. Extensive experiments on the DomainNet, CDDB, and CORe50 datasets demonstrate the effectiveness of our method.

Songlin Dong, Yuhang He, Zhengdong Zhou et al.

Current research on class-incremental learning primarily focuses on single-label classification tasks. However, real-world applications often involve multi-label scenarios, such as image retrieval and medical imaging. Therefore, this paper focuses on the challenging yet practical multi-label class-incremental learning (MLCIL) problem. In addition to the challenge of catastrophic forgetting, MLCIL encounters issues related to feature confusion, encompassing inter-session and intra-feature confusion. To address these problems, we propose a novel MLCIL approach called class-independent increment (CLIN). Specifically, in contrast to existing methods that extract image-level features, we propose a class-independent incremental network (CINet) to extract multiple class-level embeddings for multi-label samples. It learns and preserves the knowledge of different classes by constructing class-specific tokens. On this basis, we develop two novel loss functions, optimizing the learning of class-specific tokens and class-level embeddings, respectively. These losses aim to distinguish between new and old classes, further alleviating the problem of feature confusion. Extensive experiments on MS-COCO and PASCAL VOC datasets demonstrate the effectiveness of our method for improving recognition performance and mitigating forgetting on various MLCIL tasks.

Yifan Yu, Shaokun Wang, Yuhang He et al.

Continual Novel Class Discovery (CNCD) aims to continually discover novel classes without labels while maintaining the recognition capability for previously learned classes. The main challenges faced by CNCD include the feature-discrepancy problem, the inter-session confusion problem, etc. In this paper, we propose a novel Feature Enhancement and Adaptation method for the CNCD to tackle the above challenges, which consists of a guide-to-novel framework, a centroid-to-samples similarity constraint (CSS), and a boundary-aware prototype constraint (BAP). More specifically, the guide-to-novel framework is established to continually discover novel classes under the guidance of prior distribution. Afterward, the CSS is designed to constrain the relationship between centroid-to-samples similarities of different classes, thereby enhancing the distinctiveness of features among novel classes. Finally, the BAP is proposed to keep novel class features aware of the positions of other class prototypes during incremental sessions, and better adapt novel class features to the shared feature space. Experimental results on three benchmark datasets demonstrate the superiority of our method, especially in more challenging protocols with more incremental sessions.

Yuhang He

Positional encoding has become the de facto standard for grounding deep neural networks on discrete point-wise positions, and it has achieved remarkable success in tasks where the input can be represented as a one-dimensional sequence. However, extending this concept to 2D spatial geometric shapes demands carefully designed encoding strategies that account not only for shape geometry and pose, but also for compatibility with neural network learning. In this work, we address these challenges by introducing a training-free, general-purpose encoding strategy, dubbed XShapeEnc, that encodes an arbitrary spatially grounded 2D geometric shape into a compact representation exhibiting five favorable properties, including invertibility, adaptivity, and frequency richness. Specifically, a 2D spatially grounded geometric shape is decomposed into its normalized geometry within the unit disk and its pose vector, where the pose is further transformed into a harmonic pose field that also lies within the unit disk. A set of orthogonal Zernike bases is constructed to encode shape geometry and pose either independently or jointly, followed by a frequency-propagation operation to introduce high-frequency content into the encoding. We demonstrate the theoretical validity, efficiency, discriminability, and applicability of XShapeEnc via extensive analysis and experiments across a wide range of shape-aware tasks and our self-curated XShapeCorpus. We envision XShapeEnc as a foundational tool for research that goes beyond one-dimensional sequential data toward frontier 2D spatial intelligence.

Songlin Dong, Chenhao Ding, Jiangyang Li et al.

This study aims to address the problem of multi-domain task incremental learning~(MTIL), which requires that vision-language models~(VLMs) continuously acquire new knowledge while maintaining their inherent zero-shot recognition capability. Existing paradigms delegate the testing of unseen-domain samples to the original CLIP, which only prevents the degradation of the model's zero-shot capability but fails to enhance the generalization of the VLM further. To this end, we propose a novel MTIL framework, named AFA, which comprises two core modules: (1) an against forward-forgetting adapter that learns task-invariant information for each dataset in the incremental tasks to enhance the zero-shot recognition ability of VLMs; (2) an against backward-forgetting adapter that strengthens the few-shot learning capability of VLMs while supporting incremental learning. Extensive experiments demonstrate that the AFA method significantly outperforms existing state-of-the-art approaches, especially in few-shot MTIL tasks, and surpasses the inherent zero-shot performance of CLIP in terms of transferability. The code is provided in the Supplementary Material.

Jizhou Han, Chenhao Ding, Yuhang He et al.

Few-shot class-incremental Learning (FSCIL) enables models to learn new classes from limited data while retaining performance on previously learned classes. Traditional FSCIL methods often require fine-tuning parameters with limited new class data and suffer from a separation between learning new classes and utilizing old knowledge. Inspired by the analogical learning mechanisms of the human brain, we propose a novel analogical generative method. Our approach includes the Brain-Inspired Analogical Generator (BiAG), which derives new class weights from existing classes without parameter fine-tuning during incremental stages. BiAG consists of three components: Weight Self-Attention Module (WSA), Weight & Prototype Analogical Attention Module (WPAA), and Semantic Conversion Module (SCM). SCM uses Neural Collapse theory for semantic conversion, WSA supplements new class weights, and WPAA computes analogies to generate new class weights. Experiments on miniImageNet, CUB-200, and CIFAR-100 datasets demonstrate that our method achieves higher final and average accuracy compared to SOTA methods.

Chenhao Ding, Xinyuan Gao, Songlin Dong et al.

With the development of visual-language models (VLM) in downstream task applications, test-time adaptation methods based on VLM have attracted increasing attention for their ability to address changes distribution in test-time. Although prior approaches have achieved some progress, they typically either demand substantial computational resources or are constrained by the limitations of the original feature space, rendering them less effective for test-time adaptation tasks. To address these challenges, we propose a training-free feature space rotation with basis transformation for test-time adaptation. By leveraging the inherent distinctions among classes, we reconstruct the original feature space and map it to a new representation, thereby enhancing the clarity of class differences and providing more effective guidance for the model during testing. Additionally, to better capture relevant information from various classes, we maintain a dynamic queue to store representative samples. Experimental results across multiple benchmarks demonstrate that our method outperforms state-of-the-art techniques in terms of both performance and efficiency.

Chenhao Ding, Xinyuan Gao, Songlin Dong et al.

Existing prompt learning methods in Vision-Language Models (VLM) have effectively enhanced the transfer capability of VLM to downstream tasks, but they suffer from a significant decline in generalization due to severe overfitting. To address this issue, we propose a framework named LOBG for vision-language models. Specifically, we use CLIP to filter out fine-grained foreground information that might cause overfitting, thereby guiding prompts with basic visual concepts. To further mitigate overfitting, we devel oped a structural topology preservation (STP) loss at the feature level, which endows the feature space with overall plasticity, allowing effective reshaping of the feature space during optimization. Additionally, we employed hierarchical logit distilation (HLD) at the output level to constrain outputs, complementing STP at the output end. Extensive experimental results demonstrate that our method significantly improves generalization capability and alleviates overfitting compared to state-of-the-art approaches.

Songlin Dong, Jiangyang Li, Chenhao Ding et al.

Multi-label Class-Incremental Learning aims to continuously recognize novel categories in complex scenes where multiple objects co-occur. However, existing approaches often incur high computational costs due to full-parameter fine-tuning and substantial storage overhead from memory buffers, or they struggle to address feature confusion and domain discrepancies adequately. To overcome these limitations, we introduce P2L-CA, a parameter-efficient framework that integrates a Prompt-to-Label module with a Continuous Adapter module. The P2L module leverages class-specific prompts to disentangle multi-label representations while incorporating linguistic priors to enforce stable semantic-visual alignment. Meanwhile, the CA module employs lightweight adapters to mitigate domain gaps between pre-trained models and downstream tasks, thereby enhancing model plasticity. Extensive experiments across standard and challenging MLCIL settings on MS-COCO and PASCAL VOC show that P2L-CA not only achieves substantial improvements over state-of-the-art methods but also demonstrates strong generalization in CIL scenarios, all while requiring minimal trainable parameters and eliminating the need for memory buffers.

Jiangyang Li, Chenhao Ding, Songlin Dong et al.

Prompt-based continual learning methods effectively mitigate catastrophic forgetting. However, most existing methods assign a fixed set of prompts to each task, completely isolating knowledge across tasks and resulting in suboptimal parameter utilization. To address this, we consider the practical needs of continual learning and propose a prompt-sharing framework. This framework constructs a global prompt pool and introduces a task-aware gated routing mechanism that sparsely activates a subset of prompts to achieve dynamic decoupling and collaborative optimization of task-specific feature representations. Furthermore, we introduce a history-aware modulator that leverages cumulative prompt activation statistics to protect frequently used prompts from excessive updates, thereby mitigating inefficient parameter usage and knowledge forgetting. Extensive analysis and empirical results demonstrate that our approach consistently outperforms existing static allocation strategies in effectiveness and efficiency.

Qiang Wang, Xinyuan Gao, SongLin Dong et al.

We present VDC-Agent, a self-evolving framework for Video Detailed Captioning that requires neither human annotations nor larger teacher models. The agent forms a closed loop of caption generation, principle-guided scoring (score and textual suggestions), and prompt refinement. When caption quality regresses, a self-reflection path leverages the previous chain-of-thought to amend the update. Running this process on unlabeled videos produces trajectories of (caption, score) pairs. We convert the trajectories into preference tuples and filter out samples with JSON parsing errors, resulting in VDC-Agent-19K, which contains 18,886 automatically constructed pairs. We then fine-tune the base MLLM on this dataset using an easy-to-hard curriculum direct preference optimization. Built on Qwen2.5-VL-7B-Instruct, our VDC-Agent-7B attains state-of-the-art performance on the VDC benchmark with 49.08% average accuracy and 2.50 score, surpassing specialized video captioners and improving over the base model by +5.13% accuracy and +0.27 score at similar inference cost.

Jizhou Han, Shaokun Wang, Yuhang He et al.

Generalized Category Discovery (GCD) focuses on classifying known categories while simultaneously discovering novel categories from unlabeled data. However, previous GCD methods face challenges due to inconsistent optimization objectives and category confusion. This leads to feature overlap and ultimately hinders performance on novel categories. To address these issues, we propose the Neural Collapse-inspired Generalized Category Discovery (NC-GCD) framework. By pre-assigning and fixing Equiangular Tight Frame (ETF) prototypes, our method ensures an optimal geometric structure and a consistent optimization objective for both known and novel categories. We introduce a Consistent ETF Alignment Loss that unifies supervised and unsupervised ETF alignment and enhances category separability. Additionally, a Semantic Consistency Matcher (SCM) is designed to maintain stable and consistent label assignments across clustering iterations. Our method achieves strong performance on multiple GCD benchmarks, significantly enhancing novel category accuracy and demonstrating its effectiveness.

JianChao Zhao, Chenhao Ding, Songlin Dong et al.

Continual Test-Time Adaptation (CTTA) aims to enable models to adapt on-the-fly to a stream of unlabeled data under evolving distribution shifts. However, existing CTTA methods typically rely on shared model parameters across all domains, making them vulnerable to feature entanglement and catastrophic forgetting in the presence of large or non-stationary domain shifts. To address this limitation, we propose ExPaMoE, a novel framework based on an Expandable Parallel Mixture-of-Experts architecture. ExPaMoE decouples domain-general and domain-specific knowledge via a dual-branch expert design with token-guided feature separation, and dynamically expands its expert pool based on a Spectral-Aware Online Domain Discriminator (SODD) that detects distribution changes in real-time using frequency-domain cues. Extensive experiments demonstrate the superiority of ExPaMoE across diverse CTTA scenarios. We evaluate our method on standard benchmarks including CIFAR-10C, CIFAR-100C, ImageNet-C, and Cityscapes-to-ACDC for semantic segmentation. Additionally, we introduce ImageNet++, a large-scale and realistic CTTA benchmark built from multiple ImageNet-derived datasets, to better reflect long-term adaptation under complex domain evolution. ExPaMoE consistently outperforms prior arts, showing strong robustness, scalability, and resistance to forgetting.

Jizhou Han, Chenhao Ding, SongLin Dong et al.

Visual-language models (VLMs) like CLIP exhibit strong generalization but struggle with distribution shifts at test time. Existing training-free test-time adaptation (TTA) methods operate strictly within CLIP's original feature space, relying on high-confidence samples while overlooking the potential of low-confidence ones. We propose MS-TTA, a training-free approach that enhances feature representations beyond CLIP's space using a single-step k-nearest neighbors (kNN) Mean-Shift. By refining all test samples, MS-TTA improves feature compactness and class separability, leading to more stable adaptation. Additionally, a cache of refined embeddings further enhances inference by providing Mean Shift enhanced logits. Extensive evaluations on OOD and cross-dataset benchmarks demonstrate that MS-TTA consistently outperforms state-of-the-art training-free TTA methods, achieving robust adaptation without requiring additional training.

Yuhang He, Sangyun Shin, Anoop Cherian et al.

Accurately localizing 3D sound sources and estimating their semantic labels -- where the sources may not be visible, but are assumed to lie on the physical surface of objects in the scene -- have many real applications, including detecting gas leak and machinery malfunction. The audio-visual weak-correlation in such setting poses new challenges in deriving innovative methods to answer if or how we can use cross-modal information to solve the task. Towards this end, we propose to use an acoustic-camera rig consisting of a pinhole RGB-D camera and a coplanar four-channel microphone array~(Mic-Array). By using this rig to record audio-visual signals from multiviews, we can use the cross-modal cues to estimate the sound sources 3D locations. Specifically, our framework SoundLoc3D treats the task as a set prediction problem, each element in the set corresponds to a potential sound source. Given the audio-visual weak-correlation, the set representation is initially learned from a single view microphone array signal, and then refined by actively incorporating physical surface cues revealed from multiview RGB-D images. We demonstrate the efficiency and superiority of SoundLoc3D on large-scale simulated dataset, and further show its robustness to RGB-D measurement inaccuracy and ambient noise interference.

Cong Wan, Xiangyang Luo, Hao Luo et al.

Visual generation has witnessed remarkable progress in single-image tasks, yet extending these capabilities to temporal sequences remains challenging. Current approaches either build specialized video models from scratch with enormous computational costs or add separate motion modules to image generators, both requiring learning temporal dynamics anew. We observe that modern image generation models possess underutilized potential in handling structured layouts with implicit temporal understanding. Building on this insight, we introduce GRID, which reformulates temporal sequences as grid layouts, enabling holistic processing of visual sequences while leveraging existing model capabilities. Through a parallel flow-matching training strategy with coarse-to-fine scheduling, our approach achieves up to 67 faster inference speeds while using <1/1000 of the computational resources compared to specialized models. Extensive experiments demonstrate that GRID not only excels in temporal tasks from Text-to-Video to 3D Editing but also preserves strong performance in image generation, establishing itself as an efficient and versatile omni-solution for visual generation.

Yuhang He, Shitong Xu, Jia-Xing Zhong et al.

We present SPEAR, a continuous receiver-to-receiver acoustic neural warping field for spatial acoustic effects prediction in an acoustic 3D space with a single stationary audio source. Unlike traditional source-to-receiver modelling methods that require prior space acoustic properties knowledge to rigorously model audio propagation from source to receiver, we propose to predict by warping the spatial acoustic effects from one reference receiver position to another target receiver position, so that the warped audio essentially accommodates all spatial acoustic effects belonging to the target position. SPEAR can be trained in a data much more readily accessible manner, in which we simply ask two robots to independently record spatial audio at different positions. We further theoretically prove the universal existence of the warping field if and only if one audio source presents. Three physical principles are incorporated to guide SPEAR network design, leading to the learned warping field physically meaningful. We demonstrate SPEAR superiority on both synthetic, photo-realistic and real-world dataset, showing the huge potential of SPEAR to various down-stream robotic tasks.

Songlin Dong, Yingjie Chen, Yuhang He et al.

Online task-free continual learning (OTFCL) is a more challenging variant of continual learning which emphasizes the gradual shift of task boundaries and learns in an online mode. Existing methods rely on a memory buffer composed of old samples to prevent forgetting. However,the use of memory buffers not only raises privacy concerns but also hinders the efficient learning of new samples. To address this problem, we propose a novel framework called I2CANSAY that gets rid of the dependence on memory buffers and efficiently learns the knowledge of new data from one-shot samples. Concretely, our framework comprises two main modules. Firstly, the Inter-Class Analogical Augmentation (ICAN) module generates diverse pseudo-features for old classes based on the inter-class analogy of feature distributions for different new classes, serving as a substitute for the memory buffer. Secondly, the Intra-Class Significance Analysis (ISAY) module analyzes the significance of attributes for each class via its distribution standard deviation, and generates the importance vector as a correction bias for the linear classifier, thereby enhancing the capability of learning from new samples. We run our experiments on four popular image classification datasets: CoRe50, CIFAR-10, CIFAR-100, and CUB-200, our approach outperforms the prior state-of-the-art by a large margin.

Yongchao Huang, Yuhang He, Hong Ge

In this work, we introduce a novel framework which combines physics and machine learning methods to analyse acoustic signals. Three methods are developed for this task: a Bayesian inference approach for inferring the spectral acoustics characteristics, a neural-physical model which equips a neural network with forward and backward physical losses, and the non-linear least squares approach which serves as benchmark. The inferred propagation coefficient leads to the room impulse response (RIR) quantity which can be used for relocalisation with uncertainty. The simplicity and efficiency of this framework is empirically validated on simulated data.

Zhuangzhuang Dai, Yuhang He, Tran Vu et al.

Direction finding and positioning systems based on RF signals are significantly impacted by multipath propagation, particularly in indoor environments. Existing algorithms (e.g MUSIC) perform poorly in resolving Angle of Arrival (AoA) in the presence of multipath or when operating in a weak signal regime. We note that digitally sampled RF frontends allow for the easy analysis of signals, and their delayed components. Low-cost Software-Defined Radio (SDR) modules enable Channel State Information (CSI) extraction across a wide spectrum, motivating the design of an enhanced Angle-of-Arrival (AoA) solution. We propose a Deep Learning approach to deriving AoA from a single snapshot of the SDR multichannel data. We compare and contrast deep-learning based angle classification and regression models, to estimate up to two AoAs accurately. We have implemented the inference engines on different platforms to extract AoAs in real-time, demonstrating the computational tractability of our approach. To demonstrate the utility of our approach we have collected IQ (In-phase and Quadrature components) samples from a four-element Universal Linear Array (ULA) in various Light-of-Sight (LOS) and Non-Line-of-Sight (NLOS) environments, and published the dataset. Our proposed method demonstrates excellent reliability in determining number of impinging signals and realized mean absolute AoA errors less than $2^{\circ}$.

Yuhang He, Niki Trigoni, Andrew Markham

We present a new framework SoundDet, which is an end-to-end trainable and light-weight framework, for polyphonic moving sound event detection and localization. Prior methods typically approach this problem by preprocessing raw waveform into time-frequency representations, which is more amenable to process with well-established image processing pipelines. Prior methods also detect in segment-wise manner, leading to incomplete and partial detections. SoundDet takes a novel approach and directly consumes the raw, multichannel waveform and treats the spatio-temporal sound event as a complete "sound-object" to be detected. Specifically, SoundDet consists of a backbone neural network and two parallel heads for temporal detection and spatial localization, respectively. Given the large sampling rate of raw waveform, the backbone network first learns a set of phase-sensitive and frequency-selective bank of filters to explicitly retain direction-of-arrival information, whilst being highly computationally and parametrically efficient than standard 1D/2D convolution. A dense sound event proposal map is then constructed to handle the challenges of predicting events with large varying temporal duration. Accompanying the dense proposal map are a temporal overlapness map and a motion smoothness map that measure a proposal's confidence to be an event from temporal detection accuracy and movement consistency perspective. Involving the two maps guarantees SoundDet to be trained in a spatio-temporally unified manner. Experimental results on the public DCASE dataset show the advantage of SoundDet on both segment-based and our newly proposed event-based evaluation system.

Yuhang He, Wentao Yu, Jie Han et al.

In this paper, we focus on the multi-object tracking (MOT) problem of automatic driving and robot navigation. Most existing MOT methods track multiple objects using a singular RGB camera, which are prone to camera field-of-view and suffer tracking failures in complex scenarios due to background clutters and poor light conditions. To meet these challenges, we propose a MultiModality PAnoramic multi-object Tracking framework (MMPAT), which takes both 2D panorama images and 3D point clouds as input and then infers target trajectories using the multimodality data. The proposed method contains four major modules, a panorama image detection module, a multimodality data fusion module, a data association module and a trajectory inference model. We evaluate the proposed method on the JRDB dataset, where the MMPAT achieves the top performance in both the detection and tracking tasks and significantly outperforms state-of-the-art methods by a large margin (15.7 and 8.5 improvement in terms of AP and MOTA, respectively).