Chao Ma

Zongwei Wu, Jilai Zheng, Xiangxuan Ren et al.

In the realm of video object tracking, auxiliary modalities such as depth, thermal, or event data have emerged as valuable assets to complement the RGB trackers. In practice, most existing RGB trackers learn a single set of parameters to use them across datasets and applications. However, a similar single-model unification for multi-modality tracking presents several challenges. These challenges stem from the inherent heterogeneity of inputs -- each with modality-specific representations, the scarcity of multi-modal datasets, and the absence of all the modalities at all times. In this work, we introduce Un-Track, a Unified Tracker of a single set of parameters for any modality. To handle any modality, our method learns their common latent space through low-rank factorization and reconstruction techniques. More importantly, we use only the RGB-X pairs to learn the common latent space. This unique shared representation seamlessly binds all modalities together, enabling effective unification and accommodating any missing modality, all within a single transformer-based architecture. Our Un-Track achieves +8.1 absolute F-score gain, on the DepthTrack dataset, by introducing only +2.14 (over 21.50) GFLOPs with +6.6M (over 93M) parameters, through a simple yet efficient prompting strategy. Extensive comparisons on five benchmark datasets with different modalities show that Un-Track surpasses both SOTA unified trackers and modality-specific counterparts, validating our effectiveness and practicality. The source code is publicly available at https://github.com/Zongwei97/UnTrack.

Guangsheng Shi, Ruifeng Li, Chao Ma

Real-time and high-performance 3D object detection is of critical importance for autonomous driving. Recent top-performing 3D object detectors mainly rely on point-based or 3D voxel-based convolutions, which are both computationally inefficient for onboard deployment. In contrast, pillar-based methods use solely 2D convolutions, which consume less computation resources, but they lag far behind their voxel-based counterparts in detection accuracy. In this paper, by examining the primary performance gap between pillar- and voxel-based detectors, we develop a real-time and high-performance pillar-based detector, dubbed PillarNet.The proposed PillarNet consists of a powerful encoder network for effective pillar feature learning, a neck network for spatial-semantic feature fusion and the commonly used detect head. Using only 2D convolutions, PillarNet is flexible to an optional pillar size and compatible with classical 2D CNN backbones, such as VGGNet and ResNet. Additionally, PillarNet benefits from our designed orientation-decoupled IoU regression loss along with the IoU-aware prediction branch. Extensive experimental results on the large-scale nuScenes Dataset and Waymo Open Dataset demonstrate that the proposed PillarNet performs well over state-of-the-art 3D detectors in terms of effectiveness and efficiency. Code is available at \url{https://github.com/agent-sgs/PillarNet}.

Shuai Jia, Bangjie Yin, Taiping Yao et al.

Deep learning models have shown their vulnerability when dealing with adversarial attacks. Existing attacks almost perform on low-level instances, such as pixels and super-pixels, and rarely exploit semantic clues. For face recognition attacks, existing methods typically generate the l_p-norm perturbations on pixels, however, resulting in low attack transferability and high vulnerability to denoising defense models. In this work, instead of performing perturbations on the low-level pixels, we propose to generate attacks through perturbing on the high-level semantics to improve attack transferability. Specifically, a unified flexible framework, Adversarial Attributes (Adv-Attribute), is designed to generate inconspicuous and transferable attacks on face recognition, which crafts the adversarial noise and adds it into different attributes based on the guidance of the difference in face recognition features from the target. Moreover, the importance-aware attribute selection and the multi-objective optimization strategy are introduced to further ensure the balance of stealthiness and attacking strength. Extensive experiments on the FFHQ and CelebA-HQ datasets show that the proposed Adv-Attribute method achieves the state-of-the-art attacking success rates while maintaining better visual effects against recent attack methods.

Shenyuan Gao, Chunluan Zhou, Chao Ma et al.

Transformer trackers have achieved impressive advancements recently, where the attention mechanism plays an important role. However, the independent correlation computation in the attention mechanism could result in noisy and ambiguous attention weights, which inhibits further performance improvement. To address this issue, we propose an attention in attention (AiA) module, which enhances appropriate correlations and suppresses erroneous ones by seeking consensus among all correlation vectors. Our AiA module can be readily applied to both self-attention blocks and cross-attention blocks to facilitate feature aggregation and information propagation for visual tracking. Moreover, we propose a streamlined Transformer tracking framework, dubbed AiATrack, by introducing efficient feature reuse and target-background embeddings to make full use of temporal references. Experiments show that our tracker achieves state-of-the-art performance on six tracking benchmarks while running at a real-time speed.

Xingyu Ren, Alexandros Lattas, Baris Gecer et al.

Learning a dense 3D model with fine-scale details from a single facial image is highly challenging and ill-posed. To address this problem, many approaches fit smooth geometries through facial prior while learning details as additional displacement maps or personalized basis. However, these techniques typically require vast datasets of paired multi-view data or 3D scans, whereas such datasets are scarce and expensive. To alleviate heavy data dependency, we present a robust texture-guided geometric detail recovery approach using only a single in-the-wild facial image. More specifically, our method combines high-quality texture completion with the powerful expressiveness of implicit surfaces. Initially, we inpaint occluded facial parts, generate complete textures, and build an accurate multi-view dataset of the same subject. In order to estimate the detailed geometry, we define an implicit signed distance function and employ a physically-based implicit renderer to reconstruct fine geometric details from the generated multi-view images. Our method not only recovers accurate facial details but also decomposes normals, albedos, and shading parts in a self-supervised way. Finally, we register the implicit shape details to a 3D Morphable Model template, which can be used in traditional modeling and rendering pipelines. Extensive experiments demonstrate that the proposed approach can reconstruct impressive facial details from a single image, especially when compared with state-of-the-art methods trained on large datasets.

Zongwei Wu, Guillaume Allibert, Fabrice Meriaudeau et al.

RGB-D saliency detection aims to fuse multi-modal cues to accurately localize salient regions. Existing works often adopt attention modules for feature modeling, with few methods explicitly leveraging fine-grained details to merge with semantic cues. Thus, despite the auxiliary depth information, it is still challenging for existing models to distinguish objects with similar appearances but at distinct camera distances. In this paper, from a new perspective, we propose a novel Hierarchical Depth Awareness network (HiDAnet) for RGB-D saliency detection. Our motivation comes from the observation that the multi-granularity properties of geometric priors correlate well with the neural network hierarchies. To realize multi-modal and multi-level fusion, we first use a granularity-based attention scheme to strengthen the discriminatory power of RGB and depth features separately. Then we introduce a unified cross dual-attention module for multi-modal and multi-level fusion in a coarse-to-fine manner. The encoded multi-modal features are gradually aggregated into a shared decoder. Further, we exploit a multi-scale loss to take full advantage of the hierarchical information. Extensive experiments on challenging benchmark datasets demonstrate that our HiDAnet performs favorably over the state-of-the-art methods by large margins.

Qinglong Cao, Zhengqin Xu, Yuntian Chen et al.

Large pre-trained vision-language models, such as CLIP, have shown remarkable generalization capabilities across various tasks when appropriate text prompts are provided. However, adapting these models to specific domains, like remote sensing images (RSIs), medical images, etc, remains unexplored and challenging. Existing prompt learning methods often lack domain-awareness or domain-transfer mechanisms, leading to suboptimal performance due to the misinterpretation of specific images in natural image patterns. To tackle this dilemma, we proposed a \textbf{Domain-Controlled Prompt Learning} for the specific domains. Specifically, the large-scale specific domain foundation model (LSDM) is first introduced to provide essential specific domain knowledge. Using lightweight neural networks, we transfer this knowledge into domain biases, which control both the visual and language branches to obtain domain-adaptive prompts in a directly incorporating manner. Simultaneously, to overcome the existing overfitting challenge, we propose a novel noisy-adding strategy, without extra trainable parameters, to help the model escape the suboptimal solution in a global domain oscillation manner. Experimental results show our method achieves state-of-the-art performance in specific domain image recognition datasets. Our code is available at https://github.com/caoql98/DCPL.

Daniel Kunin, Atsushi Yamamura, Chao Ma et al.

In this work, we explore the maximum-margin bias of quasi-homogeneous neural networks trained with gradient flow on an exponential loss and past a point of separability. We introduce the class of quasi-homogeneous models, which is expressive enough to describe nearly all neural networks with homogeneous activations, even those with biases, residual connections, and normalization layers, while structured enough to enable geometric analysis of its gradient dynamics. Using this analysis, we generalize the existing results of maximum-margin bias for homogeneous networks to this richer class of models. We find that gradient flow implicitly favors a subset of the parameters, unlike in the case of a homogeneous model where all parameters are treated equally. We demonstrate through simple examples how this strong favoritism toward minimizing an asymmetric norm can degrade the robustness of quasi-homogeneous models. On the other hand, we conjecture that this norm-minimization discards, when possible, unnecessary higher-order parameters, reducing the model to a sparser parameterization. Lastly, by applying our theorem to sufficiently expressive neural networks with normalization layers, we reveal a universal mechanism behind the empirical phenomenon of Neural Collapse.

Cheng Zhang, Stefan Bauer, Paul Bennett et al.

We assess the ability of large language models (LLMs) to answer causal questions by analyzing their strengths and weaknesses against three types of causal question. We believe that current LLMs can answer causal questions with existing causal knowledge as combined domain experts. However, they are not yet able to provide satisfactory answers for discovering new knowledge or for high-stakes decision-making tasks with high precision. We discuss possible future directions and opportunities, such as enabling explicit and implicit causal modules as well as deep causal-aware LLMs. These will not only enable LLMs to answer many different types of causal questions for greater impact but also enable LLMs to be more trustworthy and efficient in general.

Shuai Jia, Chao Ma, Taiping Yao et al.

Various facial manipulation techniques have drawn serious public concerns in morality, security, and privacy. Although existing face forgery classifiers achieve promising performance on detecting fake images, these methods are vulnerable to adversarial examples with injected imperceptible perturbations on the pixels. Meanwhile, many face forgery detectors always utilize the frequency diversity between real and fake faces as a crucial clue. In this paper, instead of injecting adversarial perturbations into the spatial domain, we propose a frequency adversarial attack method against face forgery detectors. Concretely, we apply discrete cosine transform (DCT) on the input images and introduce a fusion module to capture the salient region of adversary in the frequency domain. Compared with existing adversarial attacks (e.g. FGSM, PGD) in the spatial domain, our method is more imperceptible to human observers and does not degrade the visual quality of the original images. Moreover, inspired by the idea of meta-learning, we also propose a hybrid adversarial attack that performs attacks in both the spatial and frequency domains. Extensive experiments indicate that the proposed method fools not only the spatial-based detectors but also the state-of-the-art frequency-based detectors effectively. In addition, the proposed frequency attack enhances the transferability across face forgery detectors as black-box attacks.

Chao Ma, Daniel Kunin, Lei Wu et al.

A quadratic approximation of neural network loss landscapes has been extensively used to study the optimization process of these networks. Though, it usually holds in a very small neighborhood of the minimum, it cannot explain many phenomena observed during the optimization process. In this work, we study the structure of neural network loss functions and its implication on optimization in a region beyond the reach of a good quadratic approximation. Numerically, we observe that neural network loss functions possesses a multiscale structure, manifested in two ways: (1) in a neighborhood of minima, the loss mixes a continuum of scales and grows subquadratically, and (2) in a larger region, the loss shows several separate scales clearly. Using the subquadratic growth, we are able to explain the Edge of Stability phenomenon [5] observed for the gradient descent (GD) method. Using the separate scales, we explain the working mechanism of learning rate decay by simple examples. Finally, we study the origin of the multiscale structure and propose that the non-convexity of the models and the non-uniformity of training data is one of the causes. By constructing a two-layer neural network problem we show that training data with different magnitudes give rise to different scales of the loss function, producing subquadratic growth and multiple separate scales.

Guangsheng Shi, Ruifeng Li, Chao Ma

The performance of point cloud 3D object detection hinges on effectively representing raw points, grid-based voxels or pillars. Recent two-stage 3D detectors typically take the point-voxel-based R-CNN paradigm, i.e., the first stage resorts to the 3D voxel-based backbone for 3D proposal generation on bird-eye-view (BEV) representation and the second stage refines them via the intermediate point representation. Their primary mechanisms involve the utilization of intermediary keypoints to restore the substantial 3D structure context from the converted BEV representation. The skilled point-voxel feature interaction, however, makes the entire detection pipeline more complex and compute-intensive. In this paper, we take a different viewpoint -- the pillar-based BEV representation owns sufficient capacity to preserve the 3D structure. In light of the latest advances in BEV-based perception, we devise a conceptually simple yet effective two-stage 3D detection architecture, named Pillar R-CNN. On top of densified BEV feature maps, Pillar R-CNN can easily introduce the feature pyramid architecture to generate 3D proposals at various scales and take the simple 2D R-CNN style detect head for box refinement. Our Pillar R-CNN performs favorably against state-of-the-art 3D detectors on the large-scale Waymo Open Dataset but at a small extra cost. It should be highlighted that further exploration into BEV perception for applications involving autonomous driving is now possible thanks to the effective and elegant Pillar R-CNN architecture.

Qinglong Cao, Yuntian Chen, Chao Ma et al.

Open-vocabulary image semantic segmentation (OVS) seeks to segment images into semantic regions across an open set of categories. Existing OVS methods commonly depend on foundational vision-language models and utilize similarity computation to tackle OVS tasks. However, these approaches are predominantly tailored to natural images and struggle with the unique characteristics of remote sensing images, such as rapidly changing orientations and significant scale variations. These challenges complicate OVS tasks in earth vision, requiring specialized approaches. To tackle this dilemma, we propose the first OVS framework specifically designed for remote sensing imagery, drawing inspiration from the distinct remote sensing traits. Particularly, to address the varying orientations, we introduce a rotation-aggregative similarity computation module that generates orientation-adaptive similarity maps as initial semantic maps. These maps are subsequently refined at both spatial and categorical levels to produce more accurate semantic maps. Additionally, to manage significant scale changes, we integrate multi-scale image features into the upsampling process, resulting in the final scale-aware semantic masks. To advance OVS in earth vision and encourage reproducible research, we establish the first open-sourced OVS benchmark for remote sensing imagery, including four public remote sensing datasets. Extensive experiments on this benchmark demonstrate our proposed method achieves state-of-the-art performance. All codes and datasets are available at https://github.com/caoql98/OVRS.

Ruibo Tu, Chao Ma, Cheng Zhang

ChatGPT has demonstrated exceptional proficiency in natural language conversation, e.g., it can answer a wide range of questions while no previous large language models can. Thus, we would like to push its limit and explore its ability to answer causal discovery questions by using a medical benchmark (Tu et al. 2019) in causal discovery.

Shengchao Zhou, Weizhou Liu, Chen Hu et al.

In the field of 3D object detection for autonomous driving, the sensor portfolio including multi-modality and single-modality is diverse and complex. Since the multi-modal methods have system complexity while the accuracy of single-modal ones is relatively low, how to make a tradeoff between them is difficult. In this work, we propose a universal cross-modality knowledge distillation framework (UniDistill) to improve the performance of single-modality detectors. Specifically, during training, UniDistill projects the features of both the teacher and the student detector into Bird's-Eye-View (BEV), which is a friendly representation for different modalities. Then, three distillation losses are calculated to sparsely align the foreground features, helping the student learn from the teacher without introducing additional cost during inference. Taking advantage of the similar detection paradigm of different detectors in BEV, UniDistill easily supports LiDAR-to-camera, camera-to-LiDAR, fusion-to-LiDAR and fusion-to-camera distillation paths. Furthermore, the three distillation losses can filter the effect of misaligned background information and balance between objects of different sizes, improving the distillation effectiveness. Extensive experiments on nuScenes demonstrate that UniDistill effectively improves the mAP and NDS of student detectors by 2.0%~3.2%.

Mingze Wang, Chao Ma

Generalization error bounds for deep neural networks trained by stochastic gradient descent (SGD) are derived by combining a dynamical control of an appropriate parameter norm and the Rademacher complexity estimate based on parameter norms. The bounds explicitly depend on the loss along the training trajectory, and work for a wide range of network architectures including multilayer perceptron (MLP) and convolutional neural networks (CNN). Compared with other algorithm-depending generalization estimates such as uniform stability-based bounds, our bounds do not require $L$-smoothness of the nonconvex loss function, and apply directly to SGD instead of Stochastic Langevin gradient descent (SGLD). Numerical results show that our bounds are non-vacuous and robust with the change of optimizer and network hyperparameters.

Zongwei Wu, Guillaume Allibert, Christophe Stolz et al.

Recent RGB-D semantic segmentation has motivated research interest thanks to the accessibility of complementary modalities from the input side. Existing works often adopt a two-stream architecture that processes photometric and geometric information in parallel, with few methods explicitly leveraging the contribution of depth cues to adjust the sampling position on RGB images. In this paper, we propose a novel framework to incorporate the depth information in the RGB convolutional neural network (CNN), termed Z-ACN (Depth-Adapted CNN). Specifically, our Z-ACN generates a 2D depth-adapted offset which is fully constrained by low-level features to guide the feature extraction on RGB images. With the generated offset, we introduce two intuitive and effective operations to replace basic CNN operators: depth-adapted convolution and depth-adapted average pooling. Extensive experiments on both indoor and outdoor semantic segmentation tasks demonstrate the effectiveness of our approach.

Chao Ma, Lexing Ying

In this paper, we show that structures similar to self-attention are natural to learn many sequence-to-sequence problems from the perspective of symmetry. Inspired by language processing applications, we study the orthogonal equivariance of seq2seq functions with knowledge, which are functions taking two inputs -- an input sequence and a ``knowledge'' -- and outputting another sequence. The knowledge consists of a set of vectors in the same embedding space as the input sequence, containing the information of the language used to process the input sequence. We show that orthogonal equivariance in the embedding space is natural for seq2seq functions with knowledge, and under such equivariance the function must take the form close to the self-attention. This shows that network structures similar to self-attention are the right structures to represent the target function of many seq2seq problems. The representation can be further refined if a ``finite information principle'' is considered, or a permutation equivariance holds for the elements of the input sequence.

Matthew Ashman, Chao Ma, Agrin Hilmkil et al.

Latent confounding has been a long-standing obstacle for causal reasoning from observational data. One popular approach is to model the data using acyclic directed mixed graphs (ADMGs), which describe ancestral relations between variables using directed and bidirected edges. However, existing methods using ADMGs are based on either linear functional assumptions or a discrete search that is complicated to use and lacks computational tractability for large datasets. In this work, we further extend the existing body of work and develop a novel gradient-based approach to learning an ADMG with non-linear functional relations from observational data. We first show that the presence of latent confounding is identifiable under the assumptions of bow-free ADMGs with non-linear additive noise models. With this insight, we propose a novel neural causal model based on autoregressive flows for ADMG learning. This not only enables us to determine complex causal structural relationships behind the data in the presence of latent confounding, but also estimate their functional relationships (hence treatment effects) simultaneously. We further validate our approach via experiments on both synthetic and real-world datasets, and demonstrate the competitive performance against relevant baselines.

Ming Cheng, Yiling Xu, Wang Shen et al.

High-speed, high-resolution stereoscopic (H2-Stereo) video allows us to perceive dynamic 3D content at fine granularity. The acquisition of H2-Stereo video, however, remains challenging with commodity cameras. Existing spatial super-resolution or temporal frame interpolation methods provide compromised solutions that lack temporal or spatial details, respectively. To alleviate this problem, we propose a dual camera system, in which one camera captures high-spatial-resolution low-frame-rate (HSR-LFR) videos with rich spatial details, and the other captures low-spatial-resolution high-frame-rate (LSR-HFR) videos with smooth temporal details. We then devise a Learned Information Fusion network (LIFnet) that exploits the cross-camera redundancies to enhance both camera views to high spatiotemporal resolution (HSTR) for reconstructing the H2-Stereo video effectively. We utilize a disparity network to transfer spatiotemporal information across views even in large disparity scenes, based on which, we propose disparity-guided flow-based warping for LSR-HFR view and complementary warping for HSR-LFR view. A multi-scale fusion method in feature domain is proposed to minimize occlusion-induced warping ghosts and holes in HSR-LFR view. The LIFnet is trained in an end-to-end manner using our collected high-quality Stereo Video dataset from YouTube. Extensive experiments demonstrate that our model outperforms existing state-of-the-art methods for both views on synthetic data and camera-captured real data with large disparity. Ablation studies explore various aspects, including spatiotemporal resolution, camera baseline, camera desynchronization, long/short exposures and applications, of our system to fully understand its capability for potential applications.

Mingze Wang, Chao Ma

The convergence of GD and SGD when training mildly parameterized neural networks starting from random initialization is studied. For a broad range of models and loss functions, including the most commonly used square loss and cross entropy loss, we prove an ``early stage convergence'' result. We show that the loss is decreased by a significant amount in the early stage of the training, and this decrease is fast. Furthurmore, for exponential type loss functions, and under some assumptions on the training data, we show global convergence of GD. Instead of relying on extreme over-parameterization, our study is based on a microscopic analysis of the activation patterns for the neurons, which helps us derive more powerful lower bounds for the gradient. The results on activation patterns, which we call ``neuron partition'', help build intuitions for understanding the behavior of neural networks' training dynamics, and may be of independent interest.

Yinglong Wang, Chao Ma, Jianzhuang Liu

Due to the difficulty in collecting paired real-world training data, image deraining is currently dominated by supervised learning with synthesized data generated by e.g., Photoshop rendering. However, the generalization to real rainy scenes is usually limited due to the gap between synthetic and real-world data. In this paper, we first statistically explore why the supervised deraining models cannot generalize well to real rainy cases, and find the substantial difference of synthetic and real rainy data. Inspired by our studies, we propose to remove rain by learning favorable deraining representations from other connected tasks. In connected tasks, the label for real data can be easily obtained. Hence, our core idea is to learn representations from real data through task transfer to improve deraining generalization. We thus term our learning strategy as \textit{task transfer learning}. If there are more than one connected tasks, we propose to reduce model size by knowledge distillation. The pretrained models for the connected tasks are treated as teachers, all their knowledge is distilled to a student network, so that we reduce the model size, meanwhile preserve effective prior representations from all the connected tasks. At last, the student network is fine-tuned with minority of paired synthetic rainy data to guide the pretrained prior representations to remove rain. Extensive experiments demonstrate that proposed task transfer learning strategy is surprisingly successful and compares favorably with state-of-the-art supervised learning methods and apparently surpass other semi-supervised deraining methods on synthetic data. Particularly, it shows superior generalization over them to real-world scenes.

Qinglong Cao, Yuntian Chen, Chao Ma et al.

Few-shot semantic segmentation (FSS) aims to segment objects of unseen classes in query images with only a few annotated support images. Existing FSS algorithms typically focus on mining category representations from the single-view support to match semantic objects of the single-view query. However, the limited annotated samples render the single-view matching struggle to perceive the reflection invariance of novel objects, which results in a restricted learning space for novel categories and further induces a biased segmentation with demoted parsing performance. To address this challenge, this paper proposes a fresh few-shot segmentation framework to mine the reflection invariance in a multi-view matching manner. Specifically, original and reflection support features from different perspectives with the same semantics are learnable fused to obtain the reflection invariance prototype with a stronger category representation ability. Simultaneously, aiming at providing better prior guidance, the Reflection Invariance Prior Mask Generation (RIPMG) module is proposed to integrate prior knowledge from different perspectives. Finally, segmentation predictions from varying views are complementarily merged in the Reflection Invariance Semantic Prediction (RISP) module to yield precise segmentation predictions. Extensive experiments on both PASCAL-$5^\textit{i}$ and COCO-$20^\textit{i}$ datasets demonstrate the effectiveness of our approach and show that our method could achieve state-of-the-art performance. Code is available at \url{https://anonymous.4open.science/r/RILFS-A4D1}

Xiaofeng Liu, Jonghye Woo, Chao Ma et al.

Delineating lesions and anatomical structure is important for image-guided interventions. Point-supervised medical image segmentation (PSS) has great potential to alleviate costly expert delineation labeling. However, due to the lack of precise size and boundary guidance, the effectiveness of PSS often falls short of expectations. Although recent vision foundational models, such as the medical segment anything model (MedSAM), have made significant advancements in bounding-box-prompted segmentation, it is not straightforward to utilize point annotation, and is prone to semantic ambiguity. In this preliminary study, we introduce an iterative framework to facilitate semantic-aware point-supervised MedSAM. Specifically, the semantic box-prompt generator (SBPG) module has the capacity to convert the point input into potential pseudo bounding box suggestions, which are explicitly refined by the prototype-based semantic similarity. This is then succeeded by a prompt-guided spatial refinement (PGSR) module that harnesses the exceptional generalizability of MedSAM to infer the segmentation mask, which also updates the box proposal seed in SBPG. Performance can be progressively improved with adequate iterations. We conducted an evaluation on BraTS2018 for the segmentation of whole brain tumors and demonstrated its superior performance compared to traditional PSS methods and on par with box-supervised methods.

Jiaqi Zhang, Joel Jennings, Agrin Hilmkil et al.

Foundation models have brought changes to the landscape of machine learning, demonstrating sparks of human-level intelligence across a diverse array of tasks. However, a gap persists in complex tasks such as causal inference, primarily due to challenges associated with intricate reasoning steps and high numerical precision requirements. In this work, we take a first step towards building causally-aware foundation models for treatment effect estimations. We propose a novel, theoretically justified method called Causal Inference with Attention (CInA), which utilizes multiple unlabeled datasets to perform self-supervised causal learning, and subsequently enables zero-shot causal inference on unseen tasks with new data. This is based on our theoretical results that demonstrate the primal-dual connection between optimal covariate balancing and self-attention, facilitating zero-shot causal inference through the final layer of a trained transformer-type architecture. We demonstrate empirically that CInA effectively generalizes to out-of-distribution datasets and various real-world datasets, matching or even surpassing traditional per-dataset methodologies. These results provide compelling evidence that our method has the potential to serve as a stepping stone for the development of causal foundation models.

YanZhao Zheng, ZhenTao Zhang, Chao Ma et al.

As LLM agent ecosystems grow, the number of available skills (tools, plugins) has reached tens of thousands, making it infeasible to inject all skills into an agent's context. This creates a need for skill routing -- retrieving the most relevant skills from a large pool given a user task. The problem is compounded by pervasive functional overlap in community skill repositories, where many skills share similar names and purposes yet differ in implementation details. Despite its practical importance, skill routing remains under-explored. Current agent architectures adopt a progressive disclosure design -- exposing only skill names and descriptions to the agent while keeping the full implementation body hidden -- implicitly treating metadata as sufficient for selection. We challenge this assumption through a systematic empirical study on a benchmark of ~$80K skills and 75 expert-verified queries. Our key finding is that the skill body (full implementation text) is the decisive signal: removing it causes 29--44 percentage point degradation across all retrieval methods, and cross-encoder attention analysis reveals 91.7% of attention concentrating on the body field. Motivated by this finding, we propose SkillRouter, a two-stage retrieve-and-rerank pipeline totaling only 1.2B parameters (0.6B encoder + 0.6B reranker). SkillRouter achieves 74.0% top-1 routing accuracy and delivers the strongest average result among the compact and zero-shot baselines we evaluate, while remaining deployable on consumer hardware.

Jilai Zheng, Pin Tang, Zhongdao Wang et al.

Perceiving the world as 3D occupancy supports embodied agents to avoid collision with any types of obstacle. While open-vocabulary image understanding has prospered recently, how to bind the predicted 3D occupancy grids with open-world semantics still remains under-explored due to limited open-world annotations. Hence, instead of building our model from scratch, we try to blend 2D foundation models, specifically a depth model MiDaS and a semantic model CLIP, to lift the semantics to 3D space, thus fulfilling 3D occupancy. However, building upon these foundation models is not trivial. First, the MiDaS faces the depth ambiguity problem, i.e., it only produces relative depth but fails to estimate bin depth for feature lifting. Second, the CLIP image features lack high-resolution pixel-level information, which limits the 3D occupancy accuracy. Third, open vocabulary is often trapped by the long-tail problem. To address these issues, we propose VEON for Vocabulary-Enhanced Occupancy predictioN by not only assembling but also adapting these foundation models. We first equip MiDaS with a Zoedepth head and low-rank adaptation (LoRA) for relative-metric-bin depth transformation while reserving beneficial depth prior. Then, a lightweight side adaptor network is attached to the CLIP vision encoder to generate high-resolution features for fine-grained 3D occupancy prediction. Moreover, we design a class reweighting strategy to give priority to the tail classes. With only 46M trainable parameters and zero manual semantic labels, VEON achieves 15.14 mIoU on Occ3D-nuScenes, and shows the capability of recognizing objects with open-vocabulary categories, meaning that our VEON is label-efficient, parameter-efficient, and precise enough.

Xirui Li, Feng Wang, Naiyan Wang et al.

In LiDAR-based 3D detection, history point clouds contain rich temporal information helpful for future prediction. In the same way, history detections should contribute to future detections. In this paper, we propose a detection enhancement method, namely FrameFusion, which improves 3D object detection results by fusing history frames. In FrameFusion, we ''forward'' history frames to the current frame and apply weighted Non-Maximum-Suppression on dense bounding boxes to obtain a fused frame with merged boxes. To ''forward'' frames, we use vehicle motion models to estimate the future pose of the bounding boxes. However, the commonly used constant velocity model fails naturally on turning vehicles, so we explore two vehicle motion models to address this issue. On Waymo Open Dataset, our FrameFusion method consistently improves the performance of various 3D detectors by about $2$ vehicle level 2 APH with negligible latency and slightly enhances the performance of the temporal fusion method MPPNet. We also conduct extensive experiments on motion model selection.

Junyi Cao, Zhichao Li, Naiyan Wang et al.

Recent studies have highlighted the promising application of NeRF in autonomous driving contexts. However, the complexity of outdoor environments, combined with the restricted viewpoints in driving scenarios, complicates the task of precisely reconstructing scene geometry. Such challenges often lead to diminished quality in reconstructions and extended durations for both training and rendering. To tackle these challenges, we present Lightning NeRF. It uses an efficient hybrid scene representation that effectively utilizes the geometry prior from LiDAR in autonomous driving scenarios. Lightning NeRF significantly improves the novel view synthesis performance of NeRF and reduces computational overheads. Through evaluations on real-world datasets, such as KITTI-360, Argoverse2, and our private dataset, we demonstrate that our approach not only exceeds the current state-of-the-art in novel view synthesis quality but also achieves a five-fold increase in training speed and a ten-fold improvement in rendering speed. Codes are available at https://github.com/VISION-SJTU/Lightning-NeRF .

Qingyu Xu, Longguang Wang, Weidong Sheng et al.

Tracking multiple tiny objects is highly challenging due to their weak appearance and limited features. Existing multi-object tracking algorithms generally focus on single-modality scenes, and overlook the complementary characteristics of tiny objects captured by multiple remote sensors. To enhance tracking performance by integrating complementary information from multiple sources, we propose a novel framework called {HGT-Track (Heterogeneous Graph Transformer based Multi-Tiny-Object Tracking)}. Specifically, we first employ a Transformer-based encoder to embed images from different modalities. Subsequently, we utilize Heterogeneous Graph Transformer to aggregate spatial and temporal information from multiple modalities to generate detection and tracking features. Additionally, we introduce a target re-detection module (ReDet) to ensure tracklet continuity by maintaining consistency across different modalities. Furthermore, this paper introduces the first benchmark VT-Tiny-MOT (Visible-Thermal Tiny Multi-Object Tracking) for RGB-T fused multiple tiny object tracking. Extensive experiments are conducted on VT-Tiny-MOT, and the results have demonstrated the effectiveness of our method. Compared to other state-of-the-art methods, our method achieves better performance in terms of MOTA (Multiple-Object Tracking Accuracy) and ID-F1 score. The code and dataset will be made available at https://github.com/xuqingyu26/HGTMT.

Weijia Zhang, Dongnan Liu, Chao Ma et al.

Monocular 3D object detection (M3OD) is a significant yet inherently challenging task in autonomous driving due to absence of explicit depth cues in a single RGB image. In this paper, we strive to boost currently underperforming monocular 3D object detectors by leveraging an abundance of unlabelled data via semi-supervised learning. Our proposed ODM3D framework entails cross-modal knowledge distillation at various levels to inject LiDAR-domain knowledge into a monocular detector during training. By identifying foreground sparsity as the main culprit behind existing methods' suboptimal training, we exploit the precise localisation information embedded in LiDAR points to enable more foreground-attentive and efficient distillation via the proposed BEV occupancy guidance mask, leading to notably improved knowledge transfer and M3OD performance. Besides, motivated by insights into why existing cross-modal GT-sampling techniques fail on our task at hand, we further design a novel cross-modal object-wise data augmentation strategy for effective RGB-LiDAR joint learning. Our method ranks 1st in both KITTI validation and test benchmarks, significantly surpassing all existing monocular methods, supervised or semi-supervised, on both BEV and 3D detection metrics.

Wenxi Li, Ruxin Zhang, Haozhe Lin et al.

The advancement of deep learning in object detection has predominantly focused on megapixel images, leaving a critical gap in the efficient processing of gigapixel images. These super high-resolution images present unique challenges due to their immense size and computational demands. To address this, we introduce 'SaccadeDet', an innovative architecture for gigapixel-level object detection, inspired by the human eye saccadic movement. The cornerstone of SaccadeDet is its ability to strategically select and process image regions, dramatically reducing computational load. This is achieved through a two-stage process: the 'saccade' stage, which identifies regions of probable interest, and the 'gaze' stage, which refines detection in these targeted areas. Our approach, evaluated on the PANDA dataset, not only achieves an 8x speed increase over the state-of-the-art methods but also demonstrates significant potential in gigapixel-level pathology analysis through its application to Whole Slide Imaging.

Jiahuan Long, Tingsong Jiang, Hanqing Liu et al.

Adversarial patches have emerged as a popular privacy-preserving approach for resisting AI-driven surveillance systems. However, their conspicuous appearance makes them difficult to deploy in real-world scenarios. In this paper, we propose a thermally activated adversarial wearable designed to ensure adaptability and effectiveness in complex real-world environments. The system integrates thermochromic dyes with flexible heating units to induce visually dynamic adversarial patterns on clothing surfaces. In its default state, the clothing appears as an ordinary black T-shirt. Upon heating via an embedded thermal unit, hidden adversarial patterns on the fabric are activated, allowing the wearer to effectively evade detection across both visible and infrared modalities. Physical experiments demonstrate that the adversarial wearable achieves rapid texture activation within 50 seconds and maintains an adversarial success rate above 80\% across diverse real-world surveillance environments. This work demonstrates a new pathway toward physically grounded, user-controllable anti-AI systems, highlighting the growing importance of proactive adversarial techniques for privacy protection in the age of ubiquitous AI surveillance.

Guoqing Wang, Pin Tang, Xiangxuan Ren et al.

Vision-Language-Action (VLA) models have recently achieved notable progress in end-to-end autonomous driving by integrating perception, reasoning, and control within a unified multimodal framework. However, they often lack explicit modeling of temporal dynamics and global world consistency, which limits their foresight and safety. In contrast, world models can simulate plausible future scenes but generally struggle to reason about or evaluate the imagined future they generate. In this work, we present VLA-World, a simple yet effective VLA world model that unifies predictive imagination with reflective reasoning to improve driving foresight. VLA-World first uses an action-derived feasible trajectory to guide the generation of the next-frame image, capturing rich spatial and temporal cues that describe how the surrounding environment evolves. The model then reasons over this self-generated future imagined frame to refine the predicted trajectory, achieving higher performance and better interpretability. To support this pipeline, we curate nuScenes-GR-20K, a generative reasoning dataset derived from nuScenes, and employ a three-stage training strategy that includes pretraining, supervised fine-tuning, and reinforcement learning. Extensive experiments demonstrate that VLA-World consistently surpasses state-of-the-art VLA and world-model baselines on both planning and future-generation benchmarks. Project page: https://vlaworld.github.io

Jingjing Jiang, Chao Ma, Xurui Song et al.

Recent advancements in multimodal large language models (MLLMs) have demonstrated exceptional performance in multimodal perception and understanding. However, leading open-source MLLMs exhibit significant limitations in complex and structured reasoning, particularly in tasks requiring deep reasoning for decision-making and problem-solving. In this work, we present Corvid, an MLLM with enhanced chain-of-thought (CoT) reasoning capabilities. Architecturally, Corvid incorporates a hybrid vision encoder for informative visual representation and a meticulously designed connector (GateMixer) to facilitate cross-modal alignment. To enhance Corvid's CoT reasoning capabilities, we introduce MCoT-Instruct-287K, a high-quality multimodal CoT instruction-following dataset, refined and standardized from diverse public reasoning sources. Leveraging this dataset, we fine-tune Corvid with a two-stage CoT-formatted training approach to progressively enhance its step-by-step reasoning abilities. Furthermore, we propose an effective inference-time scaling strategy that enables Corvid to mitigate over-reasoning and under-reasoning through self-verification. Extensive experiments demonstrate that Corvid outperforms existing o1-like MLLMs and state-of-the-art MLLMs with similar parameter scales, with notable strengths in mathematical reasoning and science problem-solving. Project page: https://mm-vl.github.io/corvid.

Weijia Zhang, Yuehao Liu, Shanyan Guan et al.

Classifier-Free Guidance (CFG) is a widely used inference-time technique to boost the image quality of diffusion models. Yet, its reliance on text conditions prevents its use in unconditional generation. We propose a simple method to enable CFG-like guidance for both conditional and unconditional generation. The key idea is to generate a perturbed prediction via simple token swap operations, and use the direction between it and the clean prediction to steer sampling towards higher-fidelity distributions. In practice, we swap pairs of most semantically dissimilar token latents in either spatial or channel dimensions. Unlike existing methods that apply perturbation in a global or less constrained manner, our approach selectively exchanges and recomposes token latents, allowing finer control over perturbation and its influence on generated samples. Experiments on MS-COCO 2014, MS-COCO 2017, and ImageNet datasets demonstrate that the proposed Self-Swap Guidance (SSG), when applied to popular diffusion models, outperforms previous condition-free methods in image fidelity and prompt alignment under different set-ups. Its fine-grained perturbation granularity also improves robustness, reducing side-effects across a wider range of perturbation strengths. Overall, SSG extends CFG to a broader scope of applications including both conditional and unconditional generation, and can be readily inserted into any diffusion model as a plug-in to gain immediate improvements.

Zepeng Wang, Chao Ma, Linjiang Zhou et al.

Different classes of safe reinforcement learning algorithms have shown satisfactory performance in various types of safety requirement scenarios. However, the existing methods mainly address one or several classes of specific safety requirement scenario problems and cannot be applied to arbitrary safety requirement scenarios. In addition, the optimization objectives of existing reinforcement learning algorithms are misaligned with the task requirements. Based on the need to address these issues, we propose $\mathrm{E^{2}CFD}$, an effective and efficient cost function design framework. $\mathrm{E^{2}CFD}$ leverages the capabilities of a large language model (LLM) to comprehend various safety scenarios and generate corresponding cost functions. It incorporates the \textit{fast performance evaluation (FPE)} method to facilitate rapid and iterative updates to the generated cost function. Through this iterative process, $\mathrm{E^{2}CFD}$ aims to obtain the most suitable cost function for policy training, tailored to the specific tasks within the safety scenario. Experiments have proven that the performance of policies trained using this framework is superior to traditional safe reinforcement learning algorithms and policies trained with carefully designed cost functions.

Jingjing Jiang, Chongjie Si, Jun Luo et al.

This paper presents a pioneering exploration of reinforcement learning (RL) via group relative policy optimization for unified multimodal large language models (ULMs), aimed at simultaneously reinforcing generation and understanding capabilities. Through systematic pilot studies, we uncover the significant potential of ULMs to enable the synergistic co-evolution of dual capabilities within a shared policy optimization framework. Building on this insight, we introduce CoRL, a co-reinforcement learning framework comprising a unified RL stage for joint optimization and a refined RL stage for task-specific enhancement. With the proposed CoRL, our resulting model, ULM-R1, achieves average improvements of 7% on three text-to-image generation datasets and 23% on nine multimodal understanding benchmarks. These results demonstrate the effectiveness of CoRL and highlight the substantial benefit of reinforcement learning in facilitating cross-task synergy and optimization for ULMs. Code is available at https://github.com/mm-vl/ULM-R1.

Meyer Scetbon, Joel Jennings, Agrin Hilmkil et al.

We propose a novel formalism for describing Structural Causal Models (SCMs) as fixed-point problems on causally ordered variables, eliminating the need for Directed Acyclic Graphs (DAGs), and establish the weakest known conditions for their unique recovery given the topological ordering (TO). Based on this, we design a two-stage causal generative model that first infers in a zero-shot manner a valid TO from observations, and then learns the generative SCM on the ordered variables. To infer TOs, we propose to amortize the learning of TOs on synthetically generated datasets by sequentially predicting the leaves of graphs seen during training. To learn SCMs, we design a transformer-based architecture that exploits a new attention mechanism enabling the modeling of causal structures, and show that this parameterization is consistent with our formalism. Finally, we conduct an extensive evaluation of each method individually, and show that when combined, our model outperforms various baselines on generated out-of-distribution problems. The code is available on \href{https://github.com/microsoft/causica/tree/main/research_experiments/fip}{Github}.

Yong Wu, YanZhao Zheng, TianZe Xu et al.

LLM-based agents show strong potential for long-horizon reasoning, yet their context size is limited by deployment factors (e.g., memory, latency, and cost), yielding a constrained context budget. As interaction histories grow, this induces a trade-off between retaining past information and staying within the context limit. To address this challenge, we propose Budget-Aware Context Management (BACM), which formulates context management as a sequential decision problem with a context budget constraint. It enables agents to assess the available budget before incorporating new observations and decide when and how much of the interaction history to compress. We further develop BACM-RL, an end-to-end curriculum-based reinforcement learning approach that learns compression strategies under varying context budgets. Experiments on compositional multi-objective QA and long-horizon web browsing benchmarks show that BACM-RL consistently outperforms prior methods across model scales and task complexities, achieving over $1.6\times$ gains over strong baselines in high-complexity settings, while maintaining strong advantages as budgets shrink, where most methods exhibit a downward performance trend.

Qinglong Cao, Yuntian Chen, Chao Ma et al.

Multimodal large language models (MLLMs) have shown remarkable capabilities in multimodal perception and understanding tasks. However, their effectiveness in specialized domains, such as remote sensing and medical imaging, remains limited. A natural approach to domain adaptation is to inject domain knowledge through textual instructions, prompts, or auxiliary captions. Surprisingly, we find that such input-level domain knowledge injection yields little to no improvement on scientific multimodal tasks, even when the domain knowledge is explicitly provided. This observation suggests that current MLLMs fail to internalize domain-specific priors through language alone, and that domain knowledge must be integrated at the optimization level. Motivated by this insight, we propose a reinforcement fine-tuning framework that incorporates domain knowledge directly into the learning objective. Instead of treating domain knowledge as descriptive information, we encode it as domain-informed constraints and reward signals, shaping the model's behavior in the output space. Extensive experiments across multiple datasets in remote sensing and medical domains consistently demonstrate good performance gains, achieving state-of-the-art results on multimodal domain tasks. Our results highlight the necessity of optimization-level domain knowledge integration and reveal a fundamental limitation of textual domain conditioning in current MLLMs.

Qinglong Cao, Yuntian Chen, Chao Ma et al.

Recent studies have explored the combination of multiple LoRAs to simultaneously generate user-specified subjects and styles. However, most existing approaches fuse LoRA weights using static statistical heuristics that deviate from LoRA's original purpose of learning adaptive feature adjustments and ignore the randomness of sampled inputs. To address this, we propose a dynamic training-free fusion framework that operates throughout the generation process. During the forward pass, at each LoRA-applied layer, we dynamically compute the KL divergence between the base model's original features and those produced by subject and style LoRAs, respectively, and adaptively select the most appropriate weights for fusion. In the reverse denoising stage, we further refine the generation trajectory by dynamically applying gradient-based corrections derived from objective metrics such as CLIP and DINO scores, providing continuous semantic and stylistic guidance. By integrating these two complementary mechanisms-feature-level selection and metric-guided latent adjustment-across the entire diffusion timeline, our method dynamically achieves coherent subject-style synthesis without any retraining. Extensive experiments across diverse subject-style combinations demonstrate that our approach consistently outperforms state-of-the-art LoRA fusion methods both qualitatively and quantitatively.

Yuehao Liu, Shanyan Guan, Weijia Zhang et al.

Continual learning in multimodal large language models (MLLMs) aims to sequentially acquire knowledge while mitigating catastrophic forgetting, yet existing methods face inherent limitations: architecture-based approaches incur additional computational overhead and often generalize poorly to new tasks, rehearsal-based methods rely on storing historical data, raising privacy and storage concerns, and conventional regularization-based strategies alone are insufficient to fully prevent parameter interference. We propose Octopus, a two-stage continual learning framework based on History-Free Gradient Orthogonalization (HiFGO), which enforces gradient-level orthogonality without historical task data. Our proposed two-stage finetuning strategy decouples task adaptation from regularization, achieving a principled balance between plasticity and stability. Experiments on UCIT show that Octopus establishes state-of-the-art performance, surpassing prior SOTA by 2.14% and 6.82% in terms of Avg and Last.

Yuehao Liu, Weijia Zhang, Xuanming Shang et al.

State-of-the-art diffusion models often rely on parameter-efficient fine-tuning to perform specialized image editing tasks. However, real-world applications require continual adaptation to new tasks while preserving previously learned knowledge. Despite the practical necessity, continual learning for image editing remains largely underexplored. We propose ACE-LoRA, a dynamic regularization framework for continual image editing that effectively mitigates catastrophic forgetting. ACE-LoRA leverages Adaptive Orthogonal Decoupling to identify and orthogonalize task interference, and introduces a Rank-Invariant Historical Information Compression strategy to address scalability issues in continual updates. To facilitate continual learning in image editing and provide a standardized evaluation protocol, we introduce CIE-Bench, the first comprehensive benchmark in this domain. CIE-Bench encompasses diverse and practically relevant image editing scenarios with a balanced level of difficulty to effectively expose limitations of existing models while remaining compatible with parameter-efficient fine-tuning. Extensive experiments demonstrate that our method consistently outperforms existing baselines in terms of instruction fidelity, visual realism, and robustness to forgetting, establishing a strong foundation for continual learning in image editing.

Furong Peng, Kang Liu, Xuan Lu et al.

Graph Convolutional Neural Network (GCN), a widely adopted method for analyzing relational data, enhances node discriminability through the aggregation of neighboring information. Usually, stacking multiple layers can improve the performance of GCN by leveraging information from high-order neighbors. However, the increase of the network depth will induce the over-smoothing problem, which can be attributed to the quality and quantity of neighbors changing: (a) neighbor quality, node's neighbors become overlapping in high order, leading to aggregated information becoming indistinguishable, (b) neighbor quantity, the exponentially growing aggregated neighbors submerges the node's initial feature by recursively aggregating operations. Current solutions mainly focus on one of the above causes and seldom consider both at once. Aiming at tackling both causes of over-smoothing in one shot, we introduce a simple Two-Sided Constraint (TSC) for GCNs, comprising two straightforward yet potent techniques: random masking and contrastive constraint. The random masking acts on the representation matrix's columns to regulate the degree of information aggregation from neighbors, thus preventing the convergence of node representations. Meanwhile, the contrastive constraint, applied to the representation matrix's rows, enhances the discriminability of the nodes. Designed as a plug-in module, TSC can be easily coupled with GCN or SGC architectures. Experimental analyses on diverse real-world graph datasets verify that our approach markedly reduces the convergence of node's representation and the performance degradation in deeper GCN.

Ziming Guo, Chao Ma, Yinggang Sun et al.

Recent advancements in large language models (LLMs) have significantly advanced text-to-SQL systems. However, most LLM-based methods often narrowly focus on SQL generation, neglecting the complexities of real-world conversational queries. This oversight can lead to unreliable responses, particularly for ambiguous questions that cannot be directly addressed with SQL. To bridge this gap, we propose MMSQL, a comprehensive test suite designed to evaluate the question classification and SQL generation capabilities of LLMs by simulating real-world scenarios with diverse question types and multi-turn Q&A interactions. Using MMSQL, we assessed the performance of popular LLMs, including both open-source and closed-source models, and identified key factors impacting their performance in such scenarios. Moreover, we introduce an LLM-based multi-agent framework that employs specialized agents to identify question types and determine appropriate answering strategies. Our experiments demonstrate that this approach significantly enhances the model's ability to navigate the complexities of conversational dynamics, effectively handling the diverse and complex nature of user queries. Our dataset and code are publicly available at https://mcxiaoxiao.github.io/MMSQL.

Chao Ma, Cheng Zhang

The gold standard for causal model evaluation involves comparing model predictions with true effects estimated from randomized controlled trials (RCT). However, RCTs are not always feasible or ethical to perform. In contrast, conditionally randomized experiments based on inverse probability weighting (IPW) offer a more realistic approach but may suffer from high estimation variance. To tackle this challenge and enhance causal model evaluation in real-world conditional randomization settings, we introduce a novel low-variance estimator for causal error, dubbed as the pairs estimator. By applying the same IPW estimator to both the model and true experimental effects, our estimator effectively cancels out the variance due to IPW and achieves a smaller asymptotic variance. Empirical studies demonstrate the improved of our estimator, highlighting its potential on achieving near-RCT performance. Our method offers a simple yet powerful solution to evaluate causal inference models in conditional randomization settings without complicated modification of the IPW estimator itself, paving the way for more robust and reliable model assessments.

Xiangxuan Ren, Zhongdao Wang, Liping Hou et al.

Multimodal large language models (MLLMs) have made significant advancements in vision understanding and reasoning. However, the autoregressive Transformer architecture used by MLLMs requries tokenization on input images, which limits their ability to accurately ground objects within the 2D image space. This raises an important question: how can sequential language tokens be improved to better ground objects in 2D spatial space for MLLMs? To address this, we present a spatial representation method for grounding objects, namely GETok, that integrates a specialized vocabulary of learnable tokens into MLLMs. GETok first uses grid tokens to partition the image plane into structured spatial anchors, and then exploits offset tokens to enable precise and iterative refinement of localization predictions. By embedding spatial relationships directly into tokens, GETok significantly advances MLLMs in native 2D space reasoning without modifying the autoregressive architecture. Extensive experiments demonstrate that GETok achieves superior performance over the state-of-the-art methods across various referring tasks in both supervised fine-tuning and reinforcement learning settings.

Huanhuan Yuan, Yang Ping, Zhengqin Xu et al.

The rapid advancement of generative artificial intelligence has enabled the creation of highly realistic fake facial images, posing serious threats to personal privacy and the integrity of online information. Existing deepfake detection methods often rely on handcrafted forensic cues and complex architectures, achieving strong performance in intra-domain settings but suffering significant degradation when confronted with unseen forgery patterns. In this paper, we propose GenDF, a simple yet effective framework that transfers a powerful large-scale vision model to the deepfake detection task with a compact and neat network design. GenDF incorporates deepfake-specific representation learning to capture discriminative patterns between real and fake facial images, feature space redistribution to mitigate distribution mismatch, and a classification-invariant feature augmentation strategy to enhance generalization without introducing additional trainable parameters. Extensive experiments demonstrate that GenDF achieves state-of-the-art generalization performance in cross-domain and cross-manipulation settings while requiring only 0.28M trainable parameters, validating the effectiveness and efficiency of the proposed framework.

Xiangxuan Ren, Zhongdao Wang, Pin Tang et al.

3D object detection is fundamental for safe and robust intelligent transportation systems. Current multi-modal 3D object detectors often rely on complex architectures and training strategies to achieve higher detection accuracy. However, these methods heavily rely on the LiDAR sensor so that they suffer from large performance drops when LiDAR is absent, which compromises the robustness and safety of autonomous systems in practical scenarios. Moreover, existing multi-modal detectors face difficulties in deployment on diverse hardware platforms, such as NPUs and FPGAs, due to their reliance on 3D sparse convolution operators, which are primarily optimized for NVIDIA GPUs. To address these challenges, we reconsider the role of LiDAR in the camera-LiDAR fusion paradigm and introduce a novel multi-modal 3D detector, LiteFusion. Instead of treating LiDAR point clouds as an independent modality with a separate feature extraction backbone, LiteFusion utilizes LiDAR data as a complementary source of geometric information to enhance camera-based detection. This straightforward approach completely eliminates the reliance on a 3D backbone, making the method highly deployment-friendly. Specifically, LiteFusion integrates complementary features from LiDAR points into image features within a quaternion space, where the orthogonal constraints are well-preserved during network training. This helps model domain-specific relations across modalities, yielding a compact cross-modal embedding. Experiments on the nuScenes dataset show that LiteFusion improves the baseline vision-based detector by +20.4% mAP and +19.7% NDS with a minimal increase in parameters (1.1%) without using dedicated LiDAR encoders. Notably, even in the absence of LiDAR input, LiteFusion maintains strong results , highlighting its favorable robustness and effectiveness across diverse fusion paradigms and deployment scenarios.