Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks562 from public data
- Deep Learning for Generic Object Detection: A Survey2,921
A comprehensive survey of the recent achievements in this field brought about by deep learning techniques, covering many aspects of generic object detection: detection frameworks, object feature representation, object proposal generation, context modeling, training strategies, and evaluation metrics.
- Libra R-CNN: Towards Balanced Learning for Object Detection1,762
Libra R-CNN is proposed, a simple but effective framework towards balanced learning for object detection that integrates three novel components: IoU-balanced sampling, balanced feature pyramid, and balanced L1 loss, respectively for reducing the imbalance at sample, feature, and objective level.
- Hybrid Task Cascade for Instance Segmentation1,565
This work proposes a new framework, Hybrid Task Cascade (HTC), which differs in two important aspects: (1) instead of performing cascaded refinement on these two tasks separately, it interweaves them for a joint multi-stage processing; (2) it adopts a fully convolutional branch to provide spatial context, which can help distinguishing hard foreground from cluttered background.
- WorldSimBench: Towards Video Generation Models as World Simulators1,338
This work classify the functionalities of predictive models into a hierarchy and takes the first step in evaluating World Simulators by proposing a dual evaluation framework called WorldSimBench, encompassing human preference assessments from the visual perspective and action-level evaluations in embodied tasks.
- Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition1,222
A simple method to disentangle multi-scale graph convolutions and a unified spatial-temporal graph convolutional operator named G3D are presented and a powerful feature extractor named MS-G3D is developed based on which the model outperforms previous state-of-the-art methods on three large-scale datasets.
- Point Transformer V3: Simpler, Faster, Stronger1,171
This paper presents Point Transformer V3 (PTv3), which prioritizes simplicity and efficiency over the accuracy of certain mechanisms that are minor to the over-all performance after scaling, such as replacing the precise neighbor search by KNN with an efficient serialized neighbor mapping of point clouds organized with specific patterns.
- 1,120
- Learning Deep Feature Representations with Domain Guided Dropout for Person Re-identification997
This work presents a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks with CNNs and proposes a Domain Guided Dropout algorithm to improve the feature learning procedure.
- 995
- 990
- DVC: An End-To-End Deep Video Compression Framework895
This paper proposes the first end-to-end video compression deep model that jointly optimizes all the components for video compression, and shows that the proposed approach can outperform the widely used video coding standard H.264 in terms of PSNR and be even on par with the latest standard MS-SSIM.
- Multi-context Attention for Human Pose Estimation736
This paper proposes to incorporate convolutional neural networks with a multi-context attention mechanism into an end-to-end framework for human pose estimation and designs novel Hourglass Residual Units (HRUs) to increase the receptive field of the network.
- 675
- 668
- 653
- PAD-Net: Multi-tasks Guided Prediction-and-Distillation Network for Simultaneous Depth Estimation and Scene Parsing582
This paper proposes a novel multi-task guided prediction-and-distillation network (PAD-Net), which first predicts a set of intermediate auxiliary tasks ranging from low level to high level, and then the predictions from these intermediate Auxiliary tasks are utilized as multi-modal input via the authors' proposed multi- modal distillation modules for the final tasks.
- Equalization Loss for Long-Tailed Object Recognition577
This work proposes a simple but effective loss, named equalization loss, to tackle the problem of long-tailed rare categories by simply ignoring those gradients for rare categories, and wins the 1st place in the LVIS Challenge 2019.
- 568
- Scene Graph Generation from Objects, Phrases and Region Captions560
This work proposes a novel neural network model, termed as Multi-level Scene Description Network (denoted as MSDN), to solve the three vision tasks jointly in an end-to-end manner and shows the joint learning across three tasks with the proposed method can bring mutual improvements over previous models.
- Scene Graph Generation from Objects, Phrases and Caption Regions.560
This work proposes a novel neural network model, termed as Multi-level Scene Description Network (denoted as MSDN), to solve the three vision tasks jointly in an end-to-end manner and shows the joint learning across three tasks with the proposed method can bring mutual improvements over previous models.
- SR-LSTM: State Refinement for LSTM Towards Pedestrian Trajectory Prediction556
A data-driven state refinement module for LSTM network (SR-LSTM), which activates the utilization of the current intention of neighbors, and jointly and iteratively refines the current states of all participants in the crowd through a message passing mechanism is proposed.
- T-CNN: Tubelets With Convolutional Neural Networks for Object Detection From Videos548
A deep learning framework that incorporates temporal and contextual information from tubelets obtained in videos, which dramatically improves the baseline performance of existing still-image detection frameworks when they are applied to videos is proposed, called T-CNN.
- Learning Feature Pyramids for Human Pose Estimation536
This work designs a Pyramid Residual Module (PRMs) to enhance the invariance in scales of DCNNs and provides theoretic derivation to extend the current weight initialization scheme to multi-branch network structures.
- Attention-Aware Compositional Network for Person Re-identification502
This work introduces a novel framework called Attention-Aware Compositional Network (AACN) for person ReID, which consists of two main components: Pose-guided Part Attention (PPA) and Attention-aware Feature Composition (AFC).
- 502
- 499
- The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models468
This paper investigates entropy dynamics both theoretically and empirically and proposes two simple yet effective techniques, namely Clip-Cov and KL-Cov, which clip and apply KL penalty to tokens with high covariances respectively, helping policy escape entropy collapse and achieve better downstream performance.
- Multi-scale Continuous CRFs as Sequential Deep Networks for Monocular Depth Estimation466
This paper addresses the problem of depth estimation from a single still image by designing a novel CNN implementation of mean-field updates for continuous CRFs and demonstrates the effectiveness of the proposed approach and establishes new state of the art results on publicly available datasets.
- DeepID-Net: Deformable deep convolutional neural networks for object detection452
The proposed approach improves the mean averaged precision obtained by RCNN, which was the state-of-the-art, from 31% to 50.3% on the ILSVRC2014 detection test set.
- RS-Mamba for Large Remote Sensing Image Dense Prediction415
This work proposes the remote sensing Mamba (RSM) for dense prediction tasks in large VHR remote sensing images, specifically designed to capture the global context of remote sensing images with linear complexity, facilitating the effective processing of large VHR images.
- 3D Human Pose Estimation in the Wild by Adversarial Learning406
An adversarial learning framework is proposed, which distills the 3D human pose structures learned from the fully annotated dataset to in-the-wild images with only 2D pose annotations and designs a geometric descriptor, which computes the pairwise relative locations and distances between body joints, as a new information source for the discriminator.
- PyMAF: 3D Human Pose and Shape Regression with Pyramidal Mesh Alignment Feedback Loop402
A Pyramidal Mesh Alignment Feedback (PyMAF) loop to leverage a feature pyramid and rectify the predicted parameters explicitly based on the mesh-image alignment status in the authors' deep regressor is proposed.
- Learning Spatial Regularization with Image-Level Supervisions for Multi-label Image Classification393
Analysis of the learned SRN model demonstrates that it can effectively capture both semantic and spatial relations of labels for improving classification performance, and significantly outperforms state-of-the-arts and has strong generalization capability.
- Online Multi-object Tracking Using CNN-Based Single Object Tracker with Spatial-Temporal Attention Mechanism391
A CNN-based framework for online MOT that utilizes the merits of single object trackers in adapting appearance models and searching for target in the next frame and introduces spatial-temporal attention mechanism (STAM) to handle the drift caused by occlusion and interaction among targets.
- 383
- Whole-Body Human Pose Estimation in the Wild376
COCO-WholeBody is introduced which extends COCO dataset with whole-body annotations and is the first benchmark that has manual annotations on the entire human body, including 133 dense landmarks with 68 on the face, 42 on hands and 23 on the body and feet.
- GS3D: An Efficient 3D Object Detection Framework for Autonomous Driving361
This work leverages the off-the-shelf 2D object detector to efficiently obtain a coarse cuboid for each predicted 2D box and explores the 3D structure information of the object by employing the visual features of visible surfaces.
- The operational medium-range deterministic weather forecasting can be extended beyond a 10-day lead time358
A new deep learning-based global medium-range weather forecasting model outperforms existing machine learning models and the European Centre for Medium-Range Weather Forecasts model, producing accurate global weather forecasts beyond ten days.
- DiffBIR: Towards Blind Image Restoration with Generative Diffusion Prior353
DiffBIR decouples blind image restoration problem into two stages: degradation removal and information regeneration, and proposes IRControlNet, a region-adaptive restoration guidance that can modify the denoising process during inference without model re-training, allowing users to balance realness and fidelity through a tunable guidance scale.
- 351
- Accurate Monocular 3D Object Detection via Color-Embedded 3D Reconstruction for Autonomous Driving347
This paper proposes a monocular 3D object detection framework in the domain of autonomous driving, and proposes a multi-modal feature fusion module to embed the complementary RGB cue into the generated point clouds representation.
- Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition343
A novel compact motion representation for video action recognition, named Optical Flow guided Feature (OFF), which enables the network to distill temporal information through a fast and robust approach and is complementary to other motion modalities such as optical flow.
- Style Aggregated Network for Facial Landmark Detection339
This work proposes a style-aggregated approach to deal with the large intrinsic variance of image styles for facial landmark detection and demonstrates to perform well when compared with state-of-the-art algorithms on benchmark datasets AFLW and 300-W.
- Quality Aware Network for Set to Set Recognition339
Analysis on gradient spread of this mechanism indicates that the quality learned by the network is beneficial to set-to-set recognition and simplifies the distribution that the network needs to fit.
- GradNet: Gradient-Guided Network for Visual Object Tracking334
A novel gradient-guided network to exploit the discriminative information in gradients and update the template in the siamese network through feed-forward and backward operations and a template generalization training method is proposed to better use gradient information and avoid overfitting.
- MT-Bench-101: A Fine-Grained Benchmark for Evaluating Large Language Models in Multi-Turn Dialogues330
This work introduces MT-Bench-101, specifically designed to evaluate the fine-grained abilities of LLMs in multi-turn dialogues, and constructs a three-tier hierarchical ability taxonomy comprising 4208 turns across 1388 multi-turn dialogues in 13 distinct tasks.
- Multi-class Token Transformer for Weakly Supervised Semantic Segmentation321
A Multi-class Token Transformer, termed as MCTformer, which uses multiple class tokens to learn interactions between the class tokens and the patch tokens, which can successfully produce class-discriminative object localization maps from the class-to-patch attentions corresponding to different class tokens.
- 320
- Geometry Uncertainty Projection Network for Monocular 3D Object Detection316
A GUP Net is proposed to tackle the error amplification problem at both inference and training stages and can infer more reliable object depth than existing methods and outperforms the state-of-the-art image-based monocular 3D detectors.
- Delving into Localization Errors for Monocular 3D Object Detection309
This work quantifies the impact introduced by each sub-task and found the ‘localization error’ is the vital factor in restricting monocular 3D detection, and investigates the underlying reasons behind localization errors.
- Crowd Counting With Deep Structured Scale Integration Network289
A novel Deep Structured Scale Integration Network (DSSINet) for crowd counting is proposed, which addresses the scale variation of people by using structured feature representation learning and hierarchically structured loss function optimization.
- Factorizable Net: An Efficient Subgraph-Based Framework for Scene Graph Generation285
A subgraph-based connection graph is proposed to concisely represent the scene graph during the inference to improve the efficiency of scene graph generation and outperforms the state-of-the-art method in both accuracy and speed.
- 279
- Contextualized Spatial–Temporal Network for Taxi Origin-Destination Demand Prediction279
A novel contextualized spatial–temporal network (CSTN) for taxi origin-destination demand prediction, which consists of three components for the modeling of local spatial context (LSC), temporal evolution context (TEC), and global correlation context (GCC), respectively.
- Structured Feature Learning for Pose Estimation269
A bi-directional tree structured model is proposed, so that the feature channels at a body joint can well receive information from other joints, and the proposed framework improves feature learning substantially.
- 266
- 265
- CLIP2Point: Transfer CLIP to Point Cloud Classification with Image-Depth Pre-Training260
Experimental results show that CLIP2Point is effective in transferring CLIP knowledge to 3D vision, and a novel Gated Dual-Path Adapter (GDPA) allows the ensemble of CLIP and CLIP2Point, tuning pre-training knowledge to downstream tasks in an efficient adaptation.
- 260
- Box-Driven Class-Wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation257
This paper introduces a box-driven class-wise masking model (BCM) to remove irrelevant regions of each class and proposes a filling rate guided adaptive loss (FR-Loss) to help the model ignore the wrongly labeled pixels in proposals.
- 243
- Rethinking Pseudo-LiDAR Representation239
An in-depth investigation is performed and it is observed that the efficacy of pseudo-LiDAR representation comes from the coordinate transformation, instead of data representation itself, which means existing 2D CNN designs can be easily utilized for extracting deep features from input data and boosting 3D detection performance.
- Improving Description-Based Person Re-Identification by Multi-Granularity Image-Text Alignments238
A Multi-granularity Image-text Alignments (MIA) model is proposed to alleviate the cross-modal fine-grained problem for better similarity evaluation in description-based person Re-id and obtains the state-of-the-art performance on the CUHK-PEDES dataset.
- Learning Cross-Modal Deep Representations for Robust Pedestrian Detection237
The proposed approach outperforms the state-of-the-art on the challenging KAIST multispectral pedestrian dataset and it is competitive with previous methods on the popular Caltech dataset.
- Person Search via a Mask-Guided Two-Stream CNN Model230
A simple yet effective re-ID method is proposed, which models foreground person and original image patches individually, and obtains enriched representations from two separate CNN streams, and achieves mAP of 83.0% on the standard person search benchmark datasets.
- 229
- Geometry-enhanced pretraining on interatomic potentials215
A geometric structure learning framework that leverages self-supervised learning pretraining to enhance existing machine learning based interatomic potential models at a negligible additional computational cost and enhances MLIPs and advances the simulations of molecular systems is proposed.
- Meta-Transformer: A Unified Framework for Multimodal Learning214
Meta-Transformer is the first framework to perform unified learning across 12 modalities with unpaired data and indicates a promising future for developing unified multimodal intelligence with transformers.
- Factors in Finetuning Deep Model for Object Detection with Long-Tail Distribution214
This paper investigates many factors that influence the performance in finetuning for object detection and proposes a hierarchical feature learning scheme that cluster objects into visually similar class groups and learn deep representations for these groups separately.
- Crowd Counting using Deep Recurrent Spatial-Aware Network198
A unified neural network framework, named Deep Recurrent Spatial-Aware Network, which adaptively addresses the two issues in a learnable spatial transform module with a region-wise refinement process, and achieves an improvement on the largest dataset WorldExpo’10 and the most challenging dataset UCF-CC.
- 196
- Not All Tokens Are Equal: Human-centric Visual Analysis via Token Clustering Transformer192
A novel Vision Transformer, called Token Clustering Transformer (TCFormer), which merges tokens by progressive clustering, where the tokens can be merged from different locations with flexible shapes and sizes, which is beneficial to capturing detailed information.
- Visual Question Generation as Dual Task of Visual Question Answering181
This paper proposes an end-to-end unified model, the Invertible Question Answering Network (iQAN), to introduce question generation as a dual task of question answering to improve the VQA performance and shows that the proposed dual training framework can consistently improve model performances of many popular V QA architectures.
- 3D Object Detection From Images for Autonomous Driving: A Survey180
This paper provides the first comprehensive survey of this novel and continuously growing research field, summarizing the most commonly used pipelines for image-based 3D detection and deeply analyzing each of their components and proposing two new taxonomies to organize the state-of-the-art methods into different categories.
- 177
- Content Adaptive and Error Propagation Aware Deep Video Compression176
This work proposes a content adaptive and error propagation aware video compression system that outperforms the state-of-the-art learning based video codecs on benchmark datasets without increasing the model size or decreasing the decoding speed.
- Bilateral Reference for High-Resolution Dichotomous Image Segmentation175
This work introduces a novel bilateral reference framework (BiRefNet) for high-resolution dichotomous image segmentation (DIS) that exhibits remarkable performance, outperforming task-specific cutting-edge methods across all benchmarks.
- Leveraging Auxiliary Tasks with Affinity Learning for Weakly Supervised Semantic Segmentation175
This work proposes a novel weakly supervised multi-task framework termed as AuxSegNet, to leverage saliency detection and multi-label image classification as auxiliary tasks to improve the primary task of semantic segmentation using only image-level ground-truth labels.
- Cap4Video: What Can Auxiliary Captions Do for Text-Video Retrieval?171
Cap4Video is a novel approach to text-video retrieval, where associated captions from videos are generated using zero-shot video captioning with knowledge from web-scale pre-trained models to complement the original Query-Video matching branch for text- video retrieval.
- 168
- NDC-Scene: Boost Monocular 3D Semantic Scene Completion in Normalized Device Coordinates Space163
A novel Normalized Device Coordinates scene completion network (NDC-Scene) is devised that directly extends the 2D feature map to a Normalized device coordinates space, rather than to the world space directly, through progressive restoration of the dimension of depth with deconvolution operations.
- OASIS: Open Agent Social Interaction Simulations with One Million Agents161
OASIS is a generalizable and scalable social media simulator that replicates various social phenomena, including information spreading, group polarization, and herd effects across X and Reddit platforms, and demonstrates OASIS's potential as a powerful tool for studying complex systems in digital environments.
- 159
- Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling158
Findings show the significance of adapting TTS strategies to the specific characteristics of each task and model and indicate that TTS is a promising approach for enhancing the reasoning abilities of LLMs.
- $\beta$-DARTS: Beta-Decay Regularization for Differentiable Architecture Search155
A simple-but-efficient regularization method, termed as Beta-Decay, is proposed to regularize the DARTS-based NAS searching process and can impose constraints to keep the value and variance of activated architecture parameters from too large.
- Revisiting Classifier: Transferring Vision-Language Models for Video Recognition154
This paper revise the role of the linear classifier and replace the classifier with the different knowledge from pre-trained model, and utilizes the well-pretrained language model to generate good semantic target for efficient transferring learning.
- Improving Deep Video Compression by Resolution-Adaptive Flow Coding149
A new framework called Resolution-adaptive Flow Coding (RaFC) is proposed to effectively compress the flow maps globally and locally, in which multi-resolution representations are used for both the input flow maps and the output motion features of the MV encoder.
- Differentiable Hierarchical Graph Grouping for Multi-person Pose Estimation147
This paper proposes a novel differentiable Hierarchical Graph Grouping (HGG) method, which takes human keypoint candidates as graph nodes and clusters keypoints in a multi-layer graph neural network model and improves the performance of bottom-up pose estimation methods.
- Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization144
Multi-Objective Direct Preference Optimization (MODPO) is presented, an RL-free extension of Direct Preference Optimization (DPO) for multiple alignment objectives that folds language modeling directly into reward modeling, training language models as implicit collective reward models that combine all objectives with specific weights.
- Crafting GBD-Net for Object Detection144
This paper proposes a gated bi-directional CNN (GBD-Net) to pass messages among features from different support regions during both feature learning and feature extraction, and shows the effectiveness of different network structures and existing techniques for object detection.
- 143
- ChemLLM: A Chemical Large Language Model139
ChemLLM, a comprehensive framework that features the first LLM dedicated to chemistry, also includes ChemData, a dataset specifically designed for instruction tuning, and ChemBench, a robust benchmark covering nine essential chemistry tasks.
- EcoNAS: Finding Proxies for Economical Neural Architecture Search137
A reliable proxy is presented and a hierarchical proxy strategy is formulated that spends more computations on candidate networks that are potentially more accurate, while discards unpromising ones in early stage with a fast proxy to lead to an economical evolutionary-based NAS (EcoNAS), which achieves an impressive 400×search time reduction.
- Learning Deep Structured Multi-Scale Features using Attention-Gated CRFs for Contour Prediction134
A hierarchical deep model is introduced which produces more rich and complementary representations and the novel Attention-Gated Conditional Random Fields (AG-CRFs) are proposed to refine and robustly fuse the representations learned at different scales.
- 132
- Gated Bi-directional CNN for Object Detection131
A novel gated bi-directional CNN (GBD-Net) to pass messages between features from different support regions during both feature learning and feature extraction to validate the existence of each other by learning their nonlinear relationships.
- 130
- 130
- 128
- 126
- Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models121
A novel framework called BIKE, which utilizes the cross-modal bridge to explore bidirectional knowledge and presents a Temporal Concept Spotting mechanism that uses the Text-to-Video expertise to capture temporal saliency in a parameter-free manner, leading to enhanced video representation.
- Part-Level Graph Convolutional Network for Skeleton-Based Action Recognition120
A novel Part-Level Graph Convolutional Network (PL-GCN) is proposed to capture part-level information of skeletons, which aims at learning high-level relations between body parts while the PA block aims at highlighting the important body parts in the action.
- 120
- TransVG++: End-to-End Visual Grounding With Language Conditioned Vision Transformer116
This work proposes TransVG, which establishes multi-modal correspondences by Transformers and localizes referred regions by directly regressing box coordinates, and introduces TransVG++, which upgrades the framework to a purely Transformer-based one by leveraging Vision Transformer (ViT) for vision feature encoding.
- Deep Learning of Scene-Specific Classifier for Pedestrian Detection116
A deep model is proposed to automatically learn scene-specific features and visual patterns in static video surveillance without any manual labels from the target scene to bridge the appearance gap.
- 3D Human Mesh Regression With Dense Correspondence114
A model-free 3D human mesh estimation framework, named DecoMR, is proposed, which explicitly establishes the dense correspondence between the mesh and the local image features in the UV space (i.e., IUV image), and a novel UV map that maintains most of the neighboring relations on the original mesh surface.
- Quantization Mimic: Towards Very Tiny CNN for Object Detection114
A simple and general framework for training very tiny CNNs (e.g. VGG with the number of channels reduced to \(\frac{1}{32}\)) for object detection and the Quantization Mimic algorithm, which first quantizes the large network, then mimic a quantized small network.
- Monocular Depth Estimation Using Multi-Scale Continuous CRFs as Sequential Deep Networks112
This paper addresses the problem of monocular depth estimation from a single still image by designing a novel CNN implementation of mean-field updates for continuous CRFs and proposes two different variations, one based on a cascade of multiple CRFs, the other on a unified graphical model.
- 111
- GD-MAE: Generative Decoder for MAE Pre-Training on LiDAR Point Clouds108
This work proposes a Generative Decoder for MAE (GD-MAE) to automatically merges the surrounding context to restore the masked geometric knowledge in a hierarchical fusion manner and achieves comparable accuracy even with 20% of the labeled data on the Waymo dataset.
- GLiT: Neural Architecture Search for Global and Local Image Transformer107
This work introduces the first Neural Architecture Search (NAS) method to find a better transformer architecture for image recognition and introduces a locality module that models the local correlations in images explicitly with fewer computational cost.
- 102
- LLaMA-Berry: Pairwise Optimization for O1-like Olympiad-Level Mathematical Reasoning101
An advanced mathematical problem-solving framework, LLaMA-Berry, for enhancing the mathematical reasoning ability of Large Language Models (LLMs), which combines Monte Carlo Tree Search (MCTS) with iterative Self-Refine to optimize the reasoning path and utilizes a pairwise reward model to evaluate different paths globally.
- Masked Motion Predictors are Strong 3D Action Representation Learners101
The proposed Masked Motion Prediction (MAMP) framework takes as input the masked spatio-temporal skeleton sequence and predicts the corresponding temporal motion of the masked human joints, promoting better attention to semantically rich temporal regions.
- Fast-BEV: A Fast and Strong Bird’s-Eye View Perception Baseline100
This paper proposes a simple yet effective framework, termed Fast-BEV, which is capable of performing faster BEV perception on the on-vehicle chips of autonomous driving, and introduces a strong data augmentation strategy for both image and BEV space to avoid over-fitting.
- Layerwise Optimization by Gradient Decomposition for Continual Learning100
This work decomposed the gradient of an old task into a part shared by all old tasks and a part specific to that task, and performs optimization for the gradients of each layer separately rather than the concatenation of all gradients as in previous works.
- From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery98
This work establishes a domain-oriented synthesis of autonomous scientific discovery and positions Agentic Science as a structured paradigm for advancing AI-driven research.
- Deep Kalman Filtering Network for Video Compression Artifact Reduction98
This paper model the video artifact reduction task as aKalman filtering procedure and restore decoded frames through a deep Kalman filtering network and builds a recursive filtering scheme based on the Kalman model.
- FiT: Flexible Vision Transformer for Diffusion Model97
The Flexible Vision Transformer (FiT), a transformer architecture specifically designed for generating images with unrestricted resolutions and aspect ratios, exhibits remarkable flexibility in resolution extrapolation generation.
- STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning96
The main insights of STEERER are a dedicated Feature Selection and Inheritance Adaptor (FSIA), which selectively forwards scale-customized features at each scale, and a Masked Selection and Inheritance Loss (MSIL) that helps to achieve high-quality density maps across all scales.
- 96
- UniHCP: A Unified Model for Human-Centric Perceptions95
This work proposes UniHCP, a Unified Model for Human-Centric Perceptions, which unifies a wide range of human-centric tasks in a simplified end-to-end manner with the plain vision transformer architecture and can outperform strong baselines on several in-domain and downstream tasks by direct evaluation.
- Online Hyper-Parameter Learning for Auto-Augmentation Strategy95
This paper proposes Online Hyper-parameter Learning for Auto-Augmentation (OHL-Auto-Aug), an economical solution that learns the augmentation policy distribution along with network training that eliminates the need of re-training and dramatically reduces the cost of the overall search process.
- Multi-Person Articulated Tracking With Spatial and Temporal Embeddings93
A unified framework for multi-person pose estimation and tracking, which combines SpatialNet and TemporalNet, and model the grouping procedure into a differentiable Pose-Guided Grouping (PGG) module to make the whole part detection and grouping pipeline fully end-to-end trainable.
- CasCast: Skillful High-resolution Precipitation Nowcasting via Cascaded Modelling92
This work proposes CasCast, a cascaded framework composed of a deterministic and a probabilistic part to decouple the predictions for mesoscale precipitation distributions and small-scale patterns and shows that CasCast achieves competitive performance.
- UniPAD: A Universal Pre-Training Paradigm for Autonomous Driving90
UniPAD is presented, a novel self-supervised learning paradigm applying 3D volumetric differentiable rendering that implicitly encodes 3D space, facilitating the reconstruction of continuous 3D shape structures and the intricate appear-ance characteristics of their 2D projections.
- Zoom Out-and-In Network with Map Attention Decision for Region Proposal and Object Detection90
Experimental results on three datasets demonstrate the effectiveness of the proposed zoom-out-and-in network over other state-of-the-arts, in terms of average recall for region proposal and average precision for object detection.
- SparseFlex: High-Resolution and Arbitrary-Topology 3D Shape Modeling89
This paper introduces SparseFlex, a novel sparse-structured isosurface representation that enables differentiable mesh reconstruction at resolutions up to 10243 directly from rendering losses, and introduces a frustum-aware sectional voxel training strategy that activates only relevant voxels during rendering, dramatically reducing memory consumption and enabling high-resolution training.
- MAP-Neo: Highly Capable and Transparent Bilingual Large Language Model Series89
The MAP-Neo is the first fully open-sourced bilingual LLM with comparable performance compared to existing state-of-the-art LLMs, and is hoped to enhance and strengthen the open research community and inspire more innovations and creativities to facilitate the further improvements of LLMs.
- Multi-task machine learning improves multi-seasonal prediction of the Indian Ocean Dipole89
Hindcasts of the Iod events during the past four decades indicate that the MTL-NET can predict the IOD well up to 7-month ahead, outperforming most of world-class dynamical models used for comparison in this study.
- Many Heads Are Better Than One: Improved Scientific Idea Generation by A LLM-Based Multi-Agent System88
This work proposes an LLM-based multi-agent system, i.e., Virtual Scientists (VirSci), designed to mimic the teamwork inherent in scientific research, and investigates the collaboration mechanisms that contribute to its tendency to produce ideas with higher novelty.
- 87
- Hierarchical Online Instance Matching for Person Search87
A Hierarchical Online Instance Matching (HOIM) loss is proposed which exploits the hierarchical relationship between detection and re-ID to guide the learning of the network and justifies the effectiveness of the proposed HOIM loss on learning robust features.
- Deep Continuous Conditional Random Fields With Asymmetric Inter-Object Constraints for Online Multi-Object Tracking87
A deep continuous conditional random field (DCCRF) is proposed for solving the online MOT problem in a track-by-detection framework and is trained in an end-to-end manner for better adapting the influences of visual information as well as inter-object relations.
- Instruct-ReID: A Multi-Purpose Person Re-Identification Task with Instructions84
A new instruct-ReID task that requires the model to retrieve images according to the given image or language instructions, and a large-scale OmniReID benchmark and an adaptive triplet loss as a baseline method to facilitate research in this new setting.
- 84
- Dividing and Aggregating Network for Multi-view Action Recognition83
A new Dividing and Aggregating Network (DA-Net) for multi-view action recognition based on the conditional random field (CRF) formulation to pass message among view-specific representations from different branches to help each other.
- Point Cloud Pre-Training with Diffusion Models82
This paper considers the point cloud pre-training task as a conditional point-to-point gen-eration problem and introduces a conditional point genera-tor that aggregates the features extracted by the backbone and employs them as the condition to guide the point-to-point recovery from the noisy point cloud, thereby assisting the backbone in capturing both local and global geometric priors as well as the global point density distri-bution of the object.
- 82
- 82
- PVT-SSD: Single-Stage 3D Object Detector with Point-Voxel Transformer81
A novel Point-Voxel Transformer for single-stage 3D detection (PVT-SSD) that takes advantage of voxel-based sparse convolutions for efficient feature encoding and adaptively fuses long-range contextual and local geometric information around reference points into content queries.
- Improving One-Shot NAS by Suppressing the Posterior Fading80
This paper analyzes existing weight sharing one-shot NAS approaches from a Bayesian point of view and identifies the Posterior Fading problem, which compromises the effectiveness of shared weights, and presents a novel approach to guide the parameter posterior towards its true distribution.
- 80
- Model Compression Using Progressive Channel Pruning79
This work proposes a simple but effective channel pruning framework called Progressive Channel Pruning (PCP) to accelerate Convolutional Neural Networks (CNNs) and extends the PCP framework to prune channels for the deep transfer learning methods like Domain Adversarial Neural Network (DANN), in which the data distribution mismatch is effectively reduced.
- MCTformer+: Multi-Class Token Transformer for Weakly Supervised Semantic Segmentation78
A novel transformer-based framework to generate accurate class-specific object localization maps for weakly supervised semantic segmentation (WSSS) and introduces a Contrastive-Class-Token (CCT) module to enhance the learning of discriminative class tokens, enabling the model to better capture the unique characteristics of each class.
- AM-LFS: AutoML for Loss Function Search77
This paper proposes AutoML for Loss Function Search (AM-LFS) which leverages REINFORCE to search loss functions during the training process and proposes an efficient optimization framework which can dynamically optimize the parameters of loss function's distribution during training.
- 76
- 75
- Deep Instance Segmentation With Automotive Radar Detection Points75
This paper proposes an efficient method based on clustering of estimated semantic information to achieve instance segmentation for the sparse radar detection points and shows that the performance of the proposed approach can be further enhanced by incorporating the visual multi-layer perceptron.
- VEnhancer: Generative Space-Time Enhancement for Video Generation74
A generative space-time enhancement framework that improves the existing text-to-video results by adding more details in spatial domain and synthetic detailed motion in temporal domain, and surpasses existing state-of-the-art video super-resolution and space-time super-resolution methods in enhancing AI-generated videos.
- Graph-Based 3D Multi-Person Pose Estimation Using Multi-View Images74
This paper decomposes the task of estimating the 3D human poses of multiple persons from multiple calibrated camera views into two stages, i.e. person localization and pose estimation, and proposes three task-specific graph neural networks for effective message passing.
- Gradient Regularized Contrastive Learning for Continual Domain Adaptation73
This work studies the problem of continual domain adaptation, where the model is presented with a labelled source domain and a sequence of unlabelled target domains, and proposes Gradient Regularized Contrastive Learning (GRCL), a method that constraining the gradient not to increase the classification loss on old target domains enables the model to preserve the performance onold target domains when adapting to an in-coming target domain.
- Relational Prototypical Network for Weakly Supervised Temporal Action Localization73
This paper proposes a weakly supervised temporal action localization method on untrimmed videos based on prototypical networks that outperforms the state-of-the-art methods on action-background separation and action relation construction.
- Unsupervised Learning of Accurate Siamese Tracking71
A novel unsupervised tracking framework is presented, in which a differentiable region mask is proposed to select features as well as to implicitly penalize tracking errors on intermediate frames, and a mask-guided loss reweighting strategy to assign dynamic weights based on the quality of pseudo labels is proposed.
- Better Teacher Better Student: Dynamic Prior Knowledge for Knowledge Distillation71
The dynamic prior knowledge (DPK) is proposed, which integrates part of teacher's features as the prior knowledge before the feature distillation, which provides a fast solution in teacher model selection for any given model.
- MotionGPT-2: A General-Purpose Motion-Language Model for Motion Generation and Understanding70
The proposed MotionGPT-2 is highly adaptable to the challenging 3D holistic motion generation task, enabled by the innovative motion discretization framework, Part-Aware VQVAE, which facilitates fine-grained representations of body and hand movements.
- 69
- PonderV2: Pave the Way for 3D Foundation Model with A Universal Pre-training Paradigm68
A novel universal 3D pre-training framework designed to facilitate the acquisition of efficient 3D representation by differentiable neural rendering is introduced, thereby establishing a pathway to 3D foundational models.
- Critic-V: VLM Critics Help Catch VLM Errors in Multimodal Reasoning67
Critic-V, a novel framework inspired by the Actor-Critic paradigm to boost the reasoning capability of VLMs, provides a promising solution to enhance the reliability of VLMs, improving their performance in real-world reasoning-heavy multimodal applications such as autonomous driving and embodied intelligence.
- Depth Any Video with Scalable Synthetic Data67
The approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates-even on single frames, and proposes a depth interpolation method that enables the model to infer high-resolution video depth across sequences of up to 150 frames.
- The theoretical research of generative adversarial networks: an overview67
This paper focuses on the theoretical achievements of GAN and introduces the improved methods of combining encoder, autoencoder or VAE, and some other important issues, such as the quantify metrics of generated samples and the basic construction structure.
- AI-Enabled Parallel Assembly of Thousands of Defect-Free Neutral Atom Arrays66
An AI-enabled, rapid, constant-time-overhead rearrangement protocol that can be readily used to generate defect-free arrays of tens of thousands of atoms with current technologies and become a useful toolbox for quantum error correction.
- HiSplat: Hierarchical 3D Gaussian Splatting for Generalizable Sparse-View Reconstruction65
This paper proposes a novel framework, HiSplat, which introduces a hierarchical manner in generalizable 3D Gaussian Splatting to construct hierarchical 3D Gaussians via a coarse-to-fine strategy, and significantly enhances reconstruction quality and cross-dataset generalization compared to prior single-scale methods.
- 65
- 65
- Bi-LRFusion: Bi-Directional LiDAR-Radar Fusion for 3D Dynamic Object Detection64
This work introduces a bi-directional LiDAR-Radar fusion framework, termed Bi-LRFusion, to tackle the challenges and improve 3D detection for dynamic objects and achieves state-of-the-art performance for detecting dynamic objects.
- ViPNAS: Efficient Video Pose Estimation via Neural Architecture Search64
A novel neural architecture search (NAS) method, termed ViP-NAS, to search networks in both spatial and temporal levels for fast online video pose estimation, which is the first to search for the temporal feature fusion and automatic computation allocation in videos.
- Cache-to-Cache: Direct Semantic Communication Between Large Language Models63
The proposed Cache-to-Cache (C2C), a new paradigm for direct semantic communication between LLMs, uses a neural network to project and fuse the source model's KV-cache with that of the target model to enable direct semantic transfer.
- Accurate Fine-Grained Object Recognition with Structure-Driven Relation Graph Networks62
This paper argues that the personalized differences could decline the network’s perception of discriminative features, thus discarding some discriminative clues and degrading the FGOR performance accordingly, and proposes an end-to-end Structure-driven Relation Graph Network (SRGN), aimed at exploring and exploiting the object structure information without any additional annotations.
- Revisiting the Transferability of Supervised Pretraining: an MLP Perspective62
This paper attempts to close the transferability gap between supervised and unsupervised pretraining by adding an MLP projector before the classifier in supervised pretraining, and indicates that theMLP projector can help retain intra-class variation of visual features, decrease the feature distribution distance between pretraining and evaluation datasets, and reduce feature redundancy.
- 62
- HumanBench: Towards General Human-Centric Perception with Projector Assisted Pretraining61
A HumanBench based on existing datasets is proposed to comprehensively evaluate on the common ground the generalization abilities of different pretraining methods on 19 datasets from 6 diverse downstream tasks, including person ReID, pose estimation, human parsing, pedestrian attribute recognition, pedestrian detection, and crowd counting.
- 61
- A general temperature-guided language model to design proteins of enhanced stability and activity60
This work introduces PRIME, a deep learning model, which can suggest protein mutants with improved stability and activity without any prior experimental mutagenesis data for the specified protein, and demonstrates broad applicability in protein engineering.
- Ponder: Point Cloud Pre-training via Neural Rendering60
A point-cloud encoder is trained within a devised point-based neural renderer by comparing the rendered images with real images on massive RGB-D data to demonstrate the superiority of this approach compared to existing pre-training methods.
- Channel Pruning Guided by Classification Loss and Feature Importance59
This work proposes a new layer-by-layer channel pruning method called Channel Pruning guided by classification Loss and feature Importance (CPLI), which additionally take the classification loss into account in the channel pruned process.
- Learnable Histogram: Statistical Context Features for Deep Neural Networks57
A learnable histogram layer is proposed, which learns histogram features within deep neural networks in end-to-end training and is able to back-propagate (BP) errors, learn optimal bin centers and bin widths, and be jointly optimized with other layers in deep networks during training.
- 56
- BriNet: Towards Bridging the Intra-class and Inter-class Gaps in One-Shot Segmentation56
A framework to bridge the gap between information interaction between query and support images and to precisely localize the query objects is proposed, which outperforms other competitive methods and leads to a new state-of-the-art on both PASCAL VOC and MSCOCO dataset.
- Dense Video Captioning Using Graph-Based Sentence Summarization56
A graph-based partition-and-summarization (GPaS) framework for dense video captioning within two stages that effectively exploits the relationship between semantic words for summarization and achieves this goal by treating semantic words as the nodes in a graph.
- Structured Modeling of Joint Deep Feature and Prediction Refinement for Salient Object Detection56
This paper designs a novel cascade CRFs architecture with CNN to jointly refine deep features and predictions at each scale and progressively compute a final refined saliency map, and formulate the CRF graphical model that involves message-passing of feature-feature, feature-prediction, and prediction-predicted, from the coarse scale to the finer scale.
- Neural Network Encapsulation54
The proposed EncapNet performs favorably better against previous state-of-the-arts on CIFAR10/100, SVHN and a subset of ImageNet and an add-on to send back higher capsules as feedback is devised, achieving the optimal transport divergence between two distributions.
- 54
- 54
- GPT4Vis: What Can GPT-4 Do for Zero-shot Visual Recognition?53
Evaluation of GPT-4's linguistic and visual capabilities in zero-shot visual recognition tasks shows that GPT-4, enhanced with rich linguistic descriptions, significantly improves zero-shot recognition, offering an average top-1 accuracy increase of 7% across all datasets.
- Once Quantization-Aware Training: High Performance Extremely Low-bit Architecture Search53
This work proposes to combine Network Architecture Search methods with quantization with a shared step size to acquire a large number of quantized models and introduces a bit-inheritance scheme, which further reduces the time cost and meanwhile improves the quantization accuracy.
- FengWu-GHR: Learning the Kilometer-scale Medium-range Global Weather Forecasting52
F FengWu-GHR is presented, the first data-driven global weather forecasting model running at the 0.09$^{\circ}$ horizontal resolution and introduces a novel approach that opens the door for operating ML-based high-resolution forecasts by inheriting prior knowledge from a pretrained low-resolution model.
- Self-Paced Collaborative and Adversarial Network for Unsupervised Domain Adaptation52
This paper proposes a new unsupervised domain adaptation approach called Collaborative and Adversarial Network (CAN), which uses the domain-collaborative and domain-adversarial learning strategies for training the neural network.
- Improved Boundary Equilibrium Generative Adversarial Networks52
This work proposes an effective approach to generate images with higher quality and better diversity in BEGANs framework by adding a second loss function (a denoising loss) to the discriminator, so that the ability of discriminator in distinguishing between real and generated images is improved.
- Learning Mutual Visibility Relationship for Pedestrian Detection with a Deep Model52
A mutual visibility deep model that jointly estimates the visibility statuses of overlapping pedestrians is proposed that has the lowest average miss rate on the Caltech-Train dataset and the ETH dataset.
- FengWu-4DVar: Coupling the Data-driven Weather Forecasting Model with 4D Variational Assimilation51
The global AI weather forecasting model, FengWu, is integrated with one of the most popular assimilation algorithms, Four-Dimensional Variational (4DVar) assimilation, and an AI-based cyclic weather forecasting system, FengWu-4DVar is developed, capable of generating reasonable analysis fields for making accurate and efficient iterative predictions.
- Evolving Search Space for Neural Architecture Search51
The proposed NSE scheme achieves 77.3% top-1 retrain accuracy on ImageNet with 333M FLOPs, which yielded a state-of-the-art performance among previous auto-generated architectures that do not involve knowledge distillation or weight pruning.
- 51
- A Comprehensive Survey on 3D Content Generation50
A new taxonomy is proposed that categorizes existing approaches into three types: 3D native generative methods, 2D prior-based 3D generative methods, and hybrid 3D generative methods.
- 50
- ViP-CNN: A Visual Phrase Reasoning Convolutional Neural Network for Visual Relationship Detection.50
In ViP-CNN, the visual relationship is considered as a phrase with three components and a Visual Phrase Reasoning Structure (VPRS) is presented to set up the connection among the relationship components and help the model consider the three problems jointly.
- 45
- 45
- Inception Convolution with Efficient Dilation Search45
A new type of dilated convolution is proposed, where the convolution operations have independent dilation patterns among different axes, channels and layers, and this method achieves consistent performance gains for image recognition, object detection, instance segmentation, human detection, and human pose estimation.
- 45
- 44
- Transferring Vision-Language Models for Visual Recognition: A Classifier Perspective44
This study revisits the role of the linear classifier in the vanilla transfer learning framework, and proposes a new paradigm where the parameters of the classifier are initialized with semantic targets from the textual encoder and remain fixed during optimization.
- Reconstructing Hand-Held Objects from Monocular Video44
An implicit neural representation-based method is devised to solve the reconstruction problem and address the issues of imprecise hand pose estimation, relative hand-object motion, and insufficient geometry optimization for small objects.
- 44
- Show, Tell and Summarize: Dense Video Captioning Using Visual Cue Aided Sentence Summarization43
This work formulate the dense video captioning task as a visual cue aided sentence summarization problem and proposes a new two stage Long Short Term Memory (LSTM) approach equipped with a new hierarchical attention mechanism to summarize all generated sentences as one descriptive sentence with the aid of visual features.
- 41
- 40
- Interleaving Reasoning for Better Text-to-Image Generation40
This work introduces Interleaving Reasoning Generation (IRG), a framework that alternates between text-based thinking and image synthesis: the model first produces a text-based thinking to guide an initial image, then reflects on the result to refine fine-grained details, visual quality, and aesthetics while preserving semantics.
- ComfyBench: Benchmarking LLM-based Agents in ComfyUI for Autonomously Designing Collaborative AI Systems40
ComfyAgent is developed, a novel framework that empowers LLM-based agents to autonomously design collaborative AI systems by generating workflows and constructs a multi-agent system that cooperates to learn from existing workflows and generate new workflows for a given task.
- 40
- DaNet40
The Decompose-and-aggregate Network (DaNet) is presented, which decomposes the prediction task into one global stream and multiple local streams so that the network not only provides global perception for the camera and shape prediction, but also has detailed perception for part pose prediction.
- The Avengers: A Simple Recipe for Uniting Smaller Language Models to Challenge Proprietary Giants39
The Avengers -- a simple recipe that leverages the collective intelligence of these smaller models to deliver superior out-of-distribution generalization, and remains robust across various embedding models, clustering algorithms, ensemble strategies, and values of its sole parameter -- the number of clusters.
- Dynamic Position-aware Network for Fine-grained Image Recognition39
An end-to-end Dynamic Position-aware Network (DP-Net) is proposed to directly incorporate the position clues into visual content and dynamically align them without extra annotations, which eliminates the effect of position information for visual variances of subcategories.
- 38
- DASOT: A Unified Framework Integrating Data Association and Single Object Tracking for Online Multi-Object Tracking38
An efficient two-stage tracking method in the SOT branch, which utilizes the merits of correlation features and can simultaneously track all the existing targets within one forward propagation, is proposed.
- 38
- 37
- BN-NAS: Neural Architecture Search with Batch Normalization37
BN-NAS can significantly reduce the time required by model training and evaluation in NAS and a BN-based indicator for predicting subnet performance at a very early training stage is proposed for fast evaluation.
- 36
- 36
- 35
- 35
- Neuro-3D: Towards 3D Visual Decoding from EEG Signals34
Experiments indicate that Neuro-3D not only reconstructs colored 3D objects with high fidelity, but also learns effective neural representations that enable insightful brain region analysis that enable insightful brain region analysis.
- 34
- Hulk: A Universal Knowledge Translator for Human-Centric Tasks33
Hulk is presented, the first multimodal human-centric generalist model, capable of addressing 2D vision, 3D vision, skeleton-based, and vision-language tasks without task-specific finetuning, and achieving state-of-the-art performance on 12 benchmarks.
- Frozen CLIP Transformer Is an Efficient Point Cloud Encoder33
This paper introduces Efficient Point Cloud Learning (EPCL), an effective and efficient point cloud learner for directly training high-quality point cloud models with a frozen CLIP transformer and achieves promising performance on both indoor and outdoor benchmarks.
- The Equalization Losses: Gradient-Driven Training for Long-tailed Object Recognition33
A gradient-driven training mechanism to tackle the long-tail problem: re-balancing the positive/negative gradients dynamically according to current accumulative gradients, with a unified goal of achieving balance gradient ratios.
- 33
- 32
- 32
- Progressive Cross-Stream Cooperation in Spatial and Temporal Domain for Action Localization32
This work proposes a new progressive cross-stream cooperation (PCSC) framework that improves all three levels of tasks of spatio-temporal action localization, and proposes a new message passing approach to pass information from one stream to another stream, which also leads to better action detection and segment detection models.
- Improved generative adversarial networks with reconstruction loss32
A reconstruction loss is added to the GAN’s objective function to enforce the generator can reconstruct from the features of the discriminator, which helps to explicitly guide the generator towards to near the probable configurations of real data.
- A Survey of Scientific Large Language Models: From Data Foundations to Agent Frontiers31
A comprehensive, data-centric synthesis that reframes the development of Sci-LLMs as a co-evolution between models and their underlying data substrate, and outlines a paradigm shift toward closed-loop systems where autonomous agents based on Sci-LLMs actively experiment, validate, and contribute to a living, evolving knowledge base.
- Data-driven global ocean modeling for seasonal to decadal prediction31
ORCA-DL (Oceanic Reliable foreCAst via Deep Learning), a data-driven three-dimensional ocean model for seasonal to decadal prediction of global ocean dynamics, accurately simulates the three-dimensional structure of global ocean dynamics with high physical consistency and outperforms state-of-the-art numerical models in capturing extreme events.
- π-PrimeNovo: an accurate and efficient non-autoregressive deep learning model for de novo peptide sequencing31
π-PrimeNovo is introduced, a non-autoregressive Transformer-based model for peptide sequencing that achieves high accuracy and up to 89x faster sequencing, which enables large-scale sequencing and multiple downstream applications.
- ExtremeCast: Boosting Extreme Value Prediction for Global Weather Forecast31
This work introduces Exloss, a novel loss function that performs asymmetric optimization and highlights extreme values to obtain accurate extreme weather forecast and introduces a training-free extreme value enhancement module named ExBooster, which captures the uncertainty in prediction outcomes by employing multiple random samples, thereby increasing the hit rate of low-probability extreme events.
- Iterative Length-Regularized Direct Preference Optimization: A Case Study on Improving 7B Language Models to GPT-4 Level31
This work identifies a pitfall of vanilla iterative DPO - improved response quality can lead to increased verbosity, and introduces iterative length-regularized DPO (iLR-DPO) to penalize response length.
- 31
- RePre: Improving Self-Supervised Vision Transformer with Reconstructive Pre-training30
This paper incorporates local feature learning into self-supervised vision transformers via Reconstructive Pre-training (RePre), which extends contrastive frameworks by adding a branch for reconstructing raw image pixels in parallel with the existing contrastive objective.
- DR.VIC: Decomposition and Reasoning for Video Individual Counting30
This work proposes to conduct the pedes-trian counting from a new perspective - Video Individual Counting (VIC), which counts the total number of individual pedestrians in the given video (a person is only counted once).
- 30
- 30
- CoMAS: Co-Evolving Multi-Agent Systems via Interaction Rewards29
This work introduces Co-Evolving Multi-Agent Systems (CoMAS), a novel framework that enables agents to improve autonomously by learning from inter-agent interactions without external supervision, thereby enabling decentralized and scalable co-evolution in LLM-based agents.
- GigaGS: Scaling up Planar-Based 3D Gaussians for Large Scene Surface Reconstruction29
The first attempt to tackle the challenging task of large-scale scene surface reconstruction for large-scale scenes using 3D Gaussian Splatting, and proposes GigaGS, the first work for high-quality surface reconstruction for large-scale scenes using 3DGS.
- Uni3D-LLM: Unifying Point Cloud Perception, Generation and Editing with Large Language Models29
The efficacy of Uni3D-LLM in the comprehension, generation, and editing of point cloud has been validated and the impact of integrating a point cloud perception module on the generation and editing processes is assessed, confirming the substantial potential of the approach for practical applications.
- 29
- 29
- TCFormer: Visual Recognition via Token Clustering Transformer28
The Token Clustering Transformer (TCFormer), which generates dynamic vision tokens based on semantic meaning that represent image regions with similar semantic meanings using the same vision token, even if those regions are not adjacent.
- TASeg: Temporal Aggregation Network for LiDAR Semantic Segmentation28
A Temporal LiDAR Aggregation and Distillation (TLAD) algorithm, which leverages historical priors to assign dif-ferent aggregation steps for different classes, and a Static-Moving Switch Augmentation (SMSA) algorithm, which utilizes sufficient temporal information to enable objects to switch their motion states freely, thus greatly increasing static and moving training samples.
- Octavius: Mitigating Task Interference in MLLMs via LoRA-MoE28
Octavius is fine-tuned in three modality setups: i.
- 28
- Progressive Modality Cooperation for Multi-Modality Domain Adaptation28
This work proposes a new generic multi-modality domain adaptation framework called Progressive Modality Cooperation (PMC) to transfer the knowledge learned from the source domain to the target domain by exploiting multiple modality clues under the MMDA and MMDA-PI settings.
- Probabilistic Graph Attention Network with Conditional Kernels for Pixel-Wise Prediction28
Quantitative and qualitative results demonstrate the effectiveness of the proposed latent AG-CRF model and the overall probabilistic graph attention network with feature conditional kernels for structured feature learning and pixel-wise prediction.
- BEACON: Benchmark for Comprehensive RNA Tasks and Language Models27
This study introduces the first comprehensive RNA benchmark BEACON (BEnchmArk for COmprehensive RNA Task and Language Models), and emphasizes the superiority of single nucleotide tokenization and the effectiveness of Attention with Linear Biases (ALiBi) over traditional positional encoding methods.
- 27
- A CLIP-Powered Framework for Robust and Generalizable Data Selection26
This work proposes a novel CLIP-powered data selection framework that leverages multimodal information for more robust and generalizable sample selection and effectively removes noisy or damaged samples from the dataset, enabling it to achieve even higher performance with less data.
- From GPT-4 to Gemini and Beyond: Assessing the Landscape of MLLMs on Generalizability, Trustworthiness and Causality through Four Modalities26
This paper evaluates the closed-source GPT-4 and Gemini and 6 open-source LLMs and MLLMs and uncover 14 empirical findings that are useful to understand the capabilities and limitations of both proprietary and open-source MLLMs, towards more reliable downstream multi-modal applications.
- Category-specific Semantic Coherency Learning for Fine-grained Image Recognition26
An end-to-end Category-specific Semantic Coherency Network (CSC-Net) to semantically align the discriminative regions of the same subcategory is proposed and yields the best performance under the same settings with most competitive approaches.
- 25
- MambaDS: Near-Surface Meteorological Field Downscaling With Topography Constrained Selective State-Space Modeling25
This article pioneers the selective state-space model (SSM) into the meteorological field downscaling and proposes a novel model called MambaDS, which retains the advantages of Mamba in long-range dependency modeling and linear computational complexity while enhancing the learning ability of multivariate correlation.
- Leapfrogging Sycamore: harnessing 1432 GPUs for 7× faster quantum random circuit sampling25
This work provides the first unambiguous experimental evidence to refute Sycamore’s claim of quantum advantage, and redefines the boundary of quantum computational advantage using random circuit sampling.
- 25
- 25
- 25
- 25
- Dolphin: Moving Towards Closed-loop Auto-research through Thinking, Practice, and Feedback24
Dolphin, a closed-loop LLM-driven framework to enhance the automation level of scientific research is introduced and it is highlighted that Dolphin can automatically propose methods that are comparable to the state-of-the-art in some tasks such as 3D point classification.
- Do We Truly Need So Many Samples? Multi-LLM Repeated Sampling Efficiently Scales Test-Time Compute24
This paper presents a simple, effective, and cost-efficient strategy, named ModelSwitch, to improve LLM performance by scaling test-time compute by using sample consistency as a signal, and dynamically switches between models.
- What Can Simple Arithmetic Operations Do for Temporal Modeling?24
This work explores the potential of four simple arithmetic operations for temporal modeling by computing addition, subtraction, multiplication, and division between pairs of extracted frame features to benefit the original temporal-irrespective domain.
- 24
- 24
- Exploring Representation-Aligned Latent Space for Better Generation23
ReaLS (Representation-Aligned Latent Space), which integrates semantic priors to improve generation performance and enables more perceptual downstream tasks, such as segmentation and depth estimation, is introduced.
- GUPNet++: Geometry Uncertainty Propagation Network for Monocular 3D Object Detection23
A novel Geometry Uncertainty Propagation Network (GUPNet++) is proposed by modeling geometry projection in a probabilistic manner that models the uncertainty propagation relationship of the geometry projection during training, improving the stability and efficiency of the end-to-end model learning.
- Partial Fine-Tuning: A Successor to Full Fine-Tuning for Vision Transformers23
A novel fine-tuned angle metric is proposed to guide the selection of appropriate layers for partial fine-tuning, making it flexible to be adapted to various scenarios for more practicable partial fine-tuning.
- 23
- 23
- PhysUniBench: A Multi-Modal Physics Reasoning Benchmark at Undergraduate Level22
PhysUniBench is a large-scale multimodal benchmark designed to evaluate and improve the reasoning capabilities of multimodal large language models (MLLMs) specifically on undergraduate-level physics problems, encouraging the development of models with stronger physical reasoning, problem-solving skills, and multimodal understanding.
- Glasses-free 3D display with ultrawide viewing range using deep learning22
A glasses-free three-dimensional display, which is as large as a conventional desktop screen and has a viewing angle larger than 100°, is achieved by an artificial intelligence-powered optimization algorithm, which adapts the virtual images sent around each eye in real time.
- 22
- 22
- EPCL: Frozen CLIP Transformer is An Efficient Point Cloud Encoder22
Comprehensive experiments on 3D detection, semantic segmentation, classification and few-shot learning demonstrate that the 2D CLIP model can be an efficient point cloud backbone and the method achieves state-of-the-art accuracy on both real-world and synthetic downstream tasks.
- Part-aligned pose-guided recurrent network for action recognition22
A Part-aligned Pose-guided Recurrent Network (P2RN) for action recognition is proposed, which is a recurrent network which can exploit the body representation and simultaneously model spatio-temporal evolutions of human body parts.
- Discovery of Potent STING Inhibitors Bearing a Difluorobenzodioxol Structural Motif as Potent Anti-Inflammatory Agents21
A structural modification campaign on the tool compound 6 (H-151) based on its potential metabolic hotspots resulted in compound 66, bearing a difluorobenzodioxol moiety, being identified as one of the most potent STING inhibitors with IC50 values of 116 and 96.3 nM.
- RH20T-P: A Primitive-Level Robotic Dataset Towards Composable Generalization Agents21
This work proposes RH20T-P, a primitive-level robotic manipulation dataset, which contains about 38k video clips covering 67 diverse manipulation tasks in real-world scenarios, and standardizes a plan-execute CGA paradigm and implements an exemplar baseline called RA-P, whose positive performance on solving unseen tasks validates that the proposed dataset can offer composable generalization ability to robotic manipulation agents.
- Merging Vision Transformers from Different Tasks and Domains21
A simple-but-effective gating network is proposed that can both merge all kinds of layers and select the suitable classifier to enable the merging of the whole ViT, and a novel metric of model weight similarity is designed to realize controllable and combined weight merging.
- TAR3D: Creating High-Quality 3D Assets Via Next-Part Prediction20
TAR3D is presented, a novel framework that consists of a 3D-aware Vector Quantized-Variational AutoEncoder and a Generative Pre-trained Transformer to generate high-quality 3D assets to migrate the multimodal unification and promising learning capabilities of the next-token prediction paradigm to conditional 3D object generation.
- MindAligner: Explicit Brain Functional Alignment for Cross-Subject Visual Decoding from Limited fMRI Data20
Experiments indicate that MindAligner not only outperforms existing methods in visual decoding under data-limited conditions, but also provides valuable neuroscience insights in cross-subject functional analysis.
- 20
- HVDistill: Transferring Knowledge from Images to Point Clouds via Unsupervised Hybrid-View Distillation20
A hybrid-view-based knowledge distillation framework, termed HVDistill, to guide the feature learning of a point cloud neural network with a pre-trained image network in an unsupervised manner and achieves consistent improvements over the baseline trained from scratch and significantly outperforms the existing schemes.
- ResoNet: Robust and Explainable ENSO Forecasts with Hybrid Convolution and Transformer Networks20
It is shown that ResoNet can robustly predict ENSO at lead times of 19 months, thus outperforming existing approaches in terms of the forecast horizon and it is demonstrated for the first time that the asymmetry between El Niño and La Niña development can be captured by ResoNet.
- 20
- 20
- 20
- A spatiotemporal attention-based ResC3D model for large-scale gesture recognition20
A spatiotemporal attention-based ResC3D model for abnormal gesture recognition with large-scale videos that outperforms other state-of-the-art methods, including the basic work in 2017 Chalearn Looking at People Workshop of ICCV.
- 20
- 19
- Online test-time adaptation for better generalization of interatomic potentials to out-of-distribution data19
A dual-level self-supervised learning approach that leverages global structure and atomic local environment information to align the model with the test data is designed, allowing for more stable simulations without the need for additional training data.
- 19
- 19
- 19
- Multiple sclerosis lesion segmentation: revisiting weighting mechanisms for federated learning19
This work proposes the first FL MS lesion segmentation framework via two effective re-weighting mechanisms, where a learnable weight is assigned to each local node during the aggregation process, based on its segmentation performance.
- 19
- 19
- Cheaper Pre-training Lunch: An Efficient Paradigm for Object Detection19
The efficiency and effectiveness of Montage pre-training are validated by extensive experiments on the MS-COCO dataset, where the results indicate that the models using Montages are able to achieve on-par or even better detection performances compared with the ImageNet pre- training.
- TRB: A Novel Triplet Representation for Understanding 2D Human Body19
This paper proposes the Triplet Representation for Body --- a compact 2D human body representation, with skeleton keypoints capturing human pose information and contour keypoints containing human shape information, and proposes a two-branch network (TRB-net) with three novel techniques, namely X-structure, Directional Convolution and Pairwise mapping.
- Referring Expression Comprehension with Semantic Visual Relationship and Word Mapping19
A word2vec based word mapping method that maps these low frequency words to high frequency words with similar meaning and outperforms existing state-of-the-art methods on three referring expression comprehension datasets.
- Efficient and Universal Neural-Network Decoder for Stabilizer-Based Quantum Error Correction18
The introduction of GraphQEC, a code-agnostic decoder leveraging machine-learning on the graph structure of stabilizer codes with linear time complexity, which represents the first universal solution for real-time quantum error correction across arbitrary stabilizer codes.
- Cut2Next: Generating Next Shot via In-Context Tuning18
Next Shot Generation (NSG) is introduced: synthesizing a subsequent, high-quality shot that critically conforms to professional editing patterns while upholding rigorous cinematic continuity, validating its ability to generate high-quality, narratively expressive, and cinematically coherent subsequent shots.
- Integration of cognitive tasks into artificial general intelligence test for large models18
This work proposes increasing the complexity of AGI testing tasks commensurate with advancements in large models and emphasizing the necessity for the interpretation of test results to avoid false negatives and false positives.
- Towards Frame Rate Agnostic Multi-object Tracking18
A Frame Rate Agnostic MOT framework with a Periodic training Scheme (FAPS) to tackle the FraMOT problem for the first time and proposes a Frame Rate Agnostic Association Module (FAAM) that infers and encodes the frame rate information to aid identity matching across multi-frame-rate inputs, improving the capability of the learned model in handling complex motion-appearance relations in FraMOT.
- CompBench: Benchmarking Complex Instruction-guided Image Editing17
CompBench is introduced, a large-scale benchmark specifically designed for complex instruction-guided image editing and proposes an instruction decoupling strategy that disentangles editing intents into four key dimensions: location, appearance, dynamics, and objects, ensuring closer alignment between instructions and complex editing requirements.
- Adapter-X: A Novel General Parameter-Efficient Fine-Tuning Framework for Vision17
The proposed Adapter-X is the first to outperform full fine-tuning in both 2D image and 3D point cloud modalities with significantly fewer parameters, i.e., only 0.20% and 1.88% of original trainable parameters for 2D and 3D classification tasks.
- 17
- FiTv2: Scalable and Improved Flexible Vision Transformer for Diffusion Model17
This work conceptualizes images as sequences of tokens with dynamic sizes, rather than traditional methods that perceive images as fixed-resolution grids, which enables a flexible training strategy that seamlessly accommodates various aspect ratios during both training and inference, thus promoting resolution generalization and eliminating biases introduced by image cropping.
- 3D-QueryIS: A Query-based Framework for 3D Instance Segmentation17
This paper proposes to generate representative points in an implicit manner, and use them together with the initial queries to generate the informative instance queries, and concludes that the 3D-QueryIS is free from the accumulated errors caused by the inter-task dependencies.
- 17
- 17
- MeshCraft: Exploring Efficient and Controllable Mesh Generation with Flow-based DiTs16
MeshCraft is introduced, a novel framework for efficient and controllable mesh generation, which leverages continuous spatial diffusion to generate discrete triangle faces and integrates seamlessly with existing conditional guidance strategies, showcasing its potential to relieve artists from the time-consuming manual work involved in mesh creation.
- Evidential deep learning for interatomic potentials16
The authors introduce an evidential deep learning framework for machine learning interatomic potentials that efficiently provides robust uncertainty quantification, demonstrating its effectiveness across diverse atomic systems.
- WeatherGFM: Learning A Weather Generalist Foundation Model via In-context Learning16
This paper introduces the first generalist weather foundation model (WeatherGFM), designed to address a wide spectrum of weather understanding tasks in a unified manner and adopts a visual prompting question-answering paradigm for the training of unified weather understanding tasks.
- MM-3DScene: 3D Scene Understanding by Customizing Masked Modeling with Informative-Preserved Reconstruction and Self-Distilled Consistency16
A novel informative-preserved reconstruction is proposed, which explores local statistics to discover and preserve the representative structured points, effectively enhancing the pretext masking task for 3D scene understanding.
- 16
- FengWu-W2S: A deep learning model for seamless weather-to-subseasonal forecast of global atmosphere15
The hindcast results demonstrate that FengWu-W2S reliably predicts atmospheric conditions out to 3-6 weeks ahead, enhancing predictive capabilities for global surface air temperature, precipitation, geopotential height and intraseasonal signals such as the Madden-Julian Oscillation and North Atlantic Oscillation.
- 15
- 15
- 15
- Auxiliary Tasks Enhanced Dual-Affinity Learning for Weakly Supervised Semantic Segmentation15
A cross-task dual-affinity learning module to learn both pairwise and unary affinities, which are used to enhance the task-specific features and predictions by aggregating both query-dependent and query-independent global context for both saliency detection and semantic segmentation.
- Unsupervised contrastive learning with simple transformation for 3D point cloud data15
Comprehensive results show that the unsupervised contrastive representation learning generally outperforms current unsupervised methods, and even achieves comparable performance to supervised methods.
- 15
- SepFusion: Finding Optimal Fusion Structures for Visual Sound Separation15
This work proposes SepFusion, a novel framework that can smoothly produce optimal fusion structures for visual-sound separation and provides a series of strong models for broader applications, such as further promoting performance via model assembly, or providing suitable architectures for the separation of certain instrument classes.
- 15
- Improving Weakly Supervised Temporal Action Localization by Exploiting Multi-Resolution Information in Temporal Domain15
This work proposes a two-stage approach to generate high-quality frame-level pseudo labels by fully exploiting multi-resolution information in the temporal domain and complementary information between the appearance and motion streams, and uses a Progressive Temporal Label Refinement framework to iteratively refine the pseudo labels.
- Light field reconstruction using hierarchical features fusion15
A novel end-to-end U-Net with SAS network (U-SAS-Net) is introduced to extract and fuse hierarchical features, both local and semantic, from a relatively large receptive field while establishing the relation of the correlated sub-aperture images.
- 15
- Image postprocessing by Non-local Kuan’s filter15
The DNLK filter is extended to form the Overcomplete Dual Non-local Kuan's (OCDNLK) filter by applying to the overcomplete DCT coefficients, which shows the effectiveness of the two methods.
- LLaMA-Berry: Pairwise Optimization for Olympiad-level Mathematical Reasoning via O1-like Monte Carlo Tree Search14
The Pairwise Preference Reward Model (PPRM), which predicts pairwise preferences between solutions through instruction-following capabilities trained by Reinforcement Learning from Human Feedback (RLHF), is proposed and the Enhanced Borda Count (EBC) method is adopted to synthesize pairwise preferences into global quantile scores for evaluations.
- 14
- 14
- 14
- Automatically Predicting Material Properties with Microscopic Images: Polymer Miscibility as an Example14
This work achieves automatic miscibility recognition utilizing a convolutional neural network and transfer learning methods, and the model obtains up to 94% accuracy.
- 14
- Solving parametric high-Reynolds-number wall-bounded turbulence around airfoils governed by Reynolds-averaged Navier–Stokes equations using time-stepping-oriented neural network13
This work proposes a semi-coupled strategy to address the convergence issues caused by the turbulence model, and successfully solves turbulence around airfoils governed by the Reynolds-Averaged Navier–Stokes (RANS) equations with the Spalart–Allmaras (SA) turbulence model.
- Deep generative model for protein subcellular localization prediction13
After trained with both protein primary sequences and protein subcellular localization fluorescence images, deepGPS shows the ability to predict cytoplasmic and nuclear localizations by reporting both textual labels and generative images as outputs.
- PostCast: Generalizable Postprocessing for Precipitation Nowcasting via Unsupervised Blurriness Modeling13
An unsupervised postprocessing method is proposed to eliminate the blurriness in precipitation nowcasting without the requirement of training with the pairs of blurry predictions and corresponding ground truth and demonstrates the generality and superiority of this method.
- 13
- Diffusion Models Need Visual Priors for Image Generation13
This work proposes Diffusion on Diffusion (DoD), an innovative multi-stage generation framework that first extracts visual priors from previously generated samples, then provides rich guidance for the diffusion model leveraging visual priors from the early stages of diffusion sampling.
- Holistic-Motion2D: Scalable Whole-body Human Motion Generation in 2D Space13
This paper presents Holistic-Motion2D, the first comprehensive and large-scale benchmark for 2D whole-body motion generation, which includes over 1M in-the-wild motion sequences, each paired with high-quality whole-body/partial pose annotations and textual descriptions.
- 13
- Category-Specific Nuance Exploration Network for Fine-Grained Object Retrieval13
An end-to-end Category-specific Nuance Exploration Network (CNENet) that elaborately discovers category-specific nuances that contribute to category prediction, and semantically aligns these nuances grouped by subcategory without any additional prior knowledge, to directly emphasize the discrepancy among subcategories is proposed.
- 13
- GigaGS: 3D Gaussian Based Planar Representation for Large-Scene Surface Reconstruction12
The first attempt to tackle the challenging task of large-scale scene surface reconstruction for large-scale scenes using 3D Gaussian Splatting, and proposes GigaGS, the first work for high-quality surface reconstruction for large-scale scenes using 3DGS.
- Semi-supervised 3D Object Detection with PatchTeacher and PillarMix12
This paper proposes PatchTeacher, a semi-supervised 3D object detection method that focuses on partial scene 3D object detection to provide high-quality pseudo labels for the student, and introduces three key techniques, i.e., Patch Normalizer, Quadrant Align, and Fovea Selection, to improve the performance of PatchTeacher.
- 12
- Adaptive pessimism via target Q-value for offline reinforcement learning12
This paper designs Adaptive Pessimism via a Target Q-value (APTQ) algorithm that balances between the pessimism constraint and the RL objective, and leads the expectation of Q-value to stably converge to a given target Q-value from a reasonable quantile of the Q-value distribution of the dataset.
- 12
- Understanding Masked Autoencoders From a Local Contrastive Perspective12
A local perspective is proposed to explicitly extract a local contrastive form from MAE's reconstructive objective at the patch level and a new empirical framework is introduced, called Local Contrastive MAE (LC-MAE), to analyze both reconstructive and contrastive aspects of MAE.
- 3D Point Cloud Pre-training with Knowledge Distillation from 2D Images12
This work introduces a cross-attention mechanism to extract concept features from 3D point cloud and compare them with the semantic information from 2D images to acquire knowledge directly from the 2D representation learning model, particularly the image encoder of CLIP.
- A Strip Dilated Convolutional Network for Semantic Segmentation12
The method is conducive to extracting multi-scale strip objects that often appear in segmentation scenes, and using strip dilated convolution to further extract contextual dependencies in other directions, so that scene understanding could benefit from long-range dependencies.
- Efficient Joint-Dimensional Search with Solution Space Regularization for Real-Time Semantic Segmentation12
A novel Solution Space Regularization (SSR) loss is first proposed to effectively encourage the supernet to converge to its discrete one and a new Hierarchical and Progressive Solution Space Shrinking method is presented to further achieve high efficiency of searching.
- DETR for Pedestrian Detection.12
The underlying motives driving ED's poor performance are identified, a mechanism to leverage the less occluded visible parts of pedestrian specifically for ED is designed, and a faster bipartite match algorithm is introduced to make ED training on crowd dataset more practical.
- 12
- 12
- CMPhysBench: A Benchmark for Evaluating Large Language Models in Condensed Matter Physics11
This work introduces CMPhysBench, designed to assess the proficiency of Large Language Models in Condensed Matter Physics, as a novel Benchmark, and introduces the Scalable Expression Edit Distance (SEED) score, which provides fine-grained partial credit and yields a more accurate assessment of similarity between prediction and ground-truth.
- Agentic Jigsaw Interaction Learning for Enhancing Visual Perception and Reasoning in Vision-Language Models11
Experimental results show that AGILE not only substantially boosts performance on jigsaw tasks of varying complexity but also demonstrates strong generalization across 9 general vision tasks, achieving an average improvement of 3.1%.
- 11
- Attribute assisted teacher-critical training strategies for image captioning11
A teacher model is introduced that serves as a bridge between the ground-truth caption and the caption model by generating some easier-to-learn word proposals as soft targets to facilitate better learning processes for the Caption model.
- Climate science data can be compressed efficiently by dual-stage extreme compression with a variational auto-encoder transformer10
A 470-fold super compression of atmospheric data has been achieved through Aeolus, a two-stage deep-learning-based variational autoencoder framework that dramatically reduces the size of weather datasets, compressing the widely used 400-terabyte reanalysis atmospheric dataset to just 0.85 TeraBytes (TB).
- $\beta $-DARTS++: Bi-Level Regularization for Proxy-Robust Differentiable Architecture Search10
Beta-Decay regularization can impose constraints to keep the value and variance of activated architecture parameters from being too large, thereby ensuring fair competition among architecture parameters and making the supernet less sensitive to the impact of input on the operation set.
- Stimulative Training++: Go Beyond the Performance Limits of Residual Networks10
This work proposes a novel training scheme called stimulative training, which randomly samples a residual subnetwork and calculates the KL divergence loss between the sampled subnetwork and the given residual network for extra supervision, and proposes three improved strategies for boosting residual networks beyond their performance limits.
- 10
- 10
- Learning from pseudo-labels: deep networks improve consistency in longitudinal brain volume estimation10
DeepBVC is a deep learning powered brain volume change estimation method for assessing brain atrophy that demonstrates superior performance in reproducibility and in the context of common clinical scan variances such as imaging contrast, voxel resolution, random bias field, and signal-to-noise ratio.
- 10
- 10
- Predicting badminton outcomes through machine learning and technical action frequencies9
A predictive model based on male data exhibiting higher accuracy and predictive values, surpassing the performance of the female data model is developed, further leveraging the machine learning model’s capabilities to support tactical decision-making before badminton competitions.
- Instruct-ReID++: Towards Universal Purpose Instruction-Guided Person Re-Identification9
This paper introduces a novel instruct-ReID task that unifies 6 existing ReID tasks in one model and retrieves images based on provided visual or textual instructions and proposes a novel baseline model, IRM, with an adaptive triplet loss to handle various retrieval tasks within a unified framework.
- Advances in 3D pre-training and downstream tasks: a survey9
This paper provides a comprehensive survey of recent advances in the 3D feld, which encompasses a wide collection of topics, including diverse pre-training strategies, backbone designs and downstream tasks.
- 9
- Observation-Guided Meteorological Field Downscaling at Station Scale: A Benchmark and a New Method9
This paper extends meteorological downscaling to arbitrary scattered station scales, establish a brand new benchmark and dataset, and retrieve meteorological states at any given station location from a coarse-resolution meteorological field using a new downscaling model based on hypernetwork architecture, namely HyperDS.
- 9
- 9
- 9
- Hierarchical Graph Convolutional Network for Skeleton-Based Action Recognition9
A novel Hierarchical Graph Convolutional Network (HiGCN) is proposed to deal with hierarchical information of human actions, which is differentiable and can be plugged into GCN in an end-to-end manner.
- DispFormer: A Pretrained Transformer Incorporating Physical Constraints for Dispersion Curve Inversion8
Results show that even without labeled data, the zero‐shot DispFormer generates inversion profiles that outperform the interpolated reference model used as the pretraining target, providing a deployable initial model generator to assist traditional workflows and demonstrate the advantages of integrating physics‐informed deep learning into geophysical applications.
- DATAP-SfM: Dynamic-Aware Tracking Any Point for Robust Structure from Motion in the Wild8
This paper proposes a dynamic-aware tracking any point (DATAP) method that leverages consistent video depth and point tracking and achieves state-of-the-art performance in terms of camera pose estimation even in complex dynamic challenge scenes.
- 8
- 8
- 8
- 8
- Multi-Modal Latent Variables for Cross-Individual Primary Visual Cortex Modeling and Analysis7
A multi-modal identifiable variational autoencoder that employs a two-level disentanglement strategy to map neural activity and visual stimuli into a unified latent space and a novel score-based attribution analysis that traces latent variables back to their origins in the source data space is presented.
- A Self-Evolving AI Agent System for Climate Science7
EarthLink, the first AI agent designed as an interactive copilot for Earth scientists, automates the end-to-end research workflow, from planning and code generation to multi-scenario analysis and enables scientists to shift from laborious manual execution to strategic oversight and hypothesis generation.
- ROGRAG: A Robustly Optimized GraphRAG Framework7
This work proposes a multi-stage retrieval mechanism that integrates dual-level with logic form retrieval methods to improve retrieval robustness without increasing computational cost and introduces ROGRAG, a Robustly Optimized GraphRAG framework.
- 7
- 7
- CMT: A Cascade MAR with Topology Predictor for Multimodal Conditional CAD Generation7
A cascade MAR with topology predictor (CMT) with topology predictor (CMT) is proposed, the first multimodal framework for CAD generation based on Boundary Representation (B-Rep), which can effectively capture the “edge-counters-surface” priors that are essential in B-Reps.
- AutoMat: Enabling Automated Crystal Structure Reconstruction from Microscopy via Agentic Tool Use7
AutoMat, a failure-aware agentic controller that performs inference-time hypothesis search with closed-loop verification to convert Scanning Transmission Electron Microscopy images into simulation-ready crystal structures and downstream properties, establishes a direct pathway from microscopic characterization to atomic-scale modeling, addressing a fundamental challenge in materials science.
- 7
- Achieving Energetic Superiority Through System-Level Quantum Circuit Simulation7
A groundbreaking large-scale system technology that leverages optimization on global, node, and device levels to achieve unprecedented scalability for tensor networks, which enables the handling of large-scale tensor networks with memory capacities reaching tens of terabytes, surpassing memory space constraints on a single node.
- Efficient Architecture Search via Bi-Level Data Pruning7
This work introduces a progressive bi-level data pruning strategy that utilizes supernet prediction dynamics as the metric to gradually prune unsuitable samples for DARTS during the search, and proposes a novel Bi-level Data Pruning (BDP) paradigm that targets the weights and architecture levels of DARTS to enhance efficiency from a data perspective.
- I-Nema: a large-scale microscopic image dataset for nematode recognition7
A large-scale microscopic image dataset consisting of 9215 images and 40 species and set up a species recognition benchmark by employing state-of-the-art deep learning networks on this dataset, which, to the knowledge, is the first time in the community.
- Adaptive Hierarchical SpatioTemporal Network for Traffic Forecasting7
An Adaptive Hierarchical SpatioTemporal Network (AHSTN) is proposed to promote traffic forecasting by exploiting the spatial hierarchy and modeling multi-scale spatial correlations.
- 7
- UniG3D: A Unified 3D Object Generation Dataset7
UniG3D, a unified 3D object generation dataset constructed by employing a universal data transformation pipeline on Objaverse and ShapeNet datasets, which converts each raw 3D model into comprehensive multi-modal data representation by employing rendering engines and multi- modal models.
- 7
- LLaVA-RadZ: Can Multimodal Large Language Models Effectively Tackle Zero-shot Radiology Recognition?6
The LLaVA-RadZ is proposed, a simple yet effective framework for zero-shot medical disease recognition via utilizing the existing MLLM features, and an end-to-end training strategy, termed Decoding-Side Feature Alignment Training (DFAT) to take advantage of the characteristics of the MLLM decoder architecture and incorporate modality-specific tokens tailored for different modalities.
- PhysicsMinions: Winning Gold Medals in the Latest Physics Olympiads with a Coevolutionary Multimodal Multi-Agent System6
Evaluated on the HiPhO benchmark spanning 7 latest physics Olympiads, PhysicsMinions delivers three major breakthroughs: strong generalization, strong generalization, and scaling to human expert.
- 6
- GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation6
GraphGen is introduced, a knowledge graph-guided framework designed for three key question-answering (QA) scenarios: atomic QA, aggregated QA, and multi-hop QA, which incorporates multi-hop neighborhood sampling to capture complex relational information and employs style-controlled generation to diversify the resulting QA data.
- Faithful Contouring: Near-Lossless 3D Voxel Representation Free from Iso-surface6
Faithful Contouring is proposed, a sparse voxelized representation that supports 2048+ resolutions for arbitrary meshes, requiring neither converting meshes to field functions nor extracting the isosurface during remeshing, achieving near-lossless fidelity by preserving sharpness and internal structures.
- 6
- 6
- Where Am I and What Will I See: An Auto-Regressive Model for Spatial Localization and View Prediction6
Generative Spatial Transformer is presented, a novel auto-regressive framework that jointly addresses spatial localization and view prediction that simultaneously estimates the camera pose from a single image and predicts the view from a new camera pose, effectively bridging the gap between spatial awareness and visual prediction.
- 6
- MaskMA: Towards Zero-Shot Multi-Agent Decision Making with Mask-Based Collaborative Learning6
This work proposes a framework for masking strategies, collaborative learning strategy, and generalizable action representation that enables the model to learn effective representations and strategies by capturing the underlying correlations among agents and their actions while maintaining adaptability to dynamic scenarios.
- Unifying Visual Contrastive Learning for Object Recognition from a Graph Perspective6
This paper proposes to Unify existing unsupervised Visual Contrastive Learning methods by using a GCN layer as the predictor layer (UniVCL), which deserves two merits to un supervised learning in object recognition.
- 6
- 6
- Once Quantized for All: Progressively Searching for Quantized Efficient Models6
Once Quantized for All (OQA) is presented, a novel framework that searches for quantized efficient models and deploys their quantized weights at the same time without additional post-process without affecting the quality of the search-retrain schema.
- 6
- A surrogate model based on parametric neural network solvers for laminar flows around aerofoils5
This paper presents a solver for laminar flow around aerofoils based on TSONN, validated across various test cases, and extends the solver to parametric problems involving flow conditions and aerofoils shapes, covering nearly all laminar flow scenarios in engineering.
- UniSTD: Towards Unified Spatio-Temporal Learning across Diverse Disciplines5
UniSTD is introduced, a unified Transformer-based framework for spatiotemporal modeling, inspired by advances in recent foundation models with the two-stage pretraining-then-adaption paradigm, which demonstrates that task-agnostic pretraining on 2D vision and vision-text datasets can build a generalizable model foundation for spatiotemporal learning.
- AI-Driven Automation Can Become the Foundation of Next-Era Science of Science Research5
A preliminary multi-agent system is presented as an illustrative example to simulate research societies, showcasing AI's ability to replicate real-world research patterns and accelerate progress in Science of Science research.
- 5
- 5
- 5
- 5
- SeCo: Separating Unknown Musical Visual Sounds with Consistency Guidance5
This work proposes the "Separation-with-Consistency" (SeCo) framework, which can accomplish the separation on unknown categories by exploiting the consistency constraints, and devise an online matching strategy which can bring stable enhancements with no cost of extra parameters.
- 5
- 5
- MS Lesion Segmentation: Revisiting Weighting Mechanisms for Federated Learning5
Comparison experiments on two FL MS segmentation scenarios using public and clinical datasets have demonstrated the effectiveness of the proposed method by outperforming other FL methods significantly and indicated the superiority of the method when estimating brain volume differences estimation after lesion inpainting.
- An Empirical Study of Pseudo-Labeling for Image-based 3D Object Detection5
This work conducts extensive experiments to investigate whether the pseudo-labels can provide effective supervision for the baseline models under varying settings and explores the feasibility of a cheaper alternative, i.e. pseudo-labeling, to leverage the unlabeled data.
- Efficient Visual Recognition5
Although visual recognition has made significant progress, especially in the past several years, there is continued need for vigorous research to solve many challenging problems towards highly efficient visual recognition including achieving energy efficiency and label/ sample efficiency.
- Perceptual Image Enhancement by Relativistic Discriminant Learning With Cross-Scale Aggregated Representation5
An automatic method for image quality enhancement that relies on a multi-level connected generator and a pair-wise relativistic discriminator and can automatically enhance the low-quality image into a digital single lens reflex (DSLR) quality.
- 5
- 5
- Position: Intelligent Science Laboratory Requires the Integration of Cognitive and Embodied AI4
This position paper proposes the paradigm of Intelligent Science Laboratories (ISLs): a multi-layered, closed-loop framework that deeply integrates cognitive and embodied intelligence that unify foundation models for scientific reasoning, agent-based workflow orchestration, and embodied agents for robust physical experimentation.
- $$\hbox {I}^2$$MD: 3D Action Representation Learning with Inter- and Intra-Modal Mutual Distillation4
This work introduces a general Inter- and intra-modal mutual distillation framework, and designs the intra-modal mutual distillation (IMD) strategy, which adaptively aggregates highly-correlated neighboring features, forming local cluster-level contrasting.
- 4
- EWE: An Agentic Framework for Extreme Weather Analysis4
This work presents the Extreme Weather Expert (EWE), the first intelligent agent framework dedicated to this task, which emulates expert workflows through knowledge-guided planning, closed-loop reasoning, and a domain-tailored meteorological toolkit.
- ReconMOST: Multi-Layer Sea Temperature Reconstruction with Observations-Guided Diffusion4
This paper first pre-train an unconditional diffusion model using a large collection of historical numerical simulation data, enabling the model to attain physically consistent distribution patterns of ocean temperature fields, and extends ML-based SST reconstruction to a global, multi-layer setting.
- Reflections Unlock: Geometry-Aware Reflection Disentanglement in 3D Gaussian Splatting for Photorealistic Scenes Rendering4
Ref-Unlock is proposed, a novel geometry-aware reflection modeling framework based on 3D Gaussian Splatting, which explicitly disentangles transmitted and reflected components to better capture complex reflections and enhance geometric consistency in real-world scenes.
- Manalyzer: End-to-end Automated Meta-analysis with Multi-agent System4
A multi-agent system, Manalyzer, is proposed, which achieves end-to-end automated meta-analysis through tool calls, and achieves significant performance improvements over the LLM baseline in multi meta-analysis tasks.
- Genomics-FM: Universal Foundation Model for Versatile and Data-Efficient Functional Genomic Analysis4
Genomics-FM is developed, a genomic vocabulary driven foundation model that enables versatile and label-efficient functional genomic analysis and demonstrates the capability to outperform existing models across a comprehensive suite of tasks including genome annotation, epigenomic and expression profile prediction, and variant effect assessment.
- DocReLM: Mastering Document Retrieval with Language Model4
This work demonstrates that by utilizing large language models, a document retrieval system can achieve advanced semantic understanding capabilities, significantly outperforming existing systems.
- Identifying featured indels associated with SARS-CoV-2 fitness4
Evaluation of the growth rate of each viral lineage indicated that indels play key roles in SARS-CoV-2 evolution and deserve more attention as substitution mutations, and machine learning model based on Bayesian inference and viral lineage growth rate evaluation suggests that these featured indels exhibit selection pressure comparable to replacement mutations.
- Boosting Semi-Supervised 3D Object Detection with Semi-Sampling4
This paper designs a data augmentation method for semi-supervised learning called Semi-Sampling, which uses ground truth labels and pseudo labels to crop gt samples and pseudo samples on labeled frames and unlabeled frames, respectively, and generates a gt sample database and a pseudo sample database.
- Adversarial learning based intermediate feature refinement for semantic segmentation4
This paper proposes a novel intermediate feature refinement method for semantic segmentation based on adversarial learning, which reduces the error and redundant information contained in the teacher network in the process of knowledge distillation, enhances the correct informationcontained in the teachers network and transfers it to the student network.
- 4
- 4
- 4
- Using Age Information as a Soft Biometric Trait for Face Image Analysis4
This chapter introduces the concepts and theory behind the three main areas of age-related face image analysis and how they can be used in practical biometric applications, and analyses the difficulties involved in these applications.
- IntersectGAN4
This paper proposes a novel GAN, namely IntersectGAN, to learn multiple attributes from different image domains through an intersecting architecture, and using the CelebFaces Attributes dataset, is able to produce high quality face images possessing multiple attributes.
- 3
- ARCHE: A Novel Task to Evaluate LLMs on Latent Reasoning Chain Extraction3
A novel task named Latent Reasoning Chain Extraction (ARCHE), in which models must decompose complex reasoning arguments into combinations of standard reasoning paradigms in the form of a Reasoning Logic Tree (RLT), and proposes two logic-aware evaluation metrics.
- ShapeGen: Towards High-Quality 3D Shape Synthesis3
ShapeGen achieves high-quality image-to-3D shape generation through 3D representation and supervision improvements, resolution scaling up, and the advantages of linear transformers, and introduces an inference-time scaling strategy that enhances generation quality at test time.
- scDNAm-GPT Captures Genome-wide CpG Dependencies in Single-cell DNA methylomes to Revolutionize Epigenetic Analysis3
The results establish scDNAm-GPT as a scalable and generalizable framework for single-cell epigenomic analysis, offering new opportunities to dissect epigenetic regulation in health and disease.
- ChemBOMAS: Accelerated BO in Chemistry with LLM-Enhanced Multi-Agent System3
ChemBOMAS is introduced: a large language model (LLM)-enhanced multi-agent system that accelerates BO through synergistic data- and knowledge-driven strategies, accelerating optimization efficiency by up to 5-fold compared to baseline methods.
- 3
- 3
- 3
- High Performance Space Debris Tracking in Complex Skylight Backgrounds with a Large-Scale Dataset3
A deep learning-based Space Debris Tracking Network~(SDT-Net) to achieve highly accurate debris tracking and enhances the efficiency and stability of end-to-end model learning.
- 3
- 3
- How Much Data are Enough? Investigating Dataset Requirements for Patch-Based Brain MRI Segmentation Tasks3
This paper proposes a strategic framework for estimating the amount of annotated data required to train patch-based segmentation networks, and demonstrates that tasks involving regions of interest (ROIs) with different sizes or shapes may yield variably acceptable Dice Similarity Coefficient (DSC) scores.
- Physical formula enhanced multi-task learning for pharmacokinetics prediction3
This work develops a physical formula enhanced mul-ti-task learning (PEMAL) method that predicts four key parameters of pharmacokinetics simultaneously, and demonstrates the benefits and potential of using PEMAL in AIDD and other scenarios with data scarcity and noise.
- 3
- 3
- 3
- 3
- Location-Aware Feature Selection for Scene Text Detection3
A novel method called Location-Aware Feature Selection (LAFS) is proposed, which separately learns the confidence of different locations' features for each component and then selects the features with the highest confidence to form a combination of the most suitable features.
- 3
- Earth-o1: A Grid-free Observation-native Atmospheric World Model2
The results establish that continuous, observation-driven world models -- a new class of fully observation-native geophysical simulators -- can match the fidelity of established physical frameworks, providing a scalable data-driven foundation for a digital twin of the Earth.
- MEMTS: Internalizing Domain Knowledge via Parameterized Memory for Retrieval-Free Domain Adaptation of Time Series Foundation Models2
MEMTS is proposed, a lightweight and plug-and-play method for retrieval-free domain adaptation in time series forecasting that enables MEMTS to achieve accurate domain adaptation with constant-time inference and near-zero latency, while effectively mitigating catastrophic forgetting of general temporal patterns.
- SynCast: Synergizing Contradictions in Precipitation Nowcasting via Diffusion Sequential Preference Optimization2
This work introduces preference optimization into precipitation nowcasting for the first time, motivated by the success of reinforcement learning from human feedback in large language models and proposes SynCast, which leverages the two-stage post-training framework of Diffusion Sequential Preference Optimization (Diffusion-SPO) to progressively align conflicting metrics.
- 2
- 2
- Dynamic Base model Shift for Delta Compression2
This work proposes Dynamic Base Model Shift (DBMS), which dynamically adapts the base model to the target task before performing delta compression, and adjusts two parameters, which respectively determine the magnitude of the base model shift and the overall scale of delta compression, to boost the compression performance on each task.
- 2
- CTTS: Collective Test-Time Scaling2
Evaluations across seven mainstream benchmarks demonstrate that CTTS-MM significantly outperforms leading STTS methods and surpasses even flagship proprietary LLMs and open-source LLMs, highlighting the substantial potential of collective scaling to push the frontier of LLM inference.
- Multimodal-Guided Dynamic Dataset Pruning for Robust and Efficient Data-Centric Learning2
This work introduces a dynamic dataset pruning framework that adaptively selects training samples based on both task-driven difficulty and cross-modality semantic consistency, highlighting the potential of integrating cross-modality alignment for robust sample selection.
- 2
- 2
- A renaissance of explicit motion information mining from transformers for action recognition2
In EMIM, the desirable affinity matrix is constructed in a cost volume style, where the set of key candidate tokens is sampled from the query-based neighboring area in the next frame in a sliding-window manner and is used to aggregate contextual information for appearance modeling and is converted into motion features for motion modeling as well.
- 2
- 2
- Fast Information Streaming Handler (FisH): A Unified Seismic Neural Network for Single Station Real-Time Earthquake Early Warning2
FisH is designed to process real-time streaming seismic data and generate simultaneous results for phase picking, location estimation, and magnitude estimation in an end-to-end fashion and simplifies the overall process and leverages the nonlinear relationships between tasks for improved performance.
- 2
- Self-consistent Validation for Machine Learning Electronic Structure2
This method integrates machine learning with self-consistent field methods to achieve both low validation cost and interpret-ability and enables exploration of the model's ability with active learning and instills confidence in its integration into real-world studies.
- 2
- Saliency Guided Contrastive Learning on Scene Images2
The saliency map derived from the model's output during learning is leveraged to highlight these discriminative regions for contrastive learning to object representations in scene images and guide the whole Contrastive learning.
- Learning class-agnostic masks with cross-task refinement for weakly supervised semantic segmentation2
This work proposes to jointly learn semantic segmentation and class-agnostic masks by using image-level annotations and off-the-shelf saliency maps as supervision, and introduces a new normalization method for CAM to generate more complete class-specific localization maps.
- Towards Fair and Comprehensive Comparisons for Image-Based 3D Object Detection2
This work builds a modular-designed codebase, formulate strong training recipes, design an error diagnosis toolbox, and discuss current methods for image-based 3D object detection to facilitate future research in image-based 3D object detection.
- Relative Contrastive Loss for Unsupervised Representation Learning2
The Relative Contrastive Loss (RCL) is proposed to learn feature representation from relatively positive/negative pairs, which not only learns more real world semantic variations than the single-instance-positive methods but also respects positive-negative relativeness compared with absolute prototype-positive methods.
- 2
- 2
- 2
- High-Performance Light Field Reconstruction with Channel-wise and SAI-wise Attention2
Channel-wise and SAI-wise attention modules are proposed to enhance the feature representation at a low cost and demonstrate that the baseline network can achieve better performance with the aid of the attention modules.
- 2
- 1
- 1
- Control-R: Towards controllable test-time scaling1
This work introduces a novel test-time approach that injects structured control signals to guide reasoning from a tree search perspective that achieves state-of-the-art performance at the 32B scale while enabling a controllable Long CoT reasoning process (L-CoT).
- 1
- 1
- 1
- 1
- Point Transformer V3 Extreme: 1st Place Solution for 2024 Waymo Open Dataset Challenge in Semantic Segmentation1
This technical report significantly enhanced the performance of Point Transformer V3 on the Waymo benchmark by implementing cutting-edge, plug-and-play training and inference technologies, achieving substantial gains over the original PTv3 performance.
- UniSPAC: A Unified Segmentation Framework for Proofreading and Annotation in Connectomics1
A Unified Segmentation framework for Proofreading and Annotation in Connectomics (UniSPAC) is introduced by providing the interactive segmentation model in 2D-level and the neuron tracing model in 3D-level, combining segmentation and proofreading in a single framework.
- 1
- 1
- 1
- 1
- 1
- Automatically Predict Material Properties with Microscopic Image Example Polymer Compatibility1
This work achieves automatic miscibility recognition utilizing convolution neural network and transfer learning method and the model obtains up to 94% accuracy, and puts forward a quantitative criterion for polymer miscibility with this model.
- 1
- 1
- 1
- 1
- Learning Deep Representations for Scene Labeling with Guided Supervision1
It is argued that the large intra-class variation provides ambiguous training information and hinders the deep models' ability to learn more discriminative deep feature representations.
- Adaptive Low Resolution Pruning for fast Full Search-equivalent pattern matching1
This paper proposes a computational analysis of several incremental techniques specifically designed to enhance the efficiency of Low Resolution Pruning and proposes a novel LRP algorithm aimed at minimizing the theoretical number of operations by adaptively exploiting different incremental approaches.
- 1
- –
- Chem-R: Learning to Reason as a Chemist–
The proposed Chem-R, a general Chemical Reasoning model designed to emulate the deliberative processes of chemists, achieves state-of-the-art performance on comprehensive benchmarks, surpassing leading LLMs, including Gemini-3-Pro and Kimi-k2.5.
- Improving protein and protein interactions using pseudo-dimers derived from monomeric proteins–
This work introduces SMP, a strategy that leverages pseudo-dimers derived from monomers to improve accuracy and generalization across diverse protein interaction applications, and highlights SMP as a scalable strategy for harnessing monomeric data to advance protein complex modeling.
- –
- Generative Diffusion Prior for Unified Image and Video Restoration & Enhancement–
Experimental results demonstrate that GDP outperforms leading unsupervised methods across diverse benchmarks in both reconstruction accuracy and perceptual quality, while demonstrating robust generalization to images and videos of any size.
- Naturalistic Social Dyads Assessment In Free Play: A Multi-view Framework For Recognizing Co-located Eye Contact Between Children and Their Assessors–
It is suggested that multi-view video analysis can objectively identify clinically defined eye contact behaviors among multiple individuals during clinical free play assessments, and may support scalable behavioral coding and quantitative assessment of clinically relevant social behaviors consistent with established criteria.
- –
- –
- CENO: A Genome-Scale World Model for Evolutionary Sequence Interpretation and Programmable Regulatory Design–
CENO is introduced, a family of long-context generative genomic world models designed to preserve local DNA grammar while extending usable context to regulatory and chromatin scales and provides a genome-scale sequence world-model framework for sequence interpretation, evolutionary reasoning, gene-scale reconstruction and programmable regulatory sequence generation.
- –
- –
- A Scalable Multi-LLM Collaboration System with Retrieval-based Selection and Exploration-Exploitation-Driven Enhancement–
By integrating fifteen open-source LLMs, SMCS outperforms prevailing closed-source LLMs and exceeds the average of best results on different datasets with open-source LLMs, significantly advancing the empirical performance frontier of open-source collaboration.
- –
- –
- –
- scMOBA: A conversational single-cell Multi-Omics Brain Agent across species–
ScMOBA, a conversational single-cell Multi-Omics Brain Agent established by a large language model, a gene encoder, and a cross-attention projector achieves state-of-the-art performance in fine-grained cell type classification across different species and modalities, as well as in batch correction and multi-omics data integration.
- –
- –
- –
- –
- –
- –
- –
- –
- Fitness aligned structural modeling enables scalable virtual screening with AuroBind–
AuroBind offers a generalizable framework for structure-function learning and high-throughput molecular screening, bridging the gap between structure prediction and therapeutic discovery.
- –
- Revealing Decurve Flows for Generalized Graph Propagation–
This study addresses the limitations of the traditional analysis of message-passing, central to graph learning, by defining Generalized Propagation Neural Networks (GPNNs), a framework that unifies most propagation-based graph neural networks.
- –
- –
- –
- EMCLR '24: 1st International Workshop on Efficient Multimedia Computing under Limited Resources–
This workshop aims to bring together experts in data-efficient, label-efficient, and computation-efficient multimedia computing fields, establishing a collaborative platform to exchange recent breakthroughs and deliberate on the future direction of multimedia computing models.
- A novel federated learning framework for accurate and secure multi-center MS lesion segmentation–
A decentralized deep learning framework is designed in this work, which can bring accurate multi-center MS lesion segmentation performance without sharing the raw data.
- –
- –
- –
- –
- –
- –
- Fast pattern matching and its applications–
This thesis proposes a transform domain SVM (TDSVM) using pruning that computes SVM much faster and shows the efficiency in applying the proposed method for human detection.
- –
- An Assembly Optimization Algorithm of Scan Quantization and FDCT Based on VLIW–
It is shown that 60% of the population believes in reincarnation, while the rest have no opinion.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.