AAAI 2026 Papers — Page 35
AAAI Conference on Artificial Intelligence · 4149 papers
SigFusion: Unified Signal-Level Self-Supervised Learning Paradigm for Image Fusion
Zeyu Wang (Dalian Minzu University), Haiyu Song (Dalian Minzu University)
Image TranslationRestorationTransformerImageMultimodalityMagnetic Resonance ImagingComputed TomographyPositron Emission Tomography
🎯 What it does: Propose SigFusion, a unified signal-level self-supervised learning framework for multi-task image fusion.
Sign-Aware Multimodal Graph Recommendation
Yahong Lian (Nankai University), Tingjian Ge (University of Massachusetts Lowell)
Recommendation SystemGraph Neural NetworkContrastive LearningTextMultimodality
🎯 What it does: Proposes the Sign-Aware Multimodal Graph Recommendation (SiMGR) framework, which integrates multimodal features with a signed user-item interaction graph and uses positive and negative thresholds to separate interactions and enhance with pseudo edges.
Signal Enhancement via Multi-view Dynamic Representation and Alignment-aware Fusion
Zikun Jin (Shanxi University), Honghong Cheng (Shanxi University)
RestorationTime SeriesAudio
🎯 What it does: This paper proposes the FracFusion framework for speech and electromagnetic signal enhancement under non-stationary, low signal-to-noise ratio conditions.
Signal: Selective Interaction and Global-local Alignment for Multi-Modal Object Re-Identification
Yangyang Liu (Dalian University of Technology), Pingping Zhang (Dalian University of Technology)
RecognitionTransformerContrastive LearningMultimodality
🎯 What it does: Propose the Signal framework, which first uses SIM to select important patches from multi-modal images, then achieves global multi-modal alignment in the Gramian space via GAM, and finally performs pixel-level local alignment using the learnable offsets of LAM, thereby enhancing feature representation in multi-modal ReID.
Sim-to-Real: An Unsupervised Noise Layer for Screen-Camera Watermarking Robustness
Yufeng Wu (Hunan University), Guiling Wang (Hunan University)
Convolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Proposed a two-step screen-camera (SC) watermark noise approximation framework named Simulation-to-Real (S2R), which first generates rough noise through mathematical modeling and then refines the approximation of real SC noise using an unsupervised image-to-image network, thereby enhancing the robustness of watermarks in real-world scenarios.
Sim4Seg: Boosting Multimodal Multi-disease Medical Diagnosis Segmentation with Region-Aware Vision-Language Similarity Masks
Lingran Song (University of Macau), Jianbing Shen (University of Macau)
SegmentationExplainability and InterpretabilityTransformerLarge Language ModelMultimodalityBiomedical DataUltrasoundBenchmarkChain-of-Thought
🎯 What it does: This paper proposes the medical diagnosis segmentation (MDS) task and constructs a multimodal, multi-disease medical image segmentation and diagnostic dataset, M3DS. It introduces the Sim4Seg framework and the Region-Aware Vision-Language Similarity to Mask (RVLS2M) module, achieving segmentation and diagnosis of multimodal medical images while providing interpretable diagnostic chains.
Simba: Towards High-Fidelity and Geometrically-Consistent Point Cloud Completion via Transformation Diffusion
Lirui Zhang (Nanjing University of Aeronautics and Astronautics), Jie Qin (Nanjing University of Aeronautics and Astronautics)
RestorationGenerationDiffusion modelPoint Cloud
🎯 What it does: Propose the Simba framework, transforming the point cloud completion task into learning the distribution of geometric transformations to achieve high-fidelity geometric consistency.
SimDiff: Simpler Yet Better Diffusion Model for Time Series Point Forecasting
Hang Ding (Shanghai Jiao Tong University), Tao Yao (Shanghai Jiao Tong University)
TransformerDiffusion modelTime Series
🎯 What it does: Proposed SimDiff, a single-stage end-to-end diffusion model for time series point prediction, completely eliminating pre-trained regressors;
SimLabel: Similarity-Weighted Semi-supervision for Multi-annotator Learning with Missing Labels
Liyun Zhang, Yuta Nakashima (Cincinnati Children's Hospital Medical Center)
ClassificationRecognitionVideoMultimodality
🎯 What it does: Propose a similarity-weighted semi-supervised framework called SimLabel to address the problem of missing labels in multi-annotator learning;
SimpleDiffusion: A Lightweight and Efficient Conditional Diffusion Model for Multi-Modal Salient Object Detection
Shuo Zhang (Shanghai Dianji University), Zizhu Fan (Shanghai Ocean University)
Object DetectionTransformerDiffusion modelMultimodality
🎯 What it does: Proposed and implemented a lightweight conditional diffusion model named SimpleDiffusion for multimodal salient object detection
SimROD: A Simple Baseline for Raw Object Detection with Global and Local Enhancements
Haiyang Xie (Wuhan University), Zheng Wang (Wuhan University)
Object DetectionImage
🎯 What it does: This paper proposes the SimROD method, which directly uses RAW images for object detection, eliminating the ISP process.
Simulated Rewards, Skewed Strategies: Tracing the Acquired Preference Bias in LLM-Based Dialogue Planners
Heyan Huang (Beijing Institute of Technology), Yang Gao (Beijing Institute of Technology)
TransformerLarge Language ModelText
🎯 What it does: This paper systematically evaluates preference bias and strategy diversity of four LLM-driven dialogue planners across three task categories (price negotiation, emotional support, and charitable persuasion), while tracking the evolution of bias during training and associated ethical risks.
Simulating Dispute Mediation with LLM-Based Agents for Legal Research
Junjie Chen (Tsinghua University), Qingyao Ai (Tsinghua University)
TransformerLarge Language ModelAgentic AITextRetrieval-Augmented Generation
🎯 What it does: Developed the AgentMediation framework, utilizing LLM agents to simulate real civil dispute mediation processes, generating multi-round dialogues and solutions.
Simulating Distribution Dynamics: Liquid Temporal Feature Evolution for Single-Domain Generalized Object Detection
Zihao Zhang (Tianjin University), Yahong Han (Tianjin University)
Object DetectionDomain AdaptationRecurrent Neural NetworkImageOrdinary Differential Equation
🎯 What it does: Proposed the Liquid Time Feature Evolution (LTFE) method for single-source domain generalization object detection;
Simulating Human-Like Counseling: A Path- and Scenario-Guided Framework for Psychological Support Dialogue
Yuanchen Shi (Soochow University), Fang Kong (Jiangsu University)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningAgentic AIPrompt EngineeringText
🎯 What it does: This paper proposes a path-guided simulation framework, PGSim, to construct a large-scale, structured, and context-rich Chinese psychological support dialogue dataset, CPsDD, and develops a multi-modal intelligent agent system, CADSS, based on this dataset, capable of generating context-controllable psychological counseling dialogues.
Simulation-Driven Railway Delay Prediction: An Imitation Learning Approach
Clément Elliker (École Polytechnique), Albert Bifet (University of Waikato)
OptimizationTransformerReinforcement LearningTabularTime Series
🎯 What it does: Convert train delay prediction into a stochastic simulation task, training strategies through imitation learning to replicate historical state transitions, thereby achieving temporal prediction of delays.
SinBasis Networks: Matrix-Equivalent Feature Extraction for Wave-Like Optical Spectrograms
Yuzhou Zhu (Dalian University of Technology), Liang Zhou (Dalian University of Technology)
ClassificationConvolutional Neural NetworkTransformerImagePhysics Related
🎯 What it does: Propose Sin-Basis Networks, which map the weights of CNN, ViT, Capsule, and other networks to sine basis functions, enhancing the extraction of periodic features in waveform images.
SineLoRA∆: Sine-Activated Delta Compression
Cameron Gordon (Australian Institute for Machine Learning, University of Adelaide), Simon Lucey (Australian Institute for Machine Learning, University of Adelaide)
CompressionComputational EfficiencyTransformerLarge Language ModelVision Language ModelDiffusion modelImageTextMultimodality
🎯 What it does: Propose a SineLoRA ∆ method that employs sine activation on quantized low-rank adapters (LoRA) to achieve differential compression;
Single-Stage fMRI-to-3D Reconstruction via Viewpoint-Aware Embedding and Hierarchical Guidance
Xun Zhang (Delft University of Technology), Pan Wang (Delft University of Technology)
GenerationDiffusion modelAuto EncoderContrastive LearningVideoMeshBiomedical DataMagnetic Resonance Imaging
🎯 What it does: This paper proposes a single-stage end-to-end NeuroSculptor3D framework that directly reconstructs textured three-dimensional objects using fMRI signals.
SITA: A Framework for Structure-to-Instance Theorem Autoformalization
Chenyi Li (Peking University), Zaiwen Wen (Peking University)
OptimizationAI Code AssistantLarge Language ModelTextBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Designed and implemented the SITA framework, achieving automated theorem proving in Lean, including the automatic generation of definitions, instance declarations, and complete proofs from abstract structures to concrete instances.
Skeletons Speak Louder than Text: A Motion-Aware Pretraining Paradigm for Video-Based Person Re-Identification
Rifen Lin (Central South University), Min Li (Central South University)
RetrievalContrastive LearningVideoMultimodality
🎯 What it does: Proposed a new skeleton-based pre-training framework called CSIP-ReID for video person re-identification (ReID), achieving multi-modal pre-training through contrastive learning on skeleton sequences and image frames.
Skill Path: Unveiling Language Skills from Circuit Graphs
Hang Chen (Xi'an Jiaotong University), Wenya Wang (Nanyang Technological University)
Explainability and InterpretabilityTransformerLarge Language ModelText
🎯 What it does: Proposed a three-step framework (decomposition, pruning, causal mediation) to extract skill paths (Skill Path) in language models, enabling finer-grained capture of mechanisms underlying specific language skills;
SkillGen: Learning Domain Skills for In-Context Sequential Decision Making
Ruomeng Ding (University of North Carolina at Chapel Hill), Chen Zhao (Baylor University)
Large Language ModelReinforcement LearningPrompt EngineeringTextSequentialRetrieval-Augmented Generation
🎯 What it does: Propose the SkillGen framework, which constructs an action-centric domain knowledge graph using trajectories sampled from LLMs. High-value actions are extracted through TD assignment to form 'golden segments' and 'step-by-step skills.' During inference, these structured knowledge elements are combined with recent actions to generate more focused, fine-grained, and label-efficient prompts.
SkipCat: Rank-Maximized Low-Rank Compression of Large Language Models via Shared Projection and Block Skipping
Yu-Chen Lu (National Yang Ming Chiao Tung University), Kai-Chiang Wu (Skymizer Taiwan Inc.)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: Proposed a SkipCat low-rank compression framework that combines shared projection (Cat) and block skipping (Skip) techniques to compress large language models without additional training.
SkyMoE: A Vision-Language Foundation Model for Enhancing Geospatial Interpretation with Mixture of Experts
Jiaqi Liu (Jilin University), Bo Yang (Jilin University)
ClassificationRecognitionObject DetectionMixture of ExpertsVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose SkyMoE, a vision-language foundation model integrating Mixture-of-Experts (MoE), specifically designed for multi-task, multi-scale remote sensing interpretation, achieving separation and fusion of local and global features through adaptive routers and context-disentangled data augmentation.
SkySplat: Generalizable 3D Gaussian Splatting from Multi-Temporal Sparse Satellite Images
Xuejun Huang (Wuhan University), Yongjun Zhang (Wuhan University)
Depth EstimationTransformerGaussian SplattingImage
🎯 What it does: Propose a self-supervised, transferable 3D Gaussian splatting framework called SkySplat for rapidly reconstructing accurate DSMs from multi-temporal sparse satellite images;
SLCFormer: Spectral-Local Context Transformer with Physics-Grounded Flare Synthesis for Nighttime Flare Removal
Xiyu Zhu (Wuhan University of Science and Technology), Xiao Wang (Wuhan University of Science and Technology)
RestorationData SynthesisTransformerAuto EncoderImagePhysics Related
🎯 What it does: Proposed the SLCFormer framework for nighttime lens flare removal and developed a scattered flare synthesis pipeline based on ZernikeVAE.
SLD-L2S: Hierarchical Subspace Latent Diffusion for High-Fidelity Lip to Speech Synthesis
Yifan Liang (Chinese Academy of Sciences), Chengshi Zheng (Chinese Academy of Sciences)
GenerationData SynthesisDiffusion modelFlow-based ModelVideoMultimodalityAudio
🎯 What it does: This paper proposes a high-fidelity speech generation framework called SLD-L2S, which directly maps lip visual features to the continuous latent space of a pre-trained audio Codec, avoiding the loss of details that occurs with traditional Mel or SSL semantic representations.
Slender3D: Curve-Guided Multi-View Reconstruction of Slender Structures
Suqin Wang (North China Electric Power University), Dengming Zhu (Institute of Computing Technology, Chinese Academy of Sciences)
GenerationOptimizationGaussian SplattingImageMeshBiomedical Data
🎯 What it does: Proposed a multi-view geometry reconstruction framework called Slender3D tailored for slender structures.
SlideTailor: Personalized Presentation Slide Generation for Scientific Papers
Wenzheng Zeng (National University of Singapore), Hwee Tou Ng (National University of Singapore)
GenerationTransformerLarge Language ModelAgentic AIVision Language ModelTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Designed a personalized paper-to-slides generation system called SlideTailor based on an agent framework.
Sliding-Window Merging for Compacting Patch-Redundant Layers in LLMs
Xuan Ding (Chinese University of Hong Kong (Shenzhen)), Yao Zhu (Zhejiang University)
OptimizationComputational EfficiencyKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed and implemented the Sliding-Window Merging (SWM) method to dynamically merge adjacent functionally similar Transformer layers in large language models (LLMs), thereby compressing model depth without significantly compromising performance.
SlimInfer: Accelerating Long-Context LLM Inference via Dynamic Token Pruning
Lingkun Long (Beihang University), Jianlei Yang (Beihang University)
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Achieve long-context LLM inference acceleration by dynamically performing fine-grained token trimming on hidden states during the forward propagation process.
SM3Det: A Unified Model for Multi-Modal Remote Sensing Object Detection
Yuxuan Li (Nankai University), Jian Yang (Nankai University)
Object DetectionConvolutional Neural NetworkMixture of ExpertsImageMultimodality
🎯 What it does: This paper proposes the M2Det task, which is a unified model for multi-task (horizontal and rotated bounding box) object detection on multi-modal remote sensing data.
Small but Mighty: Dynamic Wavelet Expert-Guided Fine-Tuning of Large-Scale Models for Optical Remote Sensing Object Segmentation
Yanguang Sun (Nanjing University of Science and Technology), Lei Luo (Nankai University)
SegmentationTransformerSupervised Fine-TuningMixture of ExpertsImage
🎯 What it does: Propose a fine-tuning paradigm based on dynamic waveform expert guidance (WEFT), which significantly enhances the performance of optical remote sensing image (ORSI) target segmentation by freezing a large base model (UniPerceiver-L) and introducing a lightweight task-specific waveform expert extractor and expert-guided conditional adapter.
Small Models Exhibit Limited Answer Consistency in Repetition Trials of the Multiple-Choice MMLU-Redux and MedQA Benchmarks
Claudio Pinhanez (IBM Research Brazil), Marcelo Carpinette Grave (IBM Research Brazil)
TransformerLarge Language ModelSupervised Fine-TuningTextBiomedical DataBenchmark
🎯 What it does: Conduct 10 repeated inferences of open-source LLMs (2B–80B) on MMLU-Redux and MedQA multiple-choice questions, evaluating answer consistency and correlating it with accuracy.
SMART: A Surrogate Model for Predicting Application Runtime in Dragonfly Systems
Xin Wang (University of Illinois Chicago), Zhiling Lan (University of Illinois Chicago)
OptimizationComputational EfficiencyGraph Neural NetworkTransformerLarge Language ModelGraphTime Series
🎯 What it does: Proposes the SMART model, integrating graph neural networks (GNN) and large language models (LLM) to predict application iteration time in Longfly network systems, achieving high-accuracy and low-latency proxy simulation.
SmartAgent: Chain-of-User-Thought for Embodied Personalized Agent in Cyber World
Jiaqi Zhang (University of Queensland), Hongzhi Yin (University of Queensland)
Recommendation SystemRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningAgentic AIVision Language ModelVision-Language-Action ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Propose the Chain-of-User-Thought (COUT) embodied personalized reasoning paradigm and implement the SmartAgent framework to achieve GUI operation to user preference reasoning and personalized recommendation in networked environments, while constructing the SmartSpot dataset.
SmartSight: Mitigating Hallucination in Video-LLMs Without Compromising Video Understanding via Temporal Attention Collapse
Yiming Sun (Fudan University), Min Yang (Fudan University)
Explainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelVision Language ModelVideoTextMultimodality
🎯 What it does: Propose a training-free, self-introspection-based method called SmartSight, which evaluates hallucinations using multi-candidate sampling and Temporal Attention Collapse Score, and employs Visual Attention Vanishing Point to preemptively terminate low-quality generation, thereby reducing hallucinations without compromising video understanding and reasoning capabilities.
SmartSplat: Feature-Smart Gaussians for Scalable Compression of Ultra-High-Resolution Images
Linfei Li (Tongji University), Ying Shen (Shanghai Jiao Tong University)
CompressionGaussian SplattingImage
🎯 What it does: Proposed the SmartSplat framework, which utilizes feature-driven 2D Gaussian splines for ultra-high resolution image compression;
SMIDT: High-Performance Inference Framework for MoE Models with Dynamic Top-K Routing
Zewen Jin (University of Science and Technology of China), Cheng Li (University of Science and Technology of China)
OptimizationComputational EfficiencyMixture of ExpertsText
🎯 What it does: Develop the SMIDT framework to integrate dynamic Top-K routing and pipeline parallelism for efficient MoE model inference.
SMoFi: Step-wise Momentum Fusion for Split Federated Learning on Heterogeneous Data
Mingkun Yang (Delft University of Technology), Jie Yang (Delft University of Technology)
Federated LearningConvolutional Neural NetworkImage
🎯 What it does: Proposes the SMoFi framework, which aligns gradients in Split Federated Learning by synchronizing the server-side optimizer's momentum buffer at each step, addressing gradient bias caused by data heterogeneity and accelerating model convergence.
SNN-Driven Event-Based Flow and Rotation Estimation with SO(3) Refinement
Ruimin Sun (Zhejiang University), De Ma (Zhejiang University)
Pose EstimationSpiking Neural NetworkOptical FlowSequential
🎯 What it does: Built a full-pulse spiking neural network (SNN) for optical flow estimation in event cameras, and achieved camera rotation estimation and panoramic image reconstruction based on the estimated optical flow; further improved rotation accuracy through unsupervised SO(3) optimization.
SNS-Grasp: Semantic-guided Noise Scaling for Grasp Generation
Zhenhua Tang (University of Macau), Chi-Man Pun (University of Macau)
GenerationTransformerDiffusion modelMesh
🎯 What it does: Propose the SNS-Grasp framework, which generates task-related hand grasping poses using semantic-guided noise scaling in diffusion models, and enhances physical feasibility through Fine-grained Grasp Refinement.
SOAR: Semi-Supervised Open-Vocabulary Aerial Object Detection via Dual-Aware Enhanced Prior Denoising
Xu Liu (Xidian University), Licheng Jiao (Xidian University)
Object DetectionTransformerVision Language ModelContrastive LearningImage
🎯 What it does: Proposes SOAR, a semi-supervised open-vocabulary aerial object detection framework, which utilizes dual-perception prior denoising to automatically generate pseudo labels and enhance queries, significantly improving detection performance.
Socrates or Smartypants: Testing Logic Reasoning Capabilities of Large Language Models with Logic Programming-Based Test Oracles
Zihao Xu (University of New South Wales), Yuekang Li (University of New South Wales)
ClassificationData SynthesisLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: Constructed a real English logical fallacy benchmark SMARTYPAT-BENCH and proposed a Prolog-based SMARTYPAT generation framework to automatically generate subtle fallacious texts.
SODiff：Semantic-Oriented Diffusion Model for JPEG Compression Artifacts Removal
Tingyu Yang (Shanghai Jiao Tong University), Yulun Zhang (Carnegie Mellon University)
RestorationTransformerPrompt EngineeringDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: Proposes SODiff, a first-order diffusion model that removes JPEG compression artifacts by leveraging a semantically aligned image prompt extractor and a quality factor-aware temporal predictor.
Soft Conflict-Resolution Decision Transformer for Offline Multi-Task Reinforcement Learning
Shudong Wang (China University of Petroleum East China), Xiaojian Liao (Beihang University)
TransformerReinforcement LearningPrompt EngineeringSequentialBenchmark
🎯 What it does: Propose a soft conflict resolution decision transformer, SoCo-DT, for offline multi-task reinforcement learning.
SOM Directions Are Better than One: Multi-Directional Refusal Suppression in Language Models
Giorgio Piras (University of Cagliari), Battista Biggio (University of Cagliari)
Safty and PrivacyHyperparameter SearchLarge Language ModelTextBenchmark
🎯 What it does: This paper studies the refusal behavior in large language models (LLMs) and proposes a method that utilizes self-organizing maps (SOM) to extract multidimensional refusal directions from internal representations, reducing the model's refusal rate by suppressing these directions.
SOMA: Feature Gradient Enhanced Affine-Flow Matching for SAR-Optical Registration
Haodong Wang (Northwestern Polytechnical University), Yanning Zhang (Northwestern Polytechnical University)
Convolutional Neural NetworkOptical FlowImageMultimodality
🎯 What it does: Proposes the SOMA framework for precise pixel-level SAR and optical image registration, integrating the Feature Gradient Enhancer (FGE) and Global-Local Affine-Flow Matcher (GLAM).
SoMe: A Realistic Benchmark for LLM-based Social Media Agents
Dizhan Xue (Chinese Academy of Sciences), Changsheng Xu (Chinese Academy of Sciences)
Large Language ModelAgentic AITextBenchmark
🎯 What it does: Constructed the SoMe benchmark, evaluating the performance of LLM-driven social media agents on 8 tasks and providing 8 tool platforms.
Sonic4D: Spatial Audio Generation for Immersive 4D Scene Exploration
Siyi Xie (University of Science and Technology of China), Zhibo Chen (University of Science and Technology of China)
GenerationLarge Language ModelVideoMultimodalityAudio
🎯 What it does: Propose the Sonic4D framework to achieve spatial audio generation for 4D scenes, enabling users to experience immersive audio-visual effects from any viewpoint.
Sonnet: Spectral Operator Neural Network for Multivariable Time Series Forecasting
Yuxuan Shu (University College London), Vasileios Lampos (University College London)
Time SeriesBenchmark
🎯 What it does: Proposed the Sonnet model, which captures internal and external dependencies in multivariate time series in the spectral domain using learnable wavelet transforms and Koopman operators, and improves prediction accuracy through multivariate resonance attention (MVCA).
Sortblock: Similarity-Aware Feature Reuse for Diffusion Model
Hanqi Chen (International Joint Innovation Center, Electromagnetics Academy at Zhejiang University), Yi Liu (Baidu Inc)
GenerationComputational EfficiencyTransformerDiffusion modelImageVideo
🎯 What it does: Propose SortBlock, a training-free inference acceleration framework that dynamically caches block-level features in Diffusion Transformer (DiT) to reduce redundant computations.
SOSControl: Enhancing Human Motion Generation Through Saliency-Aware Symbolic Orientation and Timing Control
Ho Yin Au (Hong Kong Baptist University), Jie Chen (Hong Kong Baptist University)
GenerationDiffusion modelTextSequential
🎯 What it does: This paper proposes the Salient Orientation Symbolic (SOS) script and the SOSControl framework, achieving programmable human action generation based on salient orientation symbols. It automatically extracts sparse symbolic scripts from motion data and realizes precise pose and temporal control through ControlNet and iterative optimization in text-driven diffusion models.
Source-Free Graph Foundation Model Adaptation via Pseudo-Source Reconstruction
Liang Yang (Hebei University of Technology), Zhen Wang (Hebei University of Technology)
Domain AdaptationGraph Neural NetworkGenerative Adversarial NetworkContrastive LearningGraphBenchmark
🎯 What it does: Proposes an unsupervised graph domain adaptation method that generates pseudo-source graphs and achieves adversarial alignment with the target graph when source data is unavailable.
SPA: Achieving Consensus in LLM Alignment via Self-Priority Optimization
Yue Huang (University of Notre Dame), Xiangliang Zhang (University of Notre Dame)
OptimizationSafty and PrivacyReinforcement Learning from Human FeedbackLarge Language ModelReinforcement LearningText
🎯 What it does: This paper proposes the Self-Priority Alignment (SPA) framework, achieving trustworthy priority alignment in high-risk scenarios through a fully unsupervised approach, i.e., prioritizing safety (such as harmlessness and honesty) before enhancing helpfulness.
Space Alignment Matters: The Missing Piece for Inducing Neural Collapse in Long-Tailed Learning
Jinping Wang (Wenzhou-Kean University), Zhiwu Xie (Wenzhou-Kean University)
ClassificationImageBenchmark
🎯 What it does: Introduce a spatial alignment mechanism in long-tailed learning to address the angular mismatch between feature space and classifier weight space, and significantly improve model performance through three alignment strategies.
SpaceVLLM: Endowing Multimodal Large Language Model with Spatio-Temporal Video Grounding Capability
Jiankang Wang (Renmin University of China), Yongdong Zhang (Renmin University of China)
Object DetectionObject TrackingData SynthesisTransformerLarge Language ModelVideoTextMultimodality
🎯 What it does: Propose SpaceVLLM, a framework that enables MLLM to perform spatio-temporal video grounding (STVG) through Spatio-Temporal Aware Queries and a Query-Guided Space Head.
SpaCRD: Multimodal Deep Fusion of Histology and Spatial Transcriptomics for Cancer Region Detection
Shuailin Xue (Yunnan University), Wenwen Min (Zhongnan University of Economics and Law)
SegmentationDomain AdaptationTransformerAuto EncoderImageMultimodalityBiomedical Data
🎯 What it does: Proposes SpaCRD, a framework leveraging multi-modal deep fusion and transfer learning for cross-sample, cross-platform, and cross-batch cancer tissue region (CTR) detection.
SPAN: Benchmarking and Improving Cross-Calendar Temporal Reasoning of Large Language Models
Zhongjian Miao (Li Auto Inc), Chen Wei (Li Auto Inc)
TransformerLarge Language ModelAgentic AIPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: This paper introduces the SPAN benchmark to evaluate the ability of large language models in cross-calendar time reasoning (covering ten reasoning directions, two reasoning types, and two question formats);
SPARD: Single-step Inference with Adaptive Sampling in Residual Diffusion for Human Motion Prediction
Yiming Zhang (Jilin University), Renchu Guan (Jilin University)
Pose EstimationKnowledge DistillationGraph Neural NetworkTransformerDiffusion modelScore-based ModelAuto EncoderTime SeriesSequential
🎯 What it does: The SPARD framework achieves efficient inference for human motion prediction through a single-step residual diffusion model and an adaptive noise predictor;
SPARE: Single-Pass Annotation with Reference-Guided Evaluation for Automatic Process Supervision and Reward Modelling
Md Imbesat Hassan Rizvi (Technical University of Darmstadt), Iryna Gurevych (Queen's University)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningPrompt EngineeringText
🎯 What it does: Proposes the SPARE (Single-Pass Annotation with Reference-Guided Evaluation) framework, achieving one-time alignment and evaluation of multi-step reasoning processes generated by LLMs, enabling automated process supervision.
SparK: Query-Aware Unstructured Sparsity with Recoverable KV Cache Channel Pruning
Huanxuan Liao (Chinese Academy Of Sciences), Kang Liu (Chinese Academy Of Sciences)
CompressionComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose the SPARK method, which performs training-free, recoverable channel-level non-structured sparsification on LLM KV cache to reduce memory and computational costs.
Sparse Additive Model Pruning for Order-Based Causal Structure Learning
Kentaro Kanamori (Fujitsu Limited), Takuya Takagi (Fujitsu Limited)
OptimizationComputational EfficiencyGraphBiomedical Data
🎯 What it does: This paper proposes a new sparse additive model pruning method to improve the post-pruning step in causal structure learning based on topological sorting.
Sparse Attention Across Multiple-Context KV Cache
Ziyi Cao (Harbin Institute of Technology), Bingquan Liu (Harbin Institute of Technology)
Computational EfficiencyTransformerLarge Language ModelTextBenchmark
🎯 What it does: Propose the SamKV method, achieving efficient long-sequence inference through sparsification and partial recomputation of multi-context KV Cache.
Sparse Poisson Gamma Belief Networks for High-Dimensional Sparse Count Data
Rui Huang (Harbin Institute of Technology), Sikun Yang (Great Bay University)
ClassificationComputational EfficiencyRepresentation LearningTextBiomedical Data
🎯 What it does: This paper proposes a Sparse Poisson Gamma Belief Network (SPGBN), which enhances feature extraction and missing value prediction performance for high-dimensional sparse count data by explicitly controlling inter-layer connections through sparse graph structure priors and binary masks, building upon the traditional PGBN.
Sparse Tuning Enhances Plasticity in PTM-based Continual Learning
Huan Zhang (Wuhan University), Fan Lyu (Chinese Academy of Sciences)
ClassificationComputational EfficiencyRepresentation LearningTransformerSupervised Fine-TuningPrompt EngineeringContrastive LearningImage
🎯 What it does: Proposes MIST (Mutual Information-guided Sparse Tuning), which first performs sparse fine-tuning on pre-trained models (PTM) to enhance plasticity and generalization ability in continual learning.
Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache Eviction
Yuerong Song (Fudan University), Xipeng Qiu (Fudan University)
GenerationComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose Sparse-dLLM, which enhances the inference speed of Diffusion LLM through dynamic bidirectional cache eviction and sparse attention
Sparse-Scale Transformer with Bidirectional Awareness for Time Series Forecasting
Ying Liu (Hefei University of Technology), Richang Hong (Chongqing University)
Computational EfficiencyTransformerTime Series
🎯 What it does: Propose a sparse scale Transformer (SSformer) that enhances time series forecasting performance by adaptively selecting effective time scales and combining bidirectional scale interactions.
Sparse-vDiT: Unleashing the Power of Sparse Attention to Accelerate Video Diffusion Transformers
Pengtao Chen (Fudan University), Tao Chen (StepFun)
GenerationComputational EfficiencyTransformerDiffusion modelVideoBenchmark
🎯 What it does: This paper proposes Sparse-vDiT, a framework that accelerates video diffusion transformers (vDiT) through sparse attention;
Sparse3DPR: Training-Free 3D Hierarchical Scene Parsing and Task-Adaptive Subgraph Reasoning from Sparse RGB Views
Haida Feng (Chinese Academy of Sciences), Yihong Wu (Chinese Academy of Sciences)
SegmentationComputational EfficiencyGraph Neural NetworkTransformerLarge Language ModelVision Language ModelNeural Radiance FieldImagePoint CloudGraphBenchmarkRetrieval-Augmented Generation
🎯 What it does: Construct an untrained 3D scene parsing and task-adaptive subgraph reasoning framework based on sparse RGB views
Sparse4DGS: 4D Gaussian Splatting for Sparse-Frame Dynamic Scene Reconstruction
Changyue Shi (Hangzhou Dianzi University), Jun Yu (Harbin Institute of Technology)
RestorationDepth EstimationGaussian SplattingVideo
🎯 What it does: Propose Sparse4DGS, which can reconstruct high-quality dynamic 4D scenes from sparse-frame video sequences
SparseCoop: Cooperative Perception with Kinematic-Grounded Queries
Jiahao Wang (Tsinghua University), Jianqiang Wang (Tsinghua University)
Object DetectionObject TrackingAutonomous DrivingTransformerPoint CloudBenchmark
🎯 What it does: Proposed SparseCoop, a collaboration-aware framework fully based on sparse instance queries for 3D detection and tracking.
SparseRM: A Lightweight Preference Modeling with Sparse Autoencoder
Dengcan Liu (University of Science and Technology of China), Yongdong Zhang (University of Science and Technology of China)
Explainability and InterpretabilityComputational EfficiencyRepresentation LearningReinforcement Learning from Human FeedbackLarge Language ModelReinforcement LearningAuto EncoderText
🎯 What it does: Propose SparseRM, a lightweight reward model based on sparse autoencoders (SAE), for modeling and aligning preferences of large language models (LLMs).
SparseSurf: Sparse-View 3D Gaussian Splatting for Surface Reconstruction
Meiying Gu (Beihang University), Xiao Bai (Beihang University)
GenerationConvolutional Neural NetworkGaussian SplattingImage
🎯 What it does: Accurate and detail-rich surface reconstruction under sparse views using 3D Gaussian splatting, while maintaining high-quality view synthesis.
SparseWorld: A Flexible, Adaptive, and Efficient 4D Occupancy World Model Powered by Sparse and Dynamic Queries
Chenxu Dang (Huazhong University of Science and Technology), Yan Wang (Lenovo Group Limited)
Autonomous DrivingComputational EfficiencyTransformerWorld ModelPoint Cloud
🎯 What it does: Proposes a 4D occupancy world model called SparseWorld based on sparse dynamic queries for continuous perception, prediction, and planning.
Spatial Branch-and-Bound for Computing Multiplayer Nash Equilibrium
Jakub Černý (Columbia University), Christian Kroer (Columbia University)
OptimizationGraph
🎯 What it does: Propose a complete space branch-and-bound algorithm that converts the Nash equilibrium problem in multi-player normal-form games into a polynomial complementarity problem;
Spatial-Frequency Spiking Neural Network for Underwater Object Detection
Long Chen (Dalian University of Technology), Qi Xu (Dalian University of Technology)
Object DetectionSpiking Neural NetworkImage
🎯 What it does: Propose a spatial-frequency spiking neural network (SFSNN) for low-power underwater object detection, integrating spatial features with high-frequency subband information and converting YOLOX into an SNN based on symbolic spiking neurons;
Spatial-Spectral Homogeneous Attacks on Physical-World Large Vision-Language Models
Daizong Liu (Wuhan University), Wei Hu (Peking University)
Adversarial AttackVision Language ModelMultimodality
🎯 What it does: Proposes a physical-world attack method based on printable adversarial patches, which can achieve targeted misdirection for image-text tasks by querying model outputs without knowing the internal details of large vision-language models (LVLM).
Spatial-Temporal Feedback Diffusion Guidance for Controlled Traffic Imputation
Xiaowei Mao (Beijing Jiaotong University), Huaiyu Wan (Beijing Jiaotong University)
Graph Neural NetworkDiffusion modelTime Series
🎯 What it does: Proposed a traffic missing value imputation method called FENCE based on space-time feedback diffusion guidance, and verified its superiority on real traffic datasets.
SpatialActor: Exploring Disentangled Spatial Representations for Robust Robotic Manipulation
Hao Shi (Tsinghua University), Gao Huang (Tsinghua University)
Representation LearningRobotic IntelligenceTransformerVision-Language-Action ModelMultimodalityBenchmark
🎯 What it does: Propose the SpatialActor framework to decouple semantics and geometry during robotic manipulation, and build a robust spatial representation through a semantics-guided geometric module and spatial transformer.
SpatialLogic-Bench: A Diagnostic Benchmark for Task-Oriented Spatiotemporal Reasoning
Xiaoda Yang (Zhejiang University), Xiangyu Yue (Chinese University of Hong Kong)
Vision Language ModelVideoMultimodalityBenchmark
🎯 What it does: Propose the SpatialLogic-Bench benchmark to evaluate the ability of vision-language models in task-oriented spatiotemporal logic reasoning, employing multi-scale sliding window and dual-task (forward/reverse) design, requiring the model to determine which frame is closer to task completion;
Spatially Grouped Curriculum Learning for Multi-Agent Path Finding
Thomy Phan (University of Bayreuth), Sven Koenig (University of Bayreuth)
OptimizationTransformerReinforcement LearningGraphBenchmark
🎯 What it does: This paper proposes the PARCEL method, which trains decentralized strategies for multi-agent path planning (MAPF) problems using spatial grouping, masked self-attention, and reverse curriculum learning.
Spatio-Temporal Context Learning with Temporal Difference Convolution for Moving Infrared Small Target Detection
Houzhang Fang (Xidian University), Luxin Yan (Xidian University)
Object DetectionConvolutional Neural NetworkVideo
🎯 What it does: Propose a multi-frame infrared small target detection network named TDCNet, which achieves high-precision detection of moving infrared small targets by using a TDCR module that integrates temporal difference convolution with 3D convolution, and a TDC-guided spatiotemporal attention mechanism.
Spatio-Temporal Distortion Aware Omnidirectional Video Super-Resolution
Hongyu An (University Of Chinese Academy Of Sciences), Ruiqin Xiong (ByteDance Inc)
Super ResolutionConvolutional Neural NetworkTransformerOptical FlowVideo
🎯 What it does: This paper proposes a Spatiotemporal Distortion-Aware Network (STDAN) for enhancing the super-resolution quality of low-resolution panoramic videos.
Spatio-Temporal Hierarchical Causal Models
Xintong Li (Renmin University of China), Xiao Zhou (Renmin University of China)
Graph Neural NetworkGraphTime Series
🎯 What it does: Propose a Spatio-Temporal Hierarchical Causal Model (ST-HCM) that achieves causal inference in observational data with time-invariant unobserved confounding and spatial interventions through hierarchical structures and spatiotemporal dependencies.
SpatioTemporal Difference Network for Video Depth Super-Resolution
Zhengxue Wang (Nanjing University of Science and Technology), Jian Yang (Nanjing University of Science and Technology)
Depth EstimationSuper ResolutionConvolutional Neural NetworkVideo
🎯 What it does: This paper addresses the spatial and temporal long-tail distribution problems in video depth super-resolution by proposing a dual-branch SpatioTemporal Difference Network (STDNet), achieving high-quality high-resolution depth video restoration.
Spatiotemporal-Untrammelled Mixture of Experts for Multi-Person Motion Prediction
Zheng Yin (Nanjing University of Science and Technology), Jinhui Tang (Nanjing University of Science and Technology)
Pose EstimationComputational EfficiencyGraph Neural NetworkMixture of ExpertsSequential
🎯 What it does: Propose a lightweight multi-person human motion prediction framework named ST-MoE, which adaptively captures complex spatiotemporal dependencies by combining Mixture-of-Experts (MoE) and Mamba modules.
SpeakerLM: End-to-End Versatile Speaker Diarization and Recognition with Multimodal Large Language Models
Han Yin (Alibaba Group), Xiangang Li (Alibaba Group)
RecognitionTransformerLarge Language ModelSupervised Fine-TuningTextMultimodalityAudio
🎯 What it does: Proposed SpeakerLM, an end-to-end multimodal large language model (LLM) for speaker diarization and recognition (SDR), supporting variable speaker registration.
SpecDetect: Simple, Fast, and Training-Free Detection of LLM-Generated Text via Spectral Analysis
Haitong Luo (Chinese Academy of Sciences), Yujun Zhang (Chinese Academy of Sciences)
Anomaly DetectionComputational EfficiencyText
🎯 What it does: Develop a text detection method for LLMs based on frequency domain spectral energy called SpecDetect.
SpecDiff: Accelerating Diffusion Model Inference with Self-Speculation
Jiayi Pan (Shanghai Jiao Tong University), Guohao Dai (Shanghai Jiao Tong University)
GenerationComputational EfficiencyTransformerDiffusion modelImageText
🎯 What it does: Propose SpecDiff, a training-free multi-layer feature caching method that utilizes self-speculation information to accelerate diffusion model inference
SpecQuant: Spectral Decomposition and Adaptive Truncation for Ultra-Low-Bit LLMs Quantization
Zhixiong Zhao (Shanghai Jiao Tong University), Haibing Guan (Shanghai Jiao Tong University)
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose a two-stage quantization framework called SpecQuant for large language models, first migrating outliers to the weight domain through activation smoothing, then reducing quantization error by performing channel-adaptive low-frequency Fourier truncation on the weights.
Spectral Basis Learning for Expressive Graph Neural Networks in Link Prediction
Niloofar Azizi (University of Cambridge), Horst Bischof (University of British Columbia)
Computational EfficiencyRepresentation LearningGraph Neural NetworkGraphBenchmark
🎯 What it does: Proposes a novel spectral-based learning method (LLwLC), which enhances the expressiveness of graph neural networks by embedding feature subgraphs with linear constraints (such as vertex deletion subgraphs and Neumann constraints), and applies it to the link prediction task.
Spectral Property-Driven Data Augmentation for Hyperspectral Single-Source Domain Generalization
Taiqin Chen (Harbin Institute of Technology), Yongbing Zhang (Harbin Institute of Technology)
Domain AdaptationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: Propose a spectral attribute-driven data augmentation framework (SPDDA), which generates diverse and realistic extended samples by simulating variations in the number of channels and mixing of adjacent channels in single-source hyperspectral images, thereby enhancing single-source domain generalization capabilities.
Spectrally Adaptive Channel-aware Unrolling Network for Compressed Sensing
Xiaoyang Wang (Northwestern Polytechnical University), Hongping Gan (Northwestern Polytechnical University)
RestorationOptimizationConvolutional Neural NetworkImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed a deep unfolding network SCU-Net based on the spectral projected gradient algorithm for image reconstruction in compressed sensing and magnetic resonance imaging.
Speech Recognition Model Improves Text-to-Speech Synthesis Using Fine-Grained Reward
Guansu Wang (University of Melbourne), Peijie Sun (Nanjing University of Posts and Telecommunications)
GenerationTransformerReinforcement LearningAudio
🎯 What it does: Leverages cross-attention information from a pre-trained ASR model to provide word-level rewards for a self-recursive TTS system, enabling fine-grained optimization of synthesized speech within a reinforcement learning framework.
Speech-Aware Long Context Pruning and Integration for Contextualized Automatic Speech Recognition
Yiming Rong (Key Laboratory of Cognition and Decision Intelligence for Complex Systems Institute of Automation Chinese Academy of Sciences), Bo Xu (Key Laboratory of Cognition and Decision Intelligence for Complex Systems Institute of Automation Chinese Academy of Sciences)
RecognitionTransformerLarge Language ModelSupervised Fine-TuningTextAudio
🎯 What it does: Investigated a two-stage speech-aware long context pruning and integration framework, SAP², which can actively filter and compress long text contexts to enhance speech recognition accuracy.
SPEED-Q: Staged Processing with Enhanced Distillation Towards Efficient Low-Bit On-Device VLM Quantization
Tianyu Guo (Ant Group), Chenguang Ma (Ant Group)
Computational EfficiencyKnowledge DistillationVision Language ModelMultimodalityBenchmark
🎯 What it does: Propose the SPEED-Q framework to achieve 2/4-bit weight quantization of 1–2B parameter vision-language models on edge devices;
SphereDiff: Tuning-free 360° Static and Dynamic Panorama Generation via Spherical Latent Representation
Minho Park (KAIST AI), Jaegul Choo (KAIST AI)
GenerationDiffusion modelImageText
🎯 What it does: Developed SphereDiff, a tuning-free spherical latent space framework capable of generating high-quality 360° static and dynamic wallpapers using existing diffusion models.
Spherical Geometry Diffusion: Generating High-quality 3D Face Geometry via Sphere-anchored Representations
Junyi Zhang (Zhejiang University), Min Zhang (Zhejiang University)
GenerationDiffusion modelAuto EncoderTextMesh
🎯 What it does: Propose a text-driven 3D face generation framework called Spherical Geometry Diffusion based on spherical geometric representation.