AAAI 2026 Papers — Page 33
AAAI Conference on Artificial Intelligence · 4149 papers
RSVG-ZeroOV: Exploring a Training-Free Framework for Zero-Shot Open-Vocabulary Visual Grounding in Remote Sensing Images
Ke Li, Quan Wang (Xidian University)
Object DetectionVision Language ModelDiffusion modelImageBenchmark
🎯 What it does: Proposes a training-agnostic, zero-shot remote sensing visual localization framework RSVG-ZeroOV, achieving open-vocabulary visual localization through frozen VLM and DM attention mechanisms.
RTGaze: Real-Time 3D-Aware Gaze Redirection from a Single Image
Hengfei Wang, Hyung Jin Chang (University Of Birmingham)
Image TranslationGenerationDepth EstimationConvolutional Neural NetworkTransformerNeural Radiance FieldImage
🎯 What it does: This paper proposes RTGaze, a generative model capable of achieving real-time 3D gaze redirection from a single portrait;
RTMol: Rethinking Molecule-text Alignment in a Round-trip View
Letian Chen (Shanghai Innovation Institute), Yang Yang (Shanghai Jiao Tong University)
Drug DiscoveryTransformerLarge Language ModelReinforcement LearningTextGraph
🎯 What it does: Through the self-supervised cyclic learning framework RTMol, the tasks of molecular-to-text description and text-to-molecular generation are unified, achieving bidirectional alignment between molecules and text.
Run, Ruminate, and Regulate: A Dual-process Thinking System for Vision-and-Language Navigation
Yu Zhong (Chinese Academy of Sciences), Yunji Chen (Chinese Academy of Sciences)
Autonomous DrivingRobotic IntelligenceGraph Neural NetworkTransformerLarge Language ModelVision Language ModelVision-Language-Action ModelImageTextMultimodalityChain-of-Thought
🎯 What it does: Designed and implemented a dual-process thinking framework called R3, consisting of three modules: a lightweight Runner, an LLM-driven Ruminator, and a Regulator, achieving efficient and accurate path planning in visual language navigation tasks.
Runtime Safety and Reach-avoid Prediction of Stochastic Systems via Observation-aware Barrier Functions
Shenghua Feng (Institute of Software Chinese Academy of Sciences), Fanjiang Xu (Institute of Software Chinese Academy of Sciences)
OptimizationBenchmarkPhysics RelatedStochastic Differential Equation
🎯 What it does: This paper proposes a runtime safety and achieve-avoid probability prediction framework for discrete-time stochastic systems based on observation-aware barrier functions.
S-D-RSM: Stochastic Distributed Regularized Splitting Method for Large-Scale Convex Optimization Problems
Maoran Wang, Yongxin Chen (Nanjing University Of Science And Technology)
OptimizationTabular
🎯 What it does: Proposed a new stochastic distributed regularized splitting method (S-D-RSM) for solving large-scale distributed convex optimization problems.
S-DAG: A Subject-Based Directed Acyclic Graph for Multi-Agent Heterogeneous Reasoning
Jiangwen Dong (Hong Kong Polytechnic University), Mingjin Zhang (Hong Kong Polytechnic University)
Graph Neural NetworkTransformerLarge Language ModelAgentic AIMixture of ExpertsTextBenchmark
🎯 What it does: Construct a topic-based directed acyclic graph (S-DAG) to fine-grainedly identify relevant knowledge domains in interdisciplinary problems, and deploy domain-expert large language models (LLMs) collaboratively on this graph to achieve multi-agent reasoning.
S^2-KD: Semantic-Spectral Knowledge Distillation Spatiotemporal Forecasting
Wenshuo Wang (South China University of Technology), Yihao Chen (Zhejiang University)
Computational EfficiencyKnowledge DistillationLarge Language ModelVision Language ModelMultimodalityTime SeriesBenchmark
🎯 What it does: Leverage knowledge distillation to transfer semantic causal information and frequency domain features from a multimodal teacher model (vision + text) to a lightweight visual student model, achieving efficient and semantically consistent spatiotemporal prediction.
S2-Boost: Synergistic Semantic Boosting for Coarse-to-Fine Ensemble Learning
Guanxiong He (Northwestern Polytechnical University), Feiping Nie (Northwestern Polytechnical University)
ClassificationSegmentationTransformerDiffusion modelAuto EncoderImage
🎯 What it does: This paper proposes a hierarchical semantic-enhanced ensemble learning framework called S2-Boosting based on human visual cognition.
S2-UniSeg: Fast Universal Agglomerative Pooling for Scalable Segment Anything Without Supervision
Huihui Xu (Hong Kong University of Science and Technology (Guangzhou)), Lei Zhu (Hong Kong University of Science and Technology (Guangzhou))
SegmentationKnowledge DistillationTransformerImage
🎯 What it does: Propose the S2-UniSeg framework, which employs a unified Fast Universal Agglomerative Pooling (UniAP) and Query-wise Self-Distillation for single-stage unsupervised segmentation pre-training.
S2C: A Noise-Resistant Difference Learning Framework for Unsupervised Change Detection in VHR Remote Sensing Images
Lei Ding (Information Engineering University), Jicang Lu (Information Engineering University)
TransformerContrastive LearningImageBenchmark
🎯 What it does: Developed an unsupervised change detection framework S2C based on visual foundation models and contrastive learning, achieving semantic-to-change mapping in very high-resolution (VHR) remote sensing images.
S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation
Jiechao Gao (Stanford University), Yuangang Li (University of Science and Technology of China)
GenerationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose the S2D-ALIGN method, which utilizes multi-stage shallow-to-deep auxiliary learning to achieve anatomical structure alignment for radiology report generation.
S²Drug: Bridging Protein Sequence and 3D Structure in Contrastive Representation Learning for Virtual Screening
Bowei He, Wei-Ying Ma (University Of Illinois Chicago)
Representation LearningDrug DiscoveryProtein Structure PredictionTransformerSupervised Fine-TuningContrastive LearningBiomedical Data
🎯 What it does: Propose a two-stage contrastive learning framework S Drug 2, which first pre-trains on ChemBL using protein sequence-ligand pairs, then fuses sequence and 3D structural information on PDBBind, supplemented by a binding site prediction task to enhance virtual screening and site localization performance.
S²Flow: Towards Fast and Authentic Training-Free High-Resolution Video Generation
Chaoqun Wang (South China Normal University), Xu Yang (South China Normal University)
GenerationTransformerFlow-based ModelRectified FlowVideoOrdinary Differential Equation
🎯 What it does: Proposes S Flow, a training-free high-resolution video generation framework that can achieve video synthesis at the 2560×1536 resolution without retraining the model.
S²HyRec: Self-Supervised Hypergraph Sequential Recommendation
Yuchen Liu (Ocean University of China), Yanwei Yu (Ocean University of China)
Recommendation SystemGraph Neural NetworkContrastive LearningGraphSequential
🎯 What it does: Propose a self-supervised hypergraph sequential recommendation framework S HyRec, integrating global intent tendency, temporal context intent, sequence dependency awareness, and cross-perspective self-supervised learning to enhance sequential recommendation performance.
S²Teacher: Step-by-step Teacher for Sparsely Annotated Oriented Object Detection
Yu Lin (Xiamen University), Liujuan Cao (Xiamen University)
Object DetectionImage
🎯 What it does: Studies directional object detection under sparse annotation, proposing the S Teacher framework to achieve high-precision detection in scenarios with only a few annotated instances.
S³-MSD: Large Vision-Language Model for Explainable and Generalizable Multi-modal Sarcasm Detection
Zhihong Zhu (Tencent Jarvis Lab), Xian Wu (Tencent Jarvis Lab)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelImageTextMultimodality
🎯 What it does: Designed and implemented S-3-MSD, a multimodal large vision-language model that integrates explanation generation with sarcasm detection, capable of simultaneously providing sarcasm judgments and interpretable explanations.
S³: Spiking Neurons as an Isolating Segmenter for Brain Signal Decoding
Qian Zheng (Zhejiang University), Gang Pan (Zhejiang University)
SegmentationSpiking Neural NetworkBiomedical Data
🎯 What it does: Proposed the S3 model, which uses spiking neurons for adaptive segmentation of EEG signals, taking into account individual and task differences while preserving temporal patterns.
S3Net: Spatiotemporally Separated Sparse Network for Neuromorphic Vision Processing
Ping He (Sichuan University), Huajin Tang (Zhejiang University)
ClassificationComputational EfficiencyConvolutional Neural NetworkTime SeriesSequential
🎯 What it does: A framework named S3Net is proposed for processing dynamic visual sensor (DVS) event streams, achieving efficient asynchronous processing through learnable voxel sparse coding and a spatiotemporal separation dual-branch network.
S5: Scalable Semi-Supervised Semantic Segmentation in Remote Sensing
Liang Lv (Wuhan University), Lefei Zhang (Wuhan University)
SegmentationTransformerMixture of ExpertsImageBenchmark
🎯 What it does: Propose a scalable semi-supervised semantic segmentation framework S5, which pretrains an RS base model using a large amount of unlabeled remote sensing images.
SA²GFM: Enhancing Robust Graph Foundation Models with Structure-Aware Semantic Augmentation
Junhua Shi (Beihang University), Xingcheng Fu (Guangxi Normal University)
ClassificationRepresentation LearningGraph Neural NetworkMixture of ExpertsContrastive LearningGraph
🎯 What it does: This paper proposes a graph foundation model named SA 2 GFM, achieving robustness and generalization under cross-domain and few-shot tasks through structure-aware semantic enhancement, information bottleneck pre-training, expert adaptive routing, and hierarchical structure optimization.
SABER: Switchable and Balanced Training for Efficient LLM Reasoning
Kai Zhao (Bilibili Inc), Tianjiao Li (Bilibili Inc)
Computational EfficiencyReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextChain-of-Thought
🎯 What it does: Developed a reinforcement learning-based framework called SABER, enabling large language models to adjust their thinking length according to task difficulty across four switchable reasoning modes (NoThink, FastThink, CoreThink, DeepThink), significantly reducing inference costs while maintaining accuracy.
SACO: Sequence-Aware Constrained Optimization Framework for Coupon Distribution in E-commerce
Li Kong (Renmin University of China), Bicheng Jin (ByteDance)
Recommendation SystemOptimizationTransformerReinforcement LearningTabularSequentialFinance Related
🎯 What it does: Proposed a sequence modeling-based constraint optimization framework SACO for maximizing long-term rewards in multi-user, multi-round coupon allocation.
SACodec: Asymmetric Quantization with Semantic Anchoring for Low-Bitrate High-Fidelity Neural Speech Codecs
Zhongren Dong, Zixing Zhang (Hunan University)
CompressionConvolutional Neural NetworkAuto EncoderGenerative Adversarial NetworkAudio
🎯 What it does: SACodec is a low-bitrate neural speech codec that can simultaneously provide high-fidelity audio reconstruction and semantically rich discrete tokens at 1.5 kbps.
Safe Multi-Agent Reinforcement Learning via Distributional Safety Critic and Maximum Entropy Optimization
Qiwei Liu (East China University of Science and Technology), Huaicheng Yan (East China University of Science and Technology)
OptimizationReinforcement LearningBenchmark
🎯 What it does: This paper proposes a Worst-Case Multi-Agent Soft Actor-Critic (WCMASAC) framework based on maximum entropy and distributed safety assessment, aimed at learning stochastic policies for multi-agent systems under safety constraints.
Safe RAG by RAG: Untying the Bell That RAG Rang with the RAG Hand
Xun Liang (Renmin University of China), Simin Niu (Renmin University of China)
Safty and PrivacyKnowledge DistillationContrastive LearningTextRetrieval-Augmented Generation
🎯 What it does: Developed a framework-level security solution for RAG2RAG, incorporating Detective and Judge dual modules to parallel supervise RAG generation, thereby enhancing security.
SAFE: Semantic- and Frequency-Enhanced Curriculum for Cross-Domain Deepfake Detection
Yulin Yao (Beijing University of Posts and Telecommunications), Dan Luo (Beijing University of Posts and Telecommunications)
Domain AdaptationAnomaly DetectionTransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: Built the SAFE framework, combining semantically enhanced multimodal alignment and dual-score curriculum learning, specifically designed for cross-domain deepfake detection.
SafeMIL: Learning Offline Safe Imitation Policy from Non-Preferred Trajectories
Returaj Burnwal (Indian Institute of Technology Madras), Balaraman Ravindran (Indian Institute of Technology Madras)
Reinforcement LearningSequentialBenchmark
🎯 What it does: Proposed and implemented the SafeMIL algorithm, which leverages limited non-preferred trajectories and a large number of unlabeled trajectories to learn a risk cost function through multi-instance learning. Subsequently, this cost function is used to screen preferred behaviors and train a safe policy via behavioral cloning.
SafeNLIDB: A Privacy-Preserving Safety Alignment Framework for LLM-based Natural Language Database Interfaces
Ruiheng Liu (Xi'an Research Institute of High-Tech), Bailong Yang (Xi'an Research Institute of High-Tech)
Data SynthesisSafty and PrivacyData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningTextTabularBenchmarkChain-of-Thought
🎯 What it does: Developed an end-to-end security alignment framework, SAFENLIDB, to prevent privacy leakage in LLM-driven natural language database interfaces while maintaining the reliability of SQL generation.
SafeR-CLIP: Mitigating NSFW Content in Vision-Language Models While Preserving Pre-Trained Knowledge
Adeel Yousaf (University of Central Florida), Mubarak Shah (University of Central Florida)
TransformerSupervised Fine-TuningContrastive LearningImageTextMultimodality
🎯 What it does: Perform safe fine-tuning on CLIP, proposing SafeR-CLIP which achieves migration and elimination of NSFW content through neighbor-based safety redirection.
SafeSieve: From Heuristics to Experience in Progressive Pruning for LLM-based Multi-Agent Communication
Ruijia Zhang (National University of Singapore), Qingsong Wen (Nanyang Technological University)
Computational EfficiencyTransformerLarge Language ModelAgentic AITextBenchmark
🎯 What it does: Propose SafeSieve, an algorithm for progressively adaptive sparsification of multi-agent communication graphs;
Safety Alignment of Large Language Models via Contrasting Safe and Harmful Distributions
Xiaoyun Zhang (State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences), Xing Hu (City University of Hong Kong)
Safty and PrivacyLarge Language ModelPrompt EngineeringContrastive LearningTextBenchmark
🎯 What it does: Propose an adversarial contrastive decoding (ACD) method that enhances the safety of large language models (LLMs) by learning from the contrast between safe prompts and adversarial prompts.
SafetyReminder: Reviving Delayed Safety Awareness of Vision-Language Models to Defend Against Jailbreak Attacks
Peiyuan Tang (Xi'an Jiaotong University), Zijiang James Yang (Singapore Management University)
Safty and PrivacyAdversarial AttackSupervised Fine-TuningPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: Discover that Vision-Language Models (VLMs) exhibit 'delayed safety awareness' during the generation process, and propose the SafetyReminder framework, which utilizes SAPT to learn soft prompts that activate safety awareness in the intermediate generation stage, preventing the generation of malicious content.
SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation
Paul Grimal (Université Paris-Saclay), Akihiro Sugimoto (National Institute of Informatics)
GenerationDiffusion modelFlow-based ModelImageTextMultimodality
🎯 What it does: Proposes the SAGA method, which learns a Gaussian distribution with high success rates under text prompts, directly samples from it, and performs fine-tuning to enhance the semantic alignment quality in text-to-image generation.
SAGE: Spuriousness-Aware Guided Prompt Exploration for Mitigating Multimodal Bias
Wenqian Ye (University of Virginia), Aidong Zhang (University of Virginia)
ClassificationPrompt EngineeringVision Language ModelContrastive LearningMultimodality
🎯 What it does: Studied how to suppress multimodal pseudo correlations in zero-shot classification tasks of pre-trained vision-language models such as CLIP, and proposed the SAGE method to enhance model robustness through prompt exploration.
SAGE: Structured Attribute-Guided Enhancement for GZSL
Zao Zhang (Chinese Academy of Sciences), Pin Lyu (Chinese Academy of Sciences)
ClassificationKnowledge DistillationTransformerContrastive LearningImage
🎯 What it does: Proposed the Structured Attribute-Guided Enhancement Framework (SAGE), leveraging multi-scale visual encoding, cross-modal attention, and memory pool-driven subset-aware distillation to significantly improve the performance of generalized zero-shot learning.
SageLM: A Multi-aspect and Explainable Large Language Model for Speech Judgement
Yuan Ge (Northeastern University), Jingbo Zhu (Northeastern University)
Data SynthesisExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextMultimodalityAudio
🎯 What it does: This paper proposes SageLM, an end-to-end, interpretable multi-dimensional speech evaluation model capable of simultaneously assessing semantic and acoustic features.
SalDiff-DTM: A Novel Dual-Temporal Modulated Diffusion Model for Omnidirectional Images Scanpath Prediction
Xiaohui Kong (East China Normal University), Xiongkuo Min (Shanghai Jiao Tong University)
Recurrent Neural NetworkGraph Neural NetworkDiffusion modelScore-based ModelImage
🎯 What it does: This paper proposes the SalDiff-DTM model, combining a dual-temporal modulation diffusion network, Dual-GCN, and TABiMamba module for predicting eye movement scan paths on panoramic images.
SALR: Sparsity-Aware Low-Rank Representation for Efficient Fine-Tuning of Large Language Models
Longteng Zhang (Hong Kong University of Science and Technology), Xiaowen Chu (Harbin Institute of Technology)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Develop a sparse low-rank adaptation fine-tuning framework named SALR, which combines sparse pruning with low-rank LoRA for efficient fine-tuning of large pre-trained language models.
SAM-DAQ: Segment Anything Model with Depth-guided Adaptive Queries for RGB-D Video Salient Object Detection
Jia Lin (Hangzhou Dianzi University), Jiyong Zhang (Hangzhou Dianzi University)
SegmentationTransformerVideoMultimodality
🎯 What it does: This paper proposes a RGB-D video salient object detection framework called SAM-DAQ based on SAM2, achieving salient object segmentation without manual prompts through depth-guided parallel adapters and query-driven temporal memory modules.
SAM2-OV: A Novel Detection-Only Tuning Paradigm for Open-Vocabulary Multi-Object Tracking
Yangkai Chen (Xiamen University), Hanzi Wang (Xiamen University)
Object DetectionObject TrackingTransformerVision Language ModelContrastive LearningImageVideo
🎯 What it does: Adopting a detection-only fine-tuning paradigm, leveraging SAM2's zero-shot cross-frame association to achieve open-vocabulary multi-object tracking.
SAM2MOT: A Novel Paradigm of Multi-Object Tracking by Segmentation
Junjie Jiang (Huawei Cloud), Dongsheng Jiang (Huawei Cloud)
Object TrackingSegmentationVideo
🎯 What it does: Proposes a segmentation-based multi-object tracking framework SAM2MOT, achieving zero-shot tracking
SAMCL: Empowering SAM to Continually Learn from Dynamic Domains with Extreme Storage Efficiency
Zeqing Wang (Xidian University), Fei Cheng (Xidian University)
SegmentationDomain AdaptationComputational EfficiencyPrompt EngineeringImageBiomedical Data
🎯 What it does: Propose a continual learning method named SAMCL, enabling SAM to incrementally learn and maintain performance across multiple domains.
SAME: Spatial-Aware Multimodal Egocentric Human Pose Estimation
Yurong Fu (Pico Tsinghua University), Haoqian Wang
Pose EstimationConvolutional Neural NetworkTransformerImageMultimodality
🎯 What it does: This paper proposes the SAME framework, achieving self-perspective human pose estimation through multi-modal fusion (stereo images + sparse IMU)
SAMGTD: Spatial-Aware Masked Graph Transformer-Diffusion Model for Enhanced Cell Type Deconvolution in Spatial Transcriptomics
Shilin Zhang, Xiulong Liu (Tianjin University)
Graph Neural NetworkTransformerDiffusion modelAuto EncoderContrastive LearningBiomedical Data
🎯 What it does: Propose the SAMGTD model for cell type deconvolution in spatial transcriptomics.
Sample Weighted Incomplete Multimodal Clustering Based on Graph Coarsening Label Extraction
Zhenjiao Liu (Inspur Cloud Information Technology Co., Ltd.), Liang Zhao (Dalian University of Technology)
OptimizationComputational EfficiencyGraph Neural NetworkTransformerMultimodality
🎯 What it does: This paper proposes a sample weighted incomplete multimodal clustering method based on graph coarsening label extraction (IMC-GCSW), aiming to simultaneously address the incompleteness and quality inconsistency of multimodal data.
Sample-and-Search: An Effective Algorithm for Learning-Augmented k-Median Clustering in High Dimensions
Kangke Cheng (University of Science and Technology of China), Hu Ding (University of Science and Technology of China)
OptimizationComputational EfficiencyImage
🎯 What it does: Proposed and implemented a learning-enhanced k-median clustering algorithm based on sampling and search.
Sample-specific Modality Diagnosis and Cross-modal Enhancement for Incomplete Multimodal Representations
Junsong Chen, Xinwang Liu (National University Of Defense Technology)
ClassificationRecurrent Neural NetworkTransformerImageTextMultimodalityBenchmarkAudio
🎯 What it does: Proposed the SMCIR framework, which can diagnose missing modalities for each sample and complete missing modalities through cross-modal enhancement, thereby improving the performance of multimodal sentiment analysis.
Sampling Control for Imbalanced Calibration in Semi-Supervised Learning
Senmao Tian (Beijing Jiaotong University), Shunli Zhang (Beijing Jiaotong University)
ClassificationImage
🎯 What it does: Proposes the SC-SSL framework, which suppresses class imbalance in semi-supervised learning through decoupling sampling control, with particular emphasis on feature learning for minority classes and numerical calibration;
Sampling-Free Uncertainty Quantification via Hidden State Dynamics in Language Models
Yixin Bu (Nanjing University of Aeronautics and Astronautics), Piji Li (COMAC Aircraft Design and Research Institute)
Explainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose a sampling-free internal hidden state dynamic analysis method to quantify uncertainty in the generation results of large language models during a single forward pass.
SampurNER: Fine-Grained Named Entity Recognition Dataset for 22 Indian Languages
Prachuryya Kaushik (Indian Institute of Technology Guwahati), Ashish Anand (Indian Institute of Technology Guwahati)
RecognitionTransformerSupervised Fine-TuningTextBenchmark
🎯 What it does: Proposed the entity-anchored machine translation framework EaMaTa, automatically translating the English Fine-grained NER dataset FewNERD into 22 Indian languages, and constructed the first FgNER dataset SampurNER covering all 22 languages.
SAOT: An Enhanced Locality-Aware Spectral Transformer for Solving PDEs
Chenhong Zhou (Hong Kong Baptist University), Zaifeng Yang (Agency for Science, Technology and Research)
TransformerMeshPhysics Related
🎯 What it does: Propose a spectral Transformer (SAOT) that integrates wavelet attention and Fourier attention to efficiently approximate the solution operators of PDEs.
SAPO: Self-Adaptive Process Optimization Makes Small Reasoners Stronger
Kaiyuan Chen (Yunnan University), Xuejie Zhang (Yunnan University)
OptimizationTransformerSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Proposes the Self-Adaptive Process Optimization (SAPO) framework, enabling self-evolution in small language models by identifying the first error and performing local verification.
SAQ-SAM: Semantically-Aligned Quantization for Segment Anything Model
Jing Zhang (Chinese Academy of Sciences), Qingyi Gu (Chinese Academy of Sciences)
SegmentationComputational EfficiencyKnowledge DistillationTransformerImage
🎯 What it does: In this paper, the authors propose a post-training quantization framework named SAQ-SAM for the Segment Anything Model (SAM), aiming to significantly enhance low-bit quantization performance through semantically consistent pruning and prompt-aware reconstruction.
SAR-DisentDM: A Semantic-Disentangled Diffusion Model for Limited-Data SAR Image Synthesis
Yue Yang (Sichuan University), Zijian Deng (Sichuan University)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: Designed a SAR image generation framework called SAR-DisentDM suitable for limited data environments, which can control category and orientation attributes to generate diverse and realistic SAR images.
SAR: A Structure-Aligned Reasoning Framework for Temporal Knowledge Graph Question Answering
Qianyi Hu (Central China Normal University), Shoujin Wang (University of Technology Sydney)
TransformerLarge Language ModelAgentic AIPrompt EngineeringTextGraphRetrieval-Augmented Generation
🎯 What it does: Propose the SAR framework, integrating LLM reasoning, structural alignment retrieval, and iterative verification to address structural mismatch issues in temporal knowledge graph question answering.
Sarcopenia Assessment Model Based on Dual-Source Modal Graph
Wenxian Zheng (University of Electronic Science and Technology of China), Yongguo Liu (University of Electronic Science and Technology of China)
ClassificationRepresentation LearningConvolutional Neural NetworkGraph Neural NetworkImageTabularTime SeriesBiomedical Data
🎯 What it does: Developed a dual-source feature map-based skeletal muscle assessment model DFGSE, which combines low-energy and high-energy DXA images along with blood biomarkers to evaluate muscle mass.
SASST: Leveraging Syntax-Aware Chunking and LLMs for Simultaneous Speech Translation
Zeyu Yang (Chinese University of Hong Kong), Satoshi Nakamura (Okinawa Institute of Science and Technology)
GenerationTransformerLarge Language ModelTextAudio
🎯 What it does: Proposes a syntax-based block strategy and the SASST framework to achieve end-to-end speech-to-text translation.
Sat2Flow: A Structure-Aware Diffusion Framework for Human Flow Generation from Satellite Imagery
Xiangxu Wang (Shenzhen Technology University), Jinzhou Cao (Shenzhen Technology University)
GenerationDiffusion modelContrastive LearningImageTabular
🎯 What it does: Construct a structure-preserving diffusion framework, Sat2Flow, to generate urban OD flow matrices using satellite imagery as the sole input.
SatireDecoder: Visual Cascaded Decoupling for Enhancing Satirical Image Comprehension
Yue Jiang (Fudan University), Xu Zheng (Sofia University 'St. Kliment Ohridski')
Explainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelVision Language ModelImageTextMultimodalityChain-of-Thought
🎯 What it does: Proposes a training-free SatireDecoder framework that enhances the understanding of satirical images through multi-agent visual cascading decoupling and chain-of-thought reasoning.
Satisficing and Optimal Generalised Planning via Goal Regression
Dillon Z. Chen (Vector Institute), Sheila A. McIlraith (Vector Institute)
OptimizationBenchmark
🎯 What it does: Propose a generic planning method called MOOSE based on goal regression, which can generate general plans that are directly executable or used for search pruning from training instances.
SAVER: Mitigating Hallucinations in Large Vision-Language Models via Style-Aware Visual Early Revision
Zhaoxu Li (Nanyang Technological University), Xudong Jiang (Nanyang Technological University)
GenerationExplainability and InterpretabilityTransformerVision Language ModelImage
🎯 What it does: Propose the SAVER mechanism, which dynamically corrects the generation results of large vision-language models during inference by leveraging attention patterns from early vision layers, thereby reducing the hallucination rate in stylized images.
Say More with Less: Variable-Frame-Rate Speech Tokenization via Adaptive Clustering and Implicit Duration Coding
Rui-Chen Zheng (University of Science and Technology of China), Zhen-Hua Ling (University of Science and Technology of China)
CompressionRepresentation LearningAuto EncoderGenerative Adversarial NetworkAudio
🎯 What it does: Propose VARSTok, a variable frame rate speech tokenizer that adaptively assigns token numbers based on local feature similarity.
SC-Net: Robust Correspondence Learning via Spatial and Cross-Channel Context
Shuyuan Lin (Jinan University), Jian Weng (Jinan University)
Pose EstimationComputational EfficiencyGraph Neural NetworkTransformerImage
🎯 What it does: SC-Net achieves robust learning of correspondences between two views by fusing spatial and channel contexts.
Scalable Mixed-Integer Optimization with Neural Constraints via Dual Decomposition
Shuli Zeng (University of Science and Technology of China), Xiangyang Li (University of Science and Technology of China)
OptimizationConvolutional Neural NetworkRecurrent Neural NetworkTabularBenchmarkAgriculture Related
🎯 What it does: Propose an augmented Lagrangian-based dual decomposition framework that decomposes the mixed integer programming with embedded neural networks into independent MIP subproblems and neural network subproblems, achieving scalable and modular solutions;
Scalable Multi-Objective and Meta Reinforcement Learning via Gradient Estimation
Zhenshuo Zhang (Northeastern University), Hongyang R. Zhang (Northeastern University)
Meta LearningReinforcement Learning
🎯 What it does: Propose a scalable multi-objective reinforcement learning method that first trains a meta-policy, then rapidly evaluates the adaptability of any task subset using first-order gradient estimation to obtain a task affinity matrix and cluster groups;
Scalable Privacy-Preserving Neural Network Training over Z2k via RMFE-Based Packing and Mixed-Circuit Computation
Hengcheng Zhou (Shanghai Jiao Tong University)
Computational EfficiencyImage
🎯 What it does: Propose a scalable多方privacy-preserving neural network training framework that supports any number of participants in an honest majority setting, using Shamir secret sharing over Galois rings, RMFE packing, and hybrid circuits to achieve efficient parallel training.
Scalable Semi-supervised Community Search via Graph Transformer on Attributed Heterogeneous Information Networks
Linlin Ding, Renata Borovica-Gajic (Liaoning University)
Graph Neural NetworkTransformerContrastive LearningGraph
🎯 What it does: Designed a scalable semi-supervised community search framework SCSAH, enabling efficient community discovery on attribute-heterogeneous information networks.
Scalable Solution Methods for Dec-POMDPs with Deterministic Dynamics
Yang You (United Kingdom Atomic Energy Authority), Nick Hawes (University of Oxford)
OptimizationReinforcement LearningBenchmark
🎯 What it does: Proposed the Deterministic Decentralized POMDPs (Det-Dec-POMDPs) framework and designed an efficient solver called IDPP, which can compute Nash equilibrium in large-scale problems.
Scalable Solutions to Zero-Sum Partially Observable Stochastic Games Through Belief Aggregation with Approximation Guarantees
Kim Hammar (University of Melbourne), Tansu Alpcan (University of Melbourne)
Reinforcement LearningBenchmark
🎯 What it does: Propose a new method called SAB (Shapley Iteration with Aggregated Beliefs) to approximately solve the value function of one-sided zero-sum partially observable stochastic games (POSG).
Scalable Vision-Guided Crop Yield Estimation
Harrison H Li (Harvey Mudd College), David B. Lobell (Stanford University)
Convolutional Neural NetworkImageAgriculture Related
🎯 What it does: Propose combining predictive projection inference (PPI) with computer vision models, leveraging field photos and geographic coordinates to supplement limited crop cutting measurements, thereby improving the accuracy of regional average yield estimates.
SCALAR: Scale-wise Controllable Visual Autoregressive Learning
Ryan Xu (Alibaba Group), Xiangxiang Chu (Alibaba Group)
GenerationTransformerImage
🎯 What it does: Propose the SCALAR method to achieve controllable visual autoregressive image generation
Scale-Net: A Hierarchical U-Net Framework for Cross-Scale Generalization in Multi-Task Vehicle Routing
Suyu Liu, Yew-Soon Ong (Nanyang Technological University)
OptimizationConvolutional Neural NetworkTransformerReinforcement LearningGraph
🎯 What it does: Proposed Scale-Net, a unified multi-task vehicle routing neural solver, achieving cross-scale generalization from small-scale training to large-scale instances with thousands of nodes through a hierarchical U-Net framework;
SCALE: Selective Resource Allocation for Overcoming Performance Bottlenecks in Mathematical Test-time Scaling
Yang Xiao (Hong Kong Polytechnic University), Pengfei Liu (Shanghai Jiao Tong University)
OptimizationComputational EfficiencyTextChain-of-Thought
🎯 What it does: Decompose mathematical problems into subproblems, allocate computational resources according to difficulty, and achieve selective computation during testing.
Scaling and Transferability of Annealing Strategies in Large Language Model Training
Siqi Wang (Hong Kong University of Science and Technology), Xiaomeng Li (Meituan Inc)
OptimizationHyperparameter SearchLarge Language ModelMixture of ExpertsText
🎯 What it does: Studied and verified the transferability of learning rate annealing strategies in large-scale language model training, and proposed an improved prediction framework;
Scaling Law Analysis in Federated Learning: How to Select the Optimal Model Size?
Xuanyu Chen (University of Sydney), Dong Yuan (University of Sydney)
OptimizationFederated LearningComputational EfficiencyConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper derives the generalization error upper bound for federated learning with stochastic gradient descent (FedSGD) based on PAC-Bayes theory, and uses this upper bound to obtain the analytical optimal solution for model scale, thereby studying the impact of data decentralization on the selection of large-scale model sizes.
Scaling Law for Large Wireless Models
Ziheng Liu (Beijing Jiaotong University), Enyu Shi (Beijing Jiaotong University)
OptimizationTransformerPhysics Related
🎯 What it does: Propose a novel wireless scaling law that considers channel heterogeneity and discretization granularity to predict the performance of large-scale wireless models (LWM)
Scaling Laws for Conditional Emergence of Multilingual Image Captioning via Generalization from Translation
Julian Spravil (Fraunhofer IAIS), Sven Behnke (University of Bonn)
GenerationTransformerSupervised Fine-TuningVision Language ModelContrastive LearningTextMultimodality
🎯 What it does: Investigated the use of multimodal translation task training data to achieve zero-shot generalization for image captioning in unseen languages within multilingual multitask scenarios.
Scaling LLM Speculative Decoding: Non-Autoregressive Forecasting in Large-Batch Scenarios
Luohe Shi (Wuhan University), Hai Zhao (Shanghai Jiao Tong University)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: Propose SpecFormer, a non-autoregressive draft generation architecture that combines unidirectional and bidirectional attention, designed to accelerate lossless autoregressive inference in large-batch inference scenarios;
Scaling Towards the Information Boundary of Instructions through Data Synthesizing
Li Du (Beijing Academy of Artificial Intelligence), Tengfei Pan (Beijing Academy of Artificial Intelligence)
Data SynthesisData-Centric LearningLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed a closed-loop instruction data construction framework and built a high-quality instruction dataset named Infinity Instruct Subject (InfInstruct-Sub) with approximately 1.5 million entries.
Scaling-up Perceptual Video Quality Assessment
Ziheng Jia (Shanghai Jiao Tong University), Xiongkuo Min (Shanghai Jiao Tong University)
Data-Centric LearningTransformerLarge Language ModelVision Language ModelMultimodalityBenchmark
🎯 What it does: Constructed a large-scale multi-modal instruction database OmniVQA-Chat-400K and a video quality rating dataset OmniVQA-MOS-20K, achieving data scaling through machine + human annotation methods;
SCAN: Self-Calibrated AutoregressioN for High-Quality Visual Generation
Zhanzhou Feng (Peking University), Shiliang Zhang (Ant Group)
ClassificationGenerationTransformerImage
🎯 What it does: Propose a self-calibrating autoregressive visual generative model called SCAN, which can evaluate and refine generated image patches without regenerating the entire image.
scCluBench: Comprehensive Benchmarking of Clustering Algorithms for Single-Cell RNA Sequencing
Ping Xu (Computer Network Information Center, Chinese Academy of Sciences), Yuanchun Zhou (Computer Network Information Center, Chinese Academy of Sciences)
Graph Neural NetworkTransformerAuto EncoderBiomedical DataBenchmark
🎯 What it does: Provides a unified, standardized benchmark framework for scRNA-seq clustering algorithms (scCluBench), aggregating 36 datasets covering 18 human and mouse tissues, spanning various scales and sparsity levels, and conducting systematic evaluations of 36 traditional, deep learning, graph neural network, and biology-based clustering methods.
Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition
Xiaowen Fu (Shenzhen University), Jinbao Wang (Shenzhen University)
Mixture of ExpertsNeural Radiance FieldGaussian SplattingImage
🎯 What it does: Propose Scene Experts and implement MoE-GS by splitting scenes and using a learnable router to dynamically assign anchors, enhancing the representation capability of anchor-based 3D Gaussian Splatting and improving reconstruction quality in complex scenes.
Scene-Aware Spatiotemporal Generalization: Towards Robust Temporal Action Detection Across Domains
Fangming Feng (Zhejiang University), Tao Jin (Zhejiang University)
Object DetectionDomain AdaptationTransformerVideoBenchmark
🎯 What it does: Propose the first domain generalization framework for long video spatiotemporal action detection, enhancing cross-domain generalization capabilities through two modules: Scene-Aware Video Segmentation (SAVS) and Temporal-Adaptive Normalization Perturbation (TANP).
SceneGenesis: 3D Scene Synthesis via Semantic Structural Priors and Mesh-Guided Video-Geometry Fusion
Yueming Zhao (Beihang University), Di Huang (Beihang University)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelGaussian SplattingVideoTextMesh
🎯 What it does: Propose the SceneGenesis framework, based on semantic structure initialization, geometry-conditioned multi-view video synthesis, and mesh-guided video-geometry fusion, achieving controllable and scalable 3D scene generation and editing.
SceneJailEval: A Scenario-Adaptive Multi-Dimensional Framework for Jailbreak Evaluation
Lai Jiang (Shanghai Jiao Tong University), Li Pan (Shanghai Jiao Tong University)
ClassificationSafty and PrivacyLarge Language ModelTextBenchmark
🎯 What it does: Constructed a scenario-adaptive multi-dimensional evaluation framework named SceneJailEval, and collected a high-quality jailbreak dataset containing 14 scenarios and 1,308 examples based on this framework, used to evaluate the jailbreak success rate and risk level of LLMs.
SchellingFormer: Laplacian Matrix-guided Geometric Transformer for Robust Schelling Point Detection
Yihao Chen (Nanyang Technological University), Jianmin Zheng (Alibaba Group)
Object DetectionTransformerMesh
🎯 What it does: Propose SchellingFormer, a Laplacian matrix-guided geometric Transformer for Schelling point detection on 3D meshes.
Schema-Guided Event Reasoning: A Plug-and-Play Event Reasoning Framework Based on Large Language Models
Yuying Liu (National University of Defense Technology), Bin Zhou (National University of Defense Technology)
TransformerLarge Language ModelPrompt EngineeringContrastive LearningText
🎯 What it does: Proposes a pluggable Schema-Guided Event Reasoning (SGER) framework comprising three modules: event abstraction, pattern prediction, and reasoning, which enhances the event reasoning capabilities of large language models without relying on pre-built schema libraries.
Schema-Guided Scene-Graph Reasoning Based on Multi-Agent Large Language Model System
Yiye Chen (Arm), Benjamin E Lundell (Arm)
Representation LearningTransformerLarge Language ModelAgentic AIPrompt EngineeringGraphRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose a scene graph reasoning framework SG 2 based on multi-agent large language models, achieving more precise spatial reasoning and planning tasks through alternating iterations between the reasoner and retriever;
SciMKG: A Multimodal Knowledge Graph for Science Education with Text, Image, Video and Audio
Tong Lu, Junsheng Du (Beijing Normal University)
TransformerLarge Language ModelVision Language ModelImageVideoTextMultimodalityAudio
🎯 What it does: This paper proposes an automated framework that utilizes large language models to extract concepts and perform multi-modal alignment on K12 science MOOC resources, constructing a multi-modal educational knowledge graph called SciMKG that includes text, images, videos, and audio.
SCIR: A Self-Correcting Iterative Refinement Framework for Enhanced Information Extraction Based on Schema
Yushen Fang (Huazhong University of Science and Technology), Wenqi Yang (Huazhong University of Science and Technology)
TransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmark
🎯 What it does: Propose the SCIR self-correcting iterative refinement framework, integrating Dual-Path Self-Correcting and Feedback-Driven Optimization, combined with the MBSC dataset, to achieve efficient information extraction without fine-tuning across languages;
SCo-Cloud: Satellite Constellation Collaboration for Cloud-Aware Onboard-Computed Imaging and Transmission
Jia Liu (Beijing University of Posts and Telecommunications), Shangguang Wang (Beijing University of Posts and Telecommunications)
RestorationCompressionOptimizationTransformerContrastive LearningImage
🎯 What it does: SCo-Cloud proposes a constellation architecture based on collaboration between central satellites and edge satellites, achieving cloud detection, thin cloud removal, thick cloud repositioning localization, task scheduling, and content-aware downlink;
SCoNE: Spherical Consistent Neighborhoods Ensemble for Effective and Efficient Multi-View Anomaly Detection
Yang Xu (Nanjing University), Kai Ming Ting (Deakin University)
Anomaly DetectionImageGraphTabularBenchmark
🎯 What it does: This paper proposes a novel multi-view anomaly detection method called SCoNE, which directly constructs an adaptive spherical neighborhood using multi-view samples and makes anomaly judgments based on consistent neighborhoods.
Scope Delineation Before Localization: A Two-Stage Framework for Enhancing Failure Attribution in Multi-Agent Systems
Kai Sun (Xi'an Jiaotong University), Bin Shi (Xi'an Jiaotong University)
Anomaly DetectionTransformerLarge Language ModelPrompt EngineeringTextSequentialChain-of-Thought
🎯 What it does: Proposed a two-stage failure attribution framework called Scope Delineation Before Localization (SDBL), which improves failure attribution accuracy in multi-agent systems by first delineating the failure scope and then precisely localizing the failure steps.
SCOPE: Intrinsic Semantic Space Control for Mitigating Copyright Infringement in LLMs
Zhenliang Zhang (Peking University), Xiaojun Wan (Peking University)
Safty and PrivacyLarge Language ModelAuto EncoderText
🎯 What it does: Propose a subspace control method SCOPE based on sparse autoencoders to real-time suppress LLMs from generating copyrighted content.
Score-Based Model for Low-Rank Tensor Recovery
Zhengyun Cheng (Northwestern Polytechnical University), Xiangyang Ji (Tsinghua University)
RestorationScore-based ModelImageVideoStochastic Differential Equation
🎯 What it does: This paper proposes a score-matching-based energy model that directly learns the gradient of the joint distribution between tensors and latent factors, achieving low-rank tensor completion and denoising.
SCORE: Semantic Collage by Optimizing Rendered Elements
Zefan Shao (Shenzhen University), Pengfei Xu (Shenzhen University)
GenerationOptimizationDiffusion modelScore-based ModelGaussian SplattingImageText
🎯 What it does: Built a text-prompt-based image collage generation framework named SCORE, which automatically generates semantically aligned, structurally reasonable, and preserving original image integrity collages by optimizing spatial parameters through differential rendering and Variational Score Distillation.
SCoUT: A Framework for Structured Stereotype Analysis in Language Models
Jinxuan Wu (Fudan University), Xiangyang Xue (Fudan University)
Explainability and InterpretabilityLarge Language ModelText
🎯 What it does: Propose the SCoUT framework, which utilizes Thurstone comparison judgment to reconstruct the latent utility space of warmth and ability dimensions within LLMs, locates these dimensions in attention heads through linear probing, and achieves interpretable and controllable regulation of generated content during inference via activation intervention.
SculptDrug: A Spatial Condition-Aware Bayesian Flow Model for Structure-based Drug Design
Qingsong Zhong (East China Normal University), Jilin Hu (East China Normal University)
Drug DiscoveryFlow-based ModelBiomedical Data
🎯 What it does: Proposed SculptDrug, a spatially condition-aware structural drug design model based on Bayesian flow networks, for generating drug molecules that comply with protein surface constraints.