AAAI 2026 Papers — Page 24
AAAI Conference on Artificial Intelligence · 4149 papers
Memorize Early, Then Query: Inlier-Memorization-Guided Active Outlier Detection
Minseo Kang (Sungshin Women's University), Dongha Kim (Sungshin Women's University)
Anomaly DetectionFlow-based ModelAuto EncoderImageTextTabularBenchmark
🎯 What it does: Propose the IMBoost framework, combining the inlier-memorization effect with active learning to achieve unsupervised anomaly detection.
MemoryART: Enhancing LLMs via Multi-Memory Models with Adaptive Resonance Theory for Healthcare Agents
Renke Dai (South Central Minzu University), Ah-Hwee Tan (Singapore Management University)
TransformerLarge Language ModelTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose MemoryART, a multi-memory framework based on Adaptive Resonance Theory (ART), to enhance long-term memory and retrieval capabilities of large language models (LLMs) in multi-turn medical dialogues.
Mental Model-based Generation of Lies for Insider Threat Modeling
Brittany Cates (Colorado State University), Sarath Sreedharan (Colorado State University)
Adversarial AttackBenchmark
🎯 What it does: Studied deception mechanisms in insider threat attacks, proposing a framework that generates lies and plans through model reconstruction and planning search, enabling attackers to complete hidden objectives while maintaining the supervisor's belief that they are performing expected tasks.
MentalGuide: Towards Multi-Turn, State-Aware and Strategy-Driven Conversations for Mental Health Support
Jinwei He (Beihang University), Feng Lu (Beihang University)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposed the MentalGuide framework, combining expert experience priors with LLM real-time inference for multi-round, state-aware, strategy-driven mental health dialogues.
MergeDNA: Context-Aware Genome Modeling with Dynamic Tokenization Through Token Merging
Siyuan Li (Zhejiang University), Stan Z. Li (BioMap Research)
Representation LearningTransformerLarge Language ModelSupervised Fine-TuningBiomedical Data
🎯 What it does: Proposed an adaptive DNA tokenization and context-aware hierarchical Transformer framework called MergeDNA, which realizes tokenization and representation of variable-length gene sequences through dynamic token merging techniques.
MeshA*: Efficient Path Planning with Motion Primitives
Marat Agranovskiy (St. Petersburg State University), Konstantin Yakovlev (St. Petersburg State University)
Autonomous DrivingOptimizationComputational EfficiencyBenchmark
🎯 What it does: Proposed a grid-cell-based search algorithm called MeshA*, which solves path planning problems with motion primitive constraints by searching on the grid while recording motion primitives passing through cells.
MeshSplat: Generalizable Sparse-View Surface Reconstruction via Gaussian Splatting
Hanzhi Chang (University of Science and Technology of China), Tianzhu Zhang (University of Science and Technology of China)
GenerationConvolutional Neural NetworkTransformerGaussian SplattingImageMesh
🎯 What it does: Propose the MeshSplat framework, which achieves generalizable surface reconstruction under sparse views by utilizing pixel-aligned 2D Gaussian Splatting (2DGS).
Meta Dynamic Graph for Traffic Flow Prediction
Yiqing Zou (Beijing Institute of Technology), Sijie Ruan (Beijing Institute of Technology)
Recurrent Neural NetworkGraph Neural NetworkTime Series
🎯 What it does: Propose the MetaDG framework, which utilizes dynamic node embeddings to generate time-varying graph structures for unified spatiotemporal modeling of traffic flow.
Meta-Black-Box Optimization with Bi-Space Landscape Analysis and Dual-Control Mechanism for SAEA
Yukun Du (National University of Defense Technology), Jiang Jiang (National University of Defense Technology)
OptimizationTransformerReinforcement LearningBenchmark
🎯 What it does: This paper proposes DB-SAEA, a dual-control based Meta-Black-Box optimization framework that dynamically adjusts evolutionary algorithms and filling criteria in multi-objective expensive optimization, combined with dual-space exploration landscape analysis and TabPFN surrogate model.
Meta-GAIN for Missing Data Imputation
Tao Tong (University of Electronic Science and Technology of China), Jiangzhang Gan (Hainan University)
ClassificationRestorationMeta LearningAuto EncoderGenerative Adversarial NetworkTabular
🎯 What it does: Proposes Meta-GAIN, a GAN-based missing value imputation method that enhances imputation diversity by achieving continuity in the embedding space through KL divergence, improves generalization using task regularization, and adaptively adjusts loss weights via meta-learning.
Meta-Guided Sample Reweighting for Robust Cross-Modal Hashing Retrieval with Noisy Labels
Ziang Tan (Xidian University), Erkun Yang (Xidian University)
RetrievalMeta LearningContrastive LearningMultimodality
🎯 What it does: Propose a meta-learning-based cross-modal hashing framework called MGSH, which enhances retrieval performance in the presence of label noise.
MetaAct-RL: Training Language Models for Reasoning Through Meta-Action-Based Reinforcement Learning
Zhiheng Xi (Fudan University), Xuanjing Huang (Fudan University)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBenchmark
🎯 What it does: Propose the MetaAct-RL framework, which enables language models to self-select and execute high-level meta-actions (forward reasoning, criticism, and refinement) during inference, optimized via reinforcement learning; simultaneously constructs diverse meta-action trajectories as SFT data and introduces length rewards and key-state restart mechanisms;
MetaDiT: Enabling Fine-grained Constraints in High-degree-of Freedom Metasurface Design
Hao Li (Harbin Engineering University), Andrey Bogdanov (Harbin Engineering University)
GenerationOptimizationTransformerDiffusion modelContrastive LearningImageSequentialPhysics Related
🎯 What it does: Designed a generative framework named MetaDiT to simultaneously optimize all structural parameters in high-freedom metasurfaces while precisely satisfying given high-resolution optical spectrum constraints.
MetaEval: Measuring the Discrimination of Benchmarks for Efficient LLM Evaluation
Zhuo Wang (East China Normal University), Zhenxiao Cheng (East China Normal University)
Computational EfficiencyLarge Language ModelTextBenchmark
🎯 What it does: This paper proposes the MetaEval framework to measure the discriminability of individual questions in evaluation benchmarks and achieve efficient evaluation based on this.
MetaGameBO: Hierarchical Game-Theoretic Driven Robust Meta-Learning for Bayesian Optimization
Hui Li (Beijing Jiaotong University), Jian Yu (Beijing Jiaotong University)
OptimizationMeta LearningBenchmark
🎯 What it does: Propose a hierarchical game-theoretic framework named MetaGameBO to explicitly optimize robustness for extremely difficult tasks in meta-learning with Bayesian optimization, and achieve efficient sample utilization through multi-level sample selection.
MetaGDPO: Alleviating Catastrophic Forgetting with Metacognitive Knowledge Through Group Direct Preference Optimization
Lanxue Zhang (Chinese Academy of Sciences), Yanan Cao (JIUTIAN Research)
OptimizationComputational EfficiencyMeta LearningTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Construct a 5K QA dataset incorporating metacognitive knowledge and propose Group Direct Preference Optimization (GDPO) to compress and fine-tune small LLMs, mitigating catastrophic forgetting.
MetaGPT: A Large Vision-Language Model for Meme Metaphor Understanding
Bo Xu (Dalian University of Technology), Feng Xia (RMIT University)
RecognitionRestorationGenerationExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose MetaGPT, specifically designed for identifying, extracting, and interpreting metaphors in memes, and construct the MUnd dataset containing ≈32k QA pairs.
MetaTrader: Learning to Generalize RL Trading Policies Beyond Offline Data
Haochen Yuan (Shanghai Jiao Tong University), Xiaokang Yang (China International Capital Corporation Limited)
Meta LearningReinforcement LearningTabularTime SeriesFinance Related
🎯 What it does: This paper proposes MetaTrader, a stock trading framework for partial offline reinforcement learning, aiming to enhance the generalization and return balance of trading strategies in non-stationary markets;
Methods for Optimization Problems with Markovian Stochasticity and Non-Euclidean Geometry
Vladimir Solodkin (Moscow Independent Research Institute of Artificial Intelligence), Aleksandr Beznosikov (Moscow Independent Research Institute of Artificial Intelligence)
OptimizationBenchmark
🎯 What it does: This paper proposes utilizing two accelerated first-order methods, Mirror Descent and Mirror-Prox, under any non-Euclidean geometry, specifically addressing smooth convex optimization and variational inequality problems with Markov noise, and provides theoretical convergence rates along with matching lower bounds.
METP: Multi-Granularity Integration of External Covariates for Temporal Point Processes
Boyang Li (Peking University), Xi Zhang (Peking University)
TransformerMixture of ExpertsTime SeriesFinance Related
🎯 What it does: Propose the METP framework, which integrates the lag effects of multi-granularity external covariates into the intensity function of temporal point processes to enhance event time prediction.
Metric Distortion with Preference Intensities
Mehrad Abbaszadeh (Georgia Institute of Technology), Masoud Seddighin (Tehran Institute for Advanced Studies)
Optimization
🎯 What it does: Studied the design of a voting rule named Positional Scoring Matching in ranking voting with preference intensity information, utilizing a metric space distortion framework, and proved that it achieves a metric distortion below 3 when intensity information is forcibly reported;
MF-Speech: Achieving Fine-Grained and Compositional Control in Speech Generation via Factor Disentanglement
Xinyue Yu (University of Science and Technology of China), Song Xiao (Beijing Electronic Science and Technology Institute)
GenerationGenerative Adversarial NetworkContrastive LearningAudio
🎯 What it does: Propose the MF-Speech framework, which first decomposes speech into three highly pure discrete representations of content, timbre, and emotion using a three-stream structure and multi-objective optimization. It then achieves fine-grained controllable combination of these three factors through dynamic fusion and hierarchical style adaptive normalization (HSAN), generating speech that expresses emotion and timbre while maintaining content consistency.
MFINet: Multi-view Fusion and 2D–3D Interaction Enhancement for Real-Time LiDAR Semantic Segmentation
Nan Ma (Beijing University of Technology), Yiheng Han (Beijing University of Technology)
SegmentationAutonomous DrivingComputational EfficiencyConvolutional Neural NetworkTransformerPoint Cloud
🎯 What it does: Propose MFINet, a real-time LiDAR semantic segmentation network based on three-branch multi-perspective fusion and 2D-3D interaction enhancement.
MFmamba: A Multi-function Network for Panchromatic Image Resolution Restoration Based on State-Space Model
Qian Jiang (Yunnan University), Wei Zhou (Yunnan University)
RestorationSuper ResolutionConvolutional Neural NetworkImage
🎯 What it does: Propose a multifunctional network called MFmamba, which can achieve image super-resolution, spectral recovery, and joint recovery of both tasks when only a single panchromatic image is input.
MGD:Mesh-guided Gaussians with Diffusion Priors for Dynamic Objects Reconstruction from Monocular RGB-D Video
Weixing Xie (Xiamen University), Junfeng Yao (Xiamen University)
GenerationDiffusion modelGaussian SplattingImageVideoMultimodalityMesh
🎯 What it does: Proposed a diffusion prior framework based on Mesh-guided Gaussians and depth control (MGD), achieving high-fidelity reconstruction of dynamic objects and filling unobserved regions from monocular RGB-D videos.
MGT-Prism: Enhancing Domain Generalization for Machine-Generated Text Detection via Spectral Alignment
Shengchao Liu (Xi'an Jiaotong University), Shuai Xiao (Xi'an Jiaotong University)
Anomaly DetectionTransformerText
🎯 What it does: Propose a machine-generated text detection method called MGT-Prism based on frequency domain analysis, leveraging low-frequency filtering and spectral alignment to enhance cross-domain generalization performance.
MHA2MLA-VLM: Enabling DeepSeek’s Economical Multi-Head Latent Attention Across Vision-Language Models
Xiaoran Fan (Fudan University), Tao Gui (Fudan University)
CompressionVision Language ModelMultimodality
🎯 What it does: This paper proposes a parameter-efficient, modality-aware MHA2MLA-VLM framework, which migrates existing MHA/GQA-based vision-language models to the Multi-Head Latent Attention (MLA) architecture, achieving substantial KV cache compression while maintaining the original performance.
MHED-SLAM: Multi-Scale Hybrid Encoding-Based Decoupled SLAM
Dengfang Feng (Sun Yat-sen University), Erwei Yin (Academy of Military Sciences)
Neural Radiance FieldSimultaneous Localization and MappingImagePoint Cloud
🎯 What it does: Propose MHED-SLAM, a NeRF-SLAM framework with multi-scale hybrid encoding and decoupled TSDF and density rendering.
MicLog: Towards Accurate and Efficient LLM-based Log Parsing via Progressive Meta In-Context Learning
Jianbo Yu (Southeast University), Wanyuan Wang (Chinese University of Hong Kong Shenzhen)
Meta LearningTransformerLarge Language ModelTextBenchmark
🎯 What it does: Proposed the MicLog framework, enhancing LLM performance in log parsing through Progressive Meta-ICL;
MicroEvoEval: A Systematic Evaluation Framework for Image-Based Microstructure Evolution Prediction
Qinyi Zhang (Sichuan University), Hao Wang (Sichuan University)
Recurrent Neural NetworkTransformerImageBenchmarkPhysics Related
🎯 What it does: This paper proposes and implements MicroEvoEval, a systematic evaluation framework for image-based microstructure evolution prediction, assessing the short-term and long-term prediction performance of 14 deep learning models across four representative PDE tasks.
MIDILM: A Dual-Path Model for Controllable Text-to-MIDI Generation
Shuyu Li (Dongguk University-Seoul), Yunsick Sung (Dongguk University-Seoul)
GenerationData SynthesisTransformerLarge Language ModelMixture of ExpertsTextMultimodalityAudio
🎯 What it does: This paper proposes MIDILM, a dual-path decoder model that achieves controllable generation of text and MIDI.
MIGDiff: Multi-attributes Imputations for Attribute-missing Graphs via Graph Denoising Diffusion Model
Ye Liu (South China University of Technology), Hongmin Cai (South China University of Technology)
RestorationGraph Neural NetworkDiffusion modelAuto EncoderGraph
🎯 What it does: Propose a multi-attribute missing graph attribute imputation method based on a graph denoising diffusion model (MIGDiff).
Mimic-X: A Large-Scale Motion Dataset via Fast Physics-Based Controller Adaptation
Hongyu Tao (Zhejiang University), Weiwei Xu (Zhejiang University)
Data SynthesisPose EstimationReinforcement LearningAuto EncoderVideo
🎯 What it does: This paper proposes an adaptive option framework that rapidly and realistically reconstructs 3D human motion from low-quality videos through hierarchical clustering, option strategies, and dynamic programming, thereby constructing a 52-hour high-quality physically plausible motion dataset called Mimic-X;
Mind the Gap: The Divergence Between Human and LLM-Generated Tasks
Yi-Long Lu (State Key Laboratory of General Artificial Intelligence), Wei Wang (State Key Laboratory of General Artificial Intelligence)
GenerationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Compare differences between humans and large language models (GPT-4o) in generating autonomous tasks, revealing that human tasks are driven by personal values and physical experiences, while LLM-generated tasks lack social and physical aspects, tending toward abstraction.
Mind the Third Eye! Benchmarking Privacy Awareness in MLLM-powered Smartphone Agents
Zhixin Lin, Dongliang Xu (Shandong University)
Safty and PrivacyLarge Language ModelPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Constructed a large-scale privacy-aware benchmark named SAPA-Bench to evaluate the privacy identification and response capabilities of mobile agents.
MindCross: Fast New Subject Adaptation with Limited Data for Cross-subject Video Reconstruction from Brain Signals
Xuan-Hao Liu (Shanghai Jiao Tong University), Wei-Long Zheng (Shanghai Jiao Tong University)
GenerationData SynthesisDomain AdaptationDiffusion modelVideoBiomedical Data
🎯 What it does: This paper proposes the MindCross framework, achieving cross-subject reconstruction of brain signal to video mapping under a single model, and can quickly adapt with only a small amount of new subject data.
MindSight: A Bio-Inspired Neural Architecture for Visual Restoration via Cortical Electrical Stimulation
Yongjie Zou (Lingang Laboratory), Chengyu T. Li
RestorationConvolutional Neural NetworkAuto EncoderBiomedical Data
🎯 What it does: Propose the MindSight framework, which combines deep learning with a differentiable biophysical model to achieve cortical electrical stimulation for restoring visual perception.
MindVote: When AI Meets the Wild West of Social Media Opinion
Xutao Mao (Vanderbilt University), Leyao Wang (Yale University)
Large Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmark
🎯 What it does: Propose the MindVote benchmark, using real social media voting data to predict the distribution of public opinion, addressing the issues of content, cultural, and contextual gaps in traditional questionnaire-based assessments.
Minimizing Inequity in Facility Location Games
Yuhang Guo (University of New South Wales), Houyu Zhou (University of New South Wales)
Optimization
🎯 What it does: This paper studies the game of facility location on the real line, proposing to minimize the 'maximum group effect' under the condition of strategyproofness to achieve group-level fairness;
Minimum-Cost Network Flow with Dual Predictions
Zhiyang Chen, Xia Yin (Tsinghua University)
OptimizationComputational EfficiencyConvolutional Neural NetworkMeshGraph
🎯 What it does: Proposed an algorithm integrating machine learning dual prediction into minimum cost flow solving, along with time complexity and sample complexity analysis.
Minimum-Length Conformal Prediction Sets for Ordinal Classification
Zijian Zhang (Washington State University), Yan Yan (Washington State University)
ClassificationImageTime Series
🎯 What it does: Proposed a model-free shortest covering confidence prediction method (min-CPS) and its length-regularized variant (min-RCPS) for distribution-agnostic uncertainty quantification in ordinal classification.
Minute-Long Videos with Dual Parallelisms
Zeqing Wang (National University of Singapore), Xinchao Wang (National University of Singapore)
GenerationTransformerDiffusion modelVideo
🎯 What it does: Propose a distributed inference strategy named DualParal, combining video sequence parallelism with model layer parallelism, and achieving efficient generation of long videos through block-wise denoising.
MIRA: Evaluating Multimodal AI on Complex Clinical Reasoning in Interventional Radiology
Jingxiong Li (Nanjing University of Science and Technology), Liang Xiao (Nanjing University of Science and Technology)
TransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBiomedical DataMagnetic Resonance ImagingComputed TomographyUltrasoundBenchmark
🎯 What it does: Built and released MIRA—a large-scale multimodal question-answering benchmark designed for interventional radiology, containing 184,479 medical images and approximately 1.2 million expert-generated question-answer pairs, covering open-ended, closed-ended, single-choice, and multiple-choice questions, along with expert-verified reasoning explanations.
MIRAGE: Scaling Test-Time Inference with Parallel Graph-Retrieval-Augmented Reasoning Chains
Kaiwen Wei (Chongqing University), Jiang Zhong (Chongqing University)
Explainability and InterpretabilityGraph Neural NetworkTransformerLarge Language ModelTextGraphBiomedical DataRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the MIRAGE framework, achieving scalable inference during testing in medical QA tasks through parallel multi-chain reasoning and structured knowledge graph retrieval;
MIRAGE: Towards AI-Generated Image Detection in the Wild
OuCheng Huang (Taobao & Tmall Group of Alibaba), Bo Zheng (Taobao & Tmall Group of Alibaba)
Anomaly DetectionLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelImageMultimodalityBenchmark
🎯 What it does: This paper addresses the task of detecting AI-generated images in real-world scenarios by constructing MIRAGE, a wild detection benchmark containing human-planned and multi-model synthesized images, and proposes MIRAGE-R1, a vision-language model capable of adaptive reasoning.
MirrorShield: Towards Dynamic Adaptive Defense Against Jailbreaks via Entropy-Guided Mirror Crafting
Rui Pu (Beijing University of Posts and Telecommunications), Xi Zhang (Beijing University of Posts and Telecommunications)
Adversarial AttackLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Proposed MirrorShield, which dynamically generates mirrors and uses relative input uncertainty (RIU) to defend against jailbreak attacks.
MISF: MLLM Guided Iterative Sample Filtering for Data Fault Detection
Guoying Chen (Beijing Institute of Computer Technology and Application), Kunlong Wang (Beijing Institute of Computer Technology and Application)
Anomaly DetectionData-Centric LearningConvolutional Neural NetworkLarge Language ModelDiffusion modelImage
🎯 What it does: Proposes the MISF (MLLM-Guided Iterative Sample Filtering) framework, which utilizes a multimodal large language model to generate synthetic images and a small number of real clean samples to initialize the detector. Then, it iteratively filters clean samples through Gini uncertainty and prediction consistency to achieve detection of label noise and backdoor attacks in image data.
Mitigating Content Effects on Reasoning in Language Models Through Fine-Grained Activation Steering
Marco Valentino (University of Sheffield), André Freitas (University of Sheffield)
Explainability and InterpretabilityTransformerLarge Language ModelContrastive LearningText
🎯 What it does: This paper dynamically modulates the internal activations of large language models (LLMs) to suppress content bias and enhance formal reasoning accuracy.
Mitigating Endogenous Confirmation Bias in Noisy Label Learning for Vision-Language Models
Feiyang Ning (Harbin Institute of Technology), Xinyang Chen (Harbin Institute of Technology)
Data-Centric LearningPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: To address the self-verification bias caused by pre-trained knowledge in vision-language models during noisy label learning, this paper proposes a multi-stage debiasing framework DKAF, which selects and corrects noisy samples through cross-modal pseudo labels, bimodal consistency, and debiased cross-entropy.
Mitigating Entity Hallucinations in 3D Radiology Report Generation via Dual-Stream Alignment
Lingyu Zhou (Sichuan University), Xiuyuan Xu (Sichuan University)
GenerationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBiomedical DataComputed Tomography
🎯 What it does: By constructing the dual-stream entity alignment network DEAR, the problem of entity misreporting in 3D CT report generation is addressed through fine-grained alignment of organ and lesion entities, significantly reducing hallucinations.
Mitigating Error Accumulation in Co-Speech Motion Generation via Global Rotation Diffusion and Multi-Level Constraints
Xiangyue Zhang (Tongyi Lab, Alibaba Group), Jiaxu Zhang (Nanyang Technological University)
GenerationTransformerDiffusion modelFlow-based ModelAuto EncoderMultimodalityTime Series
🎯 What it does: In full-body speech-driven co-speech action generation, the authors propose GlobalDiff, a framework for diffusion generation in the global joint rotation space;
Mitigating Error Accumulation in Knowledge Editing for Multi-Hop Question Answering
Jiaxin Guo (Peking University), Yan Zhang (Peking University)
RetrievalRepresentation LearningLarge Language ModelTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the Tree of Editing (ToE) framework, leveraging tree structures and retrieval-enhanced knowledge editing methods to address error accumulation and knowledge conflicts in multi-hop question answering;
Mitigating Hallucinations in Large Language Models via Causal Reasoning
Yuangang Li (University of Southern California), Yue Zhao (University of Southern California)
Explainability and InterpretabilityGraph Neural NetworkLarge Language ModelSupervised Fine-TuningTextGraphChain-of-Thought
🎯 What it does: This paper enhances causal inference capabilities and reduces logical hallucinations by first enabling LLMs to construct causal directed acyclic graphs (DAGs) and then performing reasoning on the graphs;
Mitigating Length Bias in RLHF Through a Causal Lens
Hyeonji Kim (Seoul National University), Sanghack Lee (Seoul National University)
Data-Centric LearningReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Propose a counterfactual data augmentation method based on a causal perspective, aiming to reduce length bias in RLHF reward models;
Mitigating Low-Quality Reasoning in MLLMs: Self-Driven Refined Multimodal CoT with Selective Thinking and Step-wise Visual Enhancement
Chongjun Tu (Fudan University), Wanli Ouyang (Shanghai Artificial Intelligent Laboratory)
Computational EfficiencyTransformerVision Language ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Developed a training-agnostic multimodal chain-of-thought framework (SDR-MCoT) that reduces low-quality reasoning and redundant computations in multimodal inference through self-driven selective thinking and incremental visual enhancement.
Mitigating Negative Flips via Margin Preserving Training
Simone Ricci (University of Florence), Alberto Del Bimbo (University of Florence)
ClassificationKnowledge DistillationRepresentation LearningConvolutional Neural NetworkTransformerImage
🎯 What it does: Proposed and studied the Edge-Preserving Training method (MPT) to address negative flips during model updates, which maintains the old class margins by adding bias in softmax and achieves balanced learning of new and old classes through dual-source focal distillation.
Mitigating Noise and Imbalance in Social Governance Graphs for Multi-Type Risk Assessment
Di Jin (Tianjin University), Dongxiao He (Tianjin University)
ClassificationGraph Neural NetworkGraph
🎯 What it does: Proposes a heterogeneous graph neural network framework named HeCoGNN to address issues such as node feature imbalance, noise propagation, and representation bias in social governance graphs;
Mitigating Perception Bias: A Training-Free Approach to Enhance LMM for Image Quality Assessment
Baoliang Chen (South China Normal University), Hanwei Zhu (City University of Macau)
TransformerPrompt EngineeringVision Language ModelImageMultimodality
🎯 What it does: This paper proposes a method without training or fine-tuning, eliminating the bias of Large Multimodal Models (LMMs) in image quality assessment (IQA) tasks by introducing 'conditional images' and 'conditional prompts' in LMMs, thereby improving the accuracy of quality judgment.
Mix-QSAM2: Mixed-Precision Quantization for High Fidelity Segmentation in Resource Constrained Scenarios
Yuzhe Duan (Xidian University), Yanhua Yang (Xidian University)
SegmentationComputational EfficiencyImageVideo
🎯 What it does: Propose an acceleration framework for SAM2, combining importance-driven mixed-precision quantization and Selective Importance-Driven Synthesis (SIS), achieving cooperative optimization of model compression and memory management.
Mixture of Ranks with Degradation-Aware Routing for One-Step Real-World Image Super-Resolution
Xiao He (Xidian University), Xinbo Gao (Xidian University)
Super ResolutionMixture of ExpertsVision Language ModelDiffusion modelGenerative Adversarial NetworkImageMultimodality
🎯 What it does: This paper proposes a sparse expert architecture called Mixture-of-Ranks (MoR), treating each rank of LoRA as an independent expert and combining a degradation estimation module with a degradation-aware load balancing loss to achieve single-step real image super-resolution.
Mixture-of-Trees: Learning to Select and Weigh Reasoning Paths for Efficient LLM Inference
Yangbo Wei (Shanghai Jiao Tong University), Lei He (Eastern Institute of Technology)
Computational EfficiencyPrompt EngineeringMixture of ExpertsTextBenchmarkChain-of-Thought
🎯 What it does: Proposed the Mixture-of-Trees (MoT) framework, combining sparse expert activation with tree-structured reasoning to achieve efficient multi-path inference;
MLLM Enriched Explainable Multiple Clustering
Shan Zhang (Shandong University), Guoxian Yu (George Mason University)
Explainability and InterpretabilityTransformerLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Proposed a multimodal clustering framework called MLLM MC based on Multimodal Large Language Models (MLLM) and Large Language Models (LLM), which utilizes text descriptions to guide clustering and fuses visual features.
MM-R1: Unleashing the Power of Unified Multimodal Large Language Models for Personalized Image Generation
Qian Liang (University of Electronic Science and Technology of China), Ning Xie (University of Electronic Science and Technology of China)
GenerationTransformerLarge Language ModelReinforcement LearningDiffusion modelImageMultimodalityChain-of-Thought
🎯 What it does: Propose the MM-R1 framework, which integrates a unified multimodal large language model with cross-modal Chain-of-Thought reasoning and reward-oriented reinforcement learning to achieve zero-shot personalized image generation.
MM4Rec: Multi-Source and Multi-Scenario Recommender for Unified User Preference
Chu-Chun Yu (National Taiwan University), Che Lin (National Taiwan University)
Recommendation SystemTransformerMixture of ExpertsTextSequential
🎯 What it does: Proposed the MM4Rec framework to achieve unified modeling for multi-source and multi-scenario recommendations.
MMAU-Pro: A Challenging and Comprehensive Benchmark for Holistic Evaluation of Audio General Intelligence
Sonal Kumar (University of Maryland), Ramani Duraiswami (University of Maryland)
BenchmarkRetrieval-Augmented GenerationChain-of-ThoughtAudio
🎯 What it does: Proposes MMAU-Pro, a comprehensive benchmark for audio intelligent evaluation, containing 5,305 expert-annotated question-answer instances, covering speech, environmental sounds, music and their combinations, and examining multidimensional capabilities such as long-duration, multi-audio, spatial, cross-cultural music, and instruction following.
MMBERT: Scaled Mixture-of-Experts Multimodal BERT for Robust Chinese Hate Speech Detection Under Cloaking Perturbations
Qiyao Xue (University of Pittsburgh), Wei Gao (University of Pittsburgh)
ClassificationTransformerMixture of ExpertsVision Language ModelImageTextMultimodalityAudio
🎯 What it does: Proposes MMBERT, a BERT framework integrating text, speech, and visual modalities, utilizing Mixture-of-Experts (MoE) dynamic routing and achieving robust Chinese hate speech detection through a three-stage progressive training approach.
MME-SCI: A Comprehensive and Challenging Science Benchmark for Multimodal Large Language Models
Jiacheng Ruan (Shanghai Jiao Tong University), Yangyang Kang (Zhejiang University)
Large Language ModelPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmarkRetrieval-Augmented Generation
🎯 What it does: Created the MME-SCI multimodal scientific evaluation benchmark, covering four disciplines (mathematics, physics, chemistry, biology), supporting five languages and three input modes, and providing 1,019 high-quality question-answer samples with 63 fine-grained knowledge point annotations.
MMG-Vid: Maximizing Marginal Gains at Segment-level and Token-level for Efficient Video LLMs
Junpeng Ma (Peking University), Shanghang Zhang (Taobao & Tmall Group of Alibaba)
Computational EfficiencyTransformerVision Language ModelVideo
🎯 What it does: Propose a training-agnostic video visual token pruning framework MMG-Vid, which significantly reduces the number of visual tokens in VLLM by maximizing marginal gains at both segment and token levels, thereby improving inference efficiency;
MMG-VL: A Vision-Language Driven Approach for Multi-Person Motion Generation
Songyuan Yang (National University of Defense Technology), Huibin Tan (National University of Defense Technology)
GenerationData SynthesisVision Language ModelDiffusion modelContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: Proposed the MMG-VL framework, which hierarchically generates 3D motion for multiple people based on a single 2D image and natural language instructions. It first uses the Scene-Aware Intent Planner to plan the spatial blueprint, then employs the Coordinated Motion Synthesizer to generate coordinated actions.
MMhops-R1: Multimodal Multi-hop Reasoning
Tao Zhang (Chinese Academy of Sciences), Weiming Hu (Tencent Inc)
RetrievalReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningMultimodalityBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose the MMhops dataset and the MMhops-R1 framework for evaluating and enhancing multimodal multi-hop reasoning capabilities.
MMIFEvol: Towards Evolutionary Multimodal Instruction Following
Haoyu Wang (Fudan University), Yanghua Xiao (Fudan University)
Large Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposed the MMIFEvol framework to evolve multi-modal instructions of varying difficulty levels from a single image and construct corresponding benchmarks;
MMMamba: A Versatile Cross-Modal in Context Fusion Framework for Pan-Sharpening and Zero-Shot Image Enhancement
Yingying Wang (Xiamen University), Haoxuan Che (Xiamen University)
RestorationSuper ResolutionImage
🎯 What it does: Propose a cross-modal context fusion framework called MMMamba based on Mamba for high-resolution multispectral image fusion and zero-shot image enhancement.
MMPG: MoE-based Adaptive Multi-Perspective Graph Fusion for Protein Representation Learning
Yusong Wang (Guangdong Institute of Intelligence Science and Technology), Prayag Tiwari (University College London)
Representation LearningGraph Neural NetworkMixture of ExpertsGraphBiomedical Data
🎯 What it does: Designed and implemented a multi-perspective graph fusion framework called MMPG, constructing protein graphs from three perspectives—physical, chemical, and geometric—and using Mixture-of-Experts for dynamic fusion;
mmPred: Radar-based Human Motion Prediction in the Dark
Junqiao Fan (Nanyang Technological University), Lihua Xie (Nanyang Technological University)
Pose EstimationTransformerDiffusion modelPoint Cloud
🎯 What it does: Propose the mmPred framework, which for the first time applies millimeter-wave radar point clouds to human motion prediction;
MMRAG-RFT: Two-stage Reinforcement Fine-tuning for Explainable Multi-modal Retrieval-augmented Generation
Shengwei Zhao (Xi'an Jiaotong University), Shaoyi Du (Xi'an Jiaotong University)
RetrievalExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose a two-stage reinforcement learning fine-tuning framework for achieving interpretable multi-modal retrieval augmented generation (MMRAG);
Mnemosyne: Accelerating Multi-Hop Question Answering via Cache Hit Order Fitting
Haizhou Du (Shanghai University of Electric Power), Lisheng Wang (Shanghai University of Electric Power)
RetrievalLarge Language ModelText
🎯 What it does: By reordering query sequences to align with cache hit order and employing a multi-granularity entity-document cache, significantly accelerates the retrieval process in multi-hop question answering.
MOBA: A Material-Oriented Backdoor Attack Against LiDAR-Based 3D Object Detection Systems
Saket Sanjeev Chaturvedi (Clemson University), Xiaoyong Yuan (Clemson University)
Autonomous DrivingAdversarial AttackMultimodalityPoint Cloud
🎯 What it does: Proposed a material-based physical backdoor attack (MOBA), which implants realizable triggers into LiDAR point cloud training data, forcing 3D object detection models to produce incorrect predictions when the triggers appear.
MoBGS: Motion Deblurring Dynamic 3D Gaussian Splatting for Blurry Monocular Video
Minh-Quan Viet Bui (Korea Advanced Institute of Science and Technology), Munchurl Kim
RestorationGenerationConvolutional Neural NetworkGaussian SplattingOptical FlowVideoOrdinary Differential Equation
🎯 What it does: Proposes the MoBGS framework, which can recover high-quality, clear novel views from blurry monocular videos.
Mobile-Agent-RAG: Driving Smart Multi-Agent Coordination with Contextual Knowledge Empowerment for Long-Horizon Mobile Automation
Yuxiang Zhou (Sun Yat-sen University), Guanbin Li (OPPO Inc.)
Autonomous DrivingRobotic IntelligenceTransformerLarge Language ModelAgentic AIVision-Language-Action ModelImageTextSequentialBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposes Mobile-Agent-RAG, a hierarchical multi-agent mobile automation framework that employs dual-layer retrieval-enhanced (Manager-RAG and Operator-RAG) modules to address strategic hallucinations in high-level planning and precise execution errors in low-level UI operations separately.
MoCast: Learning Turbulent Motions Under Physical Guidance for Precipitation Nowcasting
Binqing Wu (Zhejiang University), Ling Chen (Zhejiang University)
Convolutional Neural NetworkMixture of ExpertsDiffusion modelVideoPhysics Related
🎯 What it does: MoCast achieves high-accuracy rainfall nowcasting by decomposing turbulent motion into mean and fluctuating components and incorporating physical constraints.
MoCHA: Advanced Vision-Language Reasoning with MoE Connector and Hierarchical Group Attention
Yuqi Pang (Chinese Academy of Sciences), Chen He (Chinese Academy of Sciences)
TransformerMixture of ExpertsVision Language ModelMultimodalityBenchmark
🎯 What it does: MoCHA achieves efficient multimodal reasoning by integrating four visual encoders (CLIP, SigLIP, DINOv2, and ConvNeXt) and introducing a sparse expert connector (MoEC) and hierarchical group attention (HGA).
MODA: The First Challenging Benchmark for Multispectral Object Detection in Aerial Images
Shuaihao Han (Beijing Institute of Technology), Jianan Li (Beijing Institute of Technology)
Object DetectionConvolutional Neural NetworkImageBenchmark
🎯 What it does: This paper proposes the OSSDet framework for object detection tasks in multispectral aerial images and first constructs a large-scale multispectral object detection dataset called MODA.
Modality and Task Adaptation for Enhanced Zero-shot Composed Image Retrieval
Haiwen Li (Beijing University of Posts and Telecommunications), Zhicheng Zhao (Beijing University of Posts and Telecommunications)
RetrievalPrompt EngineeringMixture of ExpertsVision Language ModelContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: Propose a lightweight posterior adaptation framework, MoTa-Adapter, combining automatically generated text-anchored triplets to address task-modal inconsistency in zero-shot compositional image retrieval (ZS-CIR).
Modality-Aware Bias Mitigation and Invariance Learning for Unsupervised Visible-Infrared Person Re-Identification
Menglin Wang, Genlin Ji (Zhejiang University)
RecognitionRetrievalConvolutional Neural NetworkContrastive LearningImageMultimodality
🎯 What it does: For unsupervised visible-infrared person re-identification, a modality-aware Jaccard distance and split contrastive learning framework is proposed to enhance cross-modal association and representation learning.
Modality-Balanced Collaborative Distillation for Multi-Modal Domain Generalization
Xiaohan Wang (University of Electronic Science and Technology of China), Fan Zhou (University of Electronic Science and Technology of China)
Domain AdaptationKnowledge DistillationVideoMultimodalityAudio
🎯 What it does: This paper proposes the MBCD framework, which addresses the problem of over-convergence of dominant modalities caused by weight averaging in multi-modal domain generalization through three modules: adaptive modal dropout, gradient consistency constraints, and cooperative distillation, thereby achieving a flatter loss landscape and more balanced cross-modal fusion.
ModalSyncSum: Synchronizing Image and Text for Reliable Summary Generation
Xuanqi Chen (Guangdong University of Foreign Studies), Shengyi Jiang (Guangdong University of Foreign Studies)
GenerationLarge Language ModelPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: This paper proposes the ModalSyncSum framework, aiming to enhance the semantic consistency and image fidelity of multi-modal summaries by first generating image-text descriptions and performing multi-modal consistency verification.
Model Change for Description Logic Concepts
Ana Ozaki (University of Oslo), Jandson S Ribeiro
🎯 What it does: The study describes model changes in response to pointer interpretation models, including withdrawal (gradual removal of models), reception (addition of models), and revision operations (simultaneous addition and removal);
Model Counting for Dependency Quantified Boolean Formulas
Long-Hin Fung (National Taiwan University), Tony Tan (University of Liverpool)
Benchmark
🎯 What it does: Studied the model counting problem for dependency quantified Boolean formulas (DQBF), proving that even #2-DQBF with only two existential quantifiers remains #EXP-complete, and implemented a specialized counter called sharp2DQR based on BDD.
Model Editing as a Double-Edged Sword: Steering Agent Behavior Toward Beneficence or Harm
Baixiang Huang (Emory University), Kai Shu (Cisco Research)
TransformerLarge Language ModelTextBenchmark
🎯 What it does: Propose to treat the ethical behavior regulation of LLM base agents as a model editing task, and construct a three-tier BEHAVIORBENCH benchmark to systematically evaluate the effectiveness of behavior editing.
Model Whisper: Steering Vectors Unlock Large Language Models’ Potential in Test-Time
Xinyue Kang (Tsinghua University), Li Chen (Tsinghua University)
OptimizationLarge Language ModelPrompt EngineeringText
🎯 What it does: Propose Test-Time Steering Vectors (TTSV), which achieves test-time adaptation and enhances inference performance by adding learnable vectors before input embeddings without modifying LLM parameters.
Model-Agnostic Sentiment Distribution Stability Analysis for Robust LLM-Generated Texts Detection
Siyuan Li (Shanghai Jiao Tong University), Jianhua Li (Shanghai Jiao Tong University)
Anomaly DetectionTransformerLarge Language ModelText
🎯 What it does: Proposed and implemented SentiDetect, a zero-shot LLM text detection framework based on sentiment distribution stability analysis.
Modeling Item-Level Dynamic Variability with Residual Diffusion for Bundle Recommendation
Dong Zhang (Wuhan University of Technology), Jimmy Huang (York University)
Recommendation SystemDiffusion model
🎯 What it does: Proposed a model-agnostic residual diffusion framework RDiffBR, helping bundle recommendation models adapt to item-level dynamic changes.
Modeling Rapid Contextual Learning in the Visual Cortex with Fast-Weight Deep Autoencoder Networks
Yue Li (Carnegie Mellon University), Tai Sing Lee (Carnegie Mellon University)
Representation LearningTransformerAuto EncoderImage
🎯 What it does: The study utilizes Vision Transformer autoencoders combined with low-rank adapters (LoRA) to achieve rapid context learning in early visual layers, and analyzes their impact on representation geometry and attention mechanisms.
Modeling Trend Dynamics with Variational Neural ODEs for Information Popularity Prediction
Yuchen Wang (Northwestern Polytechnical University), Yang Liu (Northwestern Polytechnical University)
Recommendation SystemKnowledge DistillationRecurrent Neural NetworkGraph Neural NetworkTextTime SeriesSequentialOrdinary Differential Equation
🎯 What it does: Developed a model called VNOIP based on variational neural ODEs for predicting the future popularity of information in social networks, combining bidirectional jump ODEs to capture sequential information and model macro trends.
Modeling Uncertainty Trends for Timely Retrieval in Dynamic RAG
Bo Li (Hebei University of Technology), Wei Ye (Hebei University of Technology)
RetrievalTransformerLarge Language ModelTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose a training-free, entropy-trend-based dynamic retrieval trigger mechanism (ETC) to accurately determine when to retrieve external knowledge during the generation process.
Modelling the Effects of Hearing Loss on Neural Coding in the Auditory Midbrain with Variational Conditioning
Lloyd Pellatt (UCL Ear Institute), Nicholas A. Lesica (UCL Ear Institute)
Explainability and InterpretabilityRepresentation LearningConvolutional Neural NetworkAuto EncoderBiomedical DataAudio
🎯 What it does: Propose a variational conditional model ψ-ICNet that learns a low-dimensional representation of hearing loss directly from data using a small number of parameters and predicts neural coding in the central auditory system (midbrain);
Modulation-Based Backdoors: Leveraging Amplitude and Frequency Patterns to Attack Speaker Recognition
Hanbo Cai (Hohai University), Ying Luo (Sun Yat-sen University)
RecognitionAdversarial AttackAudio
🎯 What it does: This paper proposes an acoustic backdoor attack based on frequency modulation (FSMA) and amplitude modulation (ASMA), which deceives speaker recognition models by embedding hidden frequency or amplitude variations into speech without altering semantics.
MoE-Guided Graph Diffusion for Oriented Molecule Design
Shuochen Li (Beihang University), Lei Shi (Beihang University)
Drug DiscoveryGraph Neural NetworkReinforcement LearningMixture of ExpertsDiffusion modelBiomedical Data
🎯 What it does: Proposed a graph diffusion model named MEGD that integrates Mixture of Experts (MoE) for directed molecular design under a reinforcement learning framework.
MoE^2: A Mixture-of-Mixtures of Experts for Ensemble-Free Domain Generalization
Ahmed Radwan (University of British Columbia), Mohamed S. Shehata (University of British Columbia)
ClassificationDomain AdaptationTransformerMixture of ExpertsImage
🎯 What it does: Propose the MoE² framework, which uses a single frozen ViT backbone and dynamically combines lightweight adapter experts to synthesize customized network parameters for each input, achieving domain generalization.
MoEA-Net: Modality-Incremental Expert Aggregation Network for Retinal Prognostic Prediction
Hua Wang (Beijing University of Technology), Xiaobing Yu (Academy of Medical Sciences)
ClassificationTransformerLarge Language ModelMixture of ExpertsVision Language ModelMultimodalityTime SeriesBiomedical Data
🎯 What it does: Designed and implemented a multimodal time-series retinal image prognosis prediction framework called MoEA-Net, capable of predicting long-term changes in retinal diseases related to Age-related Macular Degeneration (AMD) and Diabetic Macular Edema (DME).