AAAI 2024 Papers — Page 3
AAAI Conference on Artificial Intelligence · 2331 papers
BARET: Balanced Attention Based Real Image Editing Driven by Target-Text Inversion
Yuming Qiao (Tsinghua University), Guo-Jun Qi (OPPO Research Institute)
Image TranslationGenerationTransformerDiffusion modelImageText
🎯 What it does: This paper proposes an efficient and controllable method for real image editing called BARET, which only requires the input of the original image and target text.
BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving
Haicheng Liao (University of Macau), Chengzhong Xu (Tsinghua University)
Autonomous DrivingRecurrent Neural NetworkTransformerMultimodalityTime Series
🎯 What it does: A behavior-aware multimodal trajectory prediction model (BAT) is proposed, which combines four modules: behavior, interaction, priority, and location, enabling high-accuracy predictions of surrounding vehicle trajectories.
Batch Normalization Is Blind to the First and Second Derivatives of the Loss
Zhanpeng Zhou (Shanghai Jiao Tong University), Quanshi Zhang (Shanghai Jiao Tong University)
Convolutional Neural NetworkImage
🎯 What it does: The paper systematically analyzes and experimentally demonstrates that batch normalization (BN) blocks the gradient propagation of the first derivative and most second derivatives of the loss function to the parameters of the earlier layers during the training phase, thereby affecting feature learning and model performance through Taylor series expansion and theoretical proof.
Bayesian Inference with Complex Knowledge Graph Evidence
Armin Toroghi (University of Toronto), Scott Sanner (University of Toronto)
Recommendation SystemGraph Neural NetworkGraph
🎯 What it does: This study investigates the use of complex logical evidence for incremental Bayesian inference on knowledge graphs and proposes the BIKG framework.
BBScore: A Brownian Bridge Based Metric for Assessing Text Coherence
Zhecheng Sheng (University of Minnesota), Dongyeop Kang (University of Minnesota)
TransformerLarge Language ModelContrastive LearningText
🎯 What it does: This paper proposes a no-reference text coherence metric called BBScore based on Brownian Bridge theory, and trains a contrastive learning sentence encoder (GPT-2 + MLP) to obtain the trajectory of the text in the latent space of the Brownian Bridge, thereby calculating the coherence score.
BCLNet: Bilateral Consensus Learning for Two-View Correspondence Pruning
Xiangyang Miao (Tongji University), Jun Yu (Hangzhou Dianzi University)
RecognitionObject DetectionConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: BCLNet is proposed, which captures local and global context in parallel through bidirectional consensus learning, achieving the pruning of correspondence between two views.
BDIQA: A New Dataset for Video Question Answering to Explore Cognitive Reasoning through Theory of Mind
Yuanyuan Mao (East China Normal University), Liang He (Shanghai International Studies University)
Data SynthesisRecommendation SystemGraph Neural NetworkTransformerVideoText
🎯 What it does: A BDIQA video question-answering dataset has been constructed, focusing on cognitive reasoning of beliefs, desires, and intentions in videos.
Behavioral Recognition of Skeletal Data Based on Targeted Dual Fusion Strategy
Xiao Yun (China University of Mining and Technology), Patrick Le Callet (Nantes Université)
RecognitionGraph Neural NetworkVideo
🎯 What it does: A lightweight skeleton action recognition model FRF-GCN is proposed, which significantly improves model performance and efficiency by combining bidirectional fusion, targeted adjacency matrices, and a parallel three-dimensional attention mechanism.
BeliefFlow: A Framework for Logic-Based Belief Diffusion via Iterated Belief Change
Nicolas Schwind (National Institute of Advanced Industrial Science and Technology), Pierre Marquis (University of Artois)
Graph
🎯 What it does: This paper proposes and analyzes the Belief Flow Networks (BFNs) framework for modeling logic-based belief propagation in agent networks, proving that consensus can be reached in strongly connected networks and providing a polynomial-time optimal 'buy-sell' scheme.
Benchmarking Large Language Models in Retrieval-Augmented Generation
Jiawei Chen (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences), Le Sun (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences)
GenerationRetrievalTransformerLarge Language ModelTextMultimodalityBenchmarkRetrieval-Augmented Generation
🎯 What it does: This paper constructs a benchmark called RGB for retrieval-augmented generation (RAG) to systematically evaluate the performance of large language models (LLMs) on four core capabilities: noise robustness, negation rejection, information integration, and counterfactual robustness.
Benchmarking Large Language Models on Controllable Generation under Diversified Instructions
Yihan Chen (University of Science and Technology of China), Zhendong Mao (University of Science and Technology of China)
GenerationTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: A benchmark called CoDI-Eval is proposed to evaluate the controllable text generation capabilities of large language models under natural language instructions.
Better than Random: Reliable NLG Human Evaluation with Constrained Active Sampling
Jie Ruan (Peking University), Yuesheng Zhu (Peking University)
GenerationTextMultimodality
🎯 What it does: This paper studies the sampling problem in natural language generation (NLG) evaluation and proposes a Constrained Active Sampling Framework (CASF). Through the collaboration of a learner, a system sampler, and a constrained controller, it selects representative samples for manual evaluation, thereby obtaining more reliable system rankings.
BEV-MAE: Bird’s Eye View Masked Autoencoders for Point Cloud Pre-training in Autonomous Driving Scenarios
Zhiwei Lin (Peking University), Ming-Hsuan Yang (University of California)
Object DetectionAutonomous DrivingAuto EncoderPoint Cloud
🎯 What it does: A self-supervised pre-training framework based on Bird's Eye View Masked Autoencoder (BEV-MAE) is proposed for 3D object detection of LiDAR point clouds in autonomous driving scenarios.
Beyond Attention: Breaking the Limits of Transformer Context Length with Recurrent Memory
Aydar Bulatov (Moscow Institute of Physics and Technology), Mikhail Burtsev (London Institute for Mathematical Sciences)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes the Recurrent Memory Transformer (RMT), which inserts learnable memory tokens into the input/output of the Transformer and implements recursive connections at the segment level, allowing for a linear expansion of the input context length up to 2 million tokens.
Beyond Entities: A Large-Scale Multi-Modal Knowledge Graph with Triplet Fact Grounding
Jingping Liu (East China University of Science and Technology), Yunwen Chen (DataGrand Inc.)
Contrastive LearningImageTextMultimodality
🎯 What it does: This paper constructs ImgFact, a multimodal knowledge graph based on images, which records 247,732 triple facts and 3,730,805 related images.
Beyond Expected Return: Accounting for Policy Reproducibility When Evaluating Reinforcement Learning Algorithms
Manon Flageat (Imperial College London), Antoine Cully (Imperial College London)
Reinforcement Learning
🎯 What it does: This study investigates the policy replicability of reinforcement learning in uncertain environments and proposes using a low confidence bound to evaluate the trade-off between performance and replicability.
Beyond Grounding: Extracting Fine-Grained Event Hierarchies across Modalities
Hammad Ayyubi (Columbia University), Shih-Fu Chang (Columbia University)
ClassificationSegmentationRetrievalTransformerContrastive LearningVideoTextMultimodality
🎯 What it does: This paper proposes the task of extracting fine-grained event hierarchical structures from two modalities: video and text, and constructs a new MultiHiEve dataset. It introduces a weakly supervised model called MASHER to accomplish this task.
Beyond Mimicking Under-Represented Emotions: Deep Data Augmentation with Emotional Subspace Constraints for EEG-Based Emotion Recognition
Zhi Zhang (Hong Kong Polytechnic University), Yan Liu (Hong Kong Polytechnic University)
ClassificationRecognitionData SynthesisGenerative Adversarial NetworkTime SeriesBiomedical Data
🎯 What it does: Designed and implemented an Emotion Subspace Constrained Generative Adversarial Network (ESC-GAN) for data augmentation of imbalanced emotional categories in EEG emotion recognition, thereby improving classification accuracy.
Beyond OOD State Actions: Supported Cross-Domain Offline Reinforcement Learning
Jinxin Liu (Westlake University), Donglin Wang (Westlake University)
Robotic IntelligenceReinforcement LearningTabular
🎯 What it does: This study focuses on cross-domain offline reinforcement learning and proposes the BOSA method, which addresses the OOD state-action and transition dynamics problem through supported constraint policy and value optimization.
Beyond Prototypes: Semantic Anchor Regularization for Better Representation Learning
Yanqi Ge (Shenzhen Institute for Advanced Study University of Electronic Science and Technology of China), Lixin Duan (Shenzhen Institute for Advanced Study University of Electronic Science and Technology of China)
SegmentationRepresentation LearningImage
🎯 What it does: A Semantic Anchor Regularization (SAR) is designed to guide feature learning through predefined class anchors, reducing prototype learning bias and enhancing inter-class separation and intra-class compactness.
Beyond the Label Itself: Latent Labels Enhance Semi-supervised Point Cloud Panoptic Segmentation
Yujun Chen (East China Normal University), Yuan Xie (Xiamen University)
Object DetectionSegmentationAutonomous DrivingPoint Cloud
🎯 What it does: To address the semi-supervised point cloud panoptic segmentation problem, this paper proposes the use of hidden 'latent labels' to enhance model performance. Specifically, it employs CylinderMix data augmentation in the LiDAR branch and uses the IPSL module to learn instance location and scale information in the image branch, integrating both into a multi-modal BEV network.
Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles
Maximilian Muschalik (LMU Munich), Eyke Hüllermeier
Explainability and InterpretabilityComputational EfficiencyTabularBenchmarkFinance Related
🎯 What it does: Proposes TreeSHAP-IQ, an efficient algorithm for computing Shapley interaction of arbitrary order for tree models;
Bi-directional Adapter for Multimodal Tracking
Bing Cao (Tianjin University), Qinghua Hu (Tianjin University)
Object TrackingTransformerImageVideoMultimodality
🎯 What it does: This paper proposes the Bi-directional Adapter for Multi-modal Tracking (BAT) framework, which utilizes a lightweight bidirectional adapter to achieve dynamic information mutual promotion and fusion between RGB and infrared (TIR) dual modalities within a pre-trained Transformer backbone.
Bi-ViT: Pushing the Limit of Vision Transformer Quantization
Yanjing Li (Beihang University), Baochang Zhang (Zhongguancun Laboratory)
ClassificationObject DetectionComputational EfficiencyKnowledge DistillationTransformerImage
🎯 What it does: This paper fully binarizes the Vision Transformer and proposes the Bi-ViT method.
Bias-Conflict Sample Synthesis and Adversarial Removal Debias Strategy for Temporal Sentence Grounding in Video
Zhaobo Qi (Harbin Institute of Technology), Qingming Huang (University of Chinese Academy of Sciences)
RecognitionData SynthesisGenerative Adversarial NetworkVideoText
🎯 What it does: This paper proposes a bias elimination framework based on adversarial training called BSSARD, which dynamically synthesizes bias conflict samples and breaks the pseudo-correlation between temporal position and semantics through a visual and textual bias generator to enhance the generalization ability of video sentence localization.
Bidirectional Temporal Plan Graph: Enabling Switchable Passing Orders for More Efficient Multi-Agent Path Finding Plan Execution
Yifan Su (Carnegie Mellon University), Jiaoyang Li (Carnegie Mellon University)
OptimizationComputational EfficiencyRobotic IntelligenceGraph
🎯 What it does: This paper proposes the Bidirectional Temporal Plan Graph (BTPG), which allows for dynamic switching of the passing order at conflict points during execution to reduce meaningless waiting and improve the execution efficiency of multi-agent path planning.
Big Learning Expectation Maximization
Yulai Cong (Sun Yat-sen University), Sijia Li (Sun Yat-sen University)
OptimizationTabular
🎯 What it does: A BigLearn-EM algorithm based on the principle of large learning is proposed, improving the EM training of GMM and overcoming the issue of poor local optima.
Bilateral Gradual Semantics for Weighted Argumentation
Zongshun Wang (Sun Yat-sen University), Yuping Shen (Sun Yat-sen University)
Graph Neural NetworkGraph
🎯 What it does: This study investigates the introduction of non-recursive rejection degrees in weighted argumentation graphs and proposes a bilateral asymptotic semantics.
Binding-Adaptive Diffusion Models for Structure-Based Drug Design
Zhilin Huang (Tsinghua University), Wenming Yang (Peng Cheng Laboratory)
Drug DiscoveryGraph Neural NetworkDiffusion modelGraphBiomedical Data
🎯 What it does: A binding adaptive diffusion model based on BINDDM is proposed for structured drug design, which automatically extracts key sub-complexes from protein-ligand complexes and enhances the generation process during diffusion.
BiPFT: Binary Pre-trained Foundation Transformer with Low-Rank Estimation of Binarization Residual Polynomials
Xingrun Xing (Institute of Automation, Chinese Academy of Sciences), Jiajun Zhang (Institute of Automation, Chinese Academy of Sciences)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes the first binary pre-trained foundational transformer for natural language understanding tasks, BiPFT, which significantly reduces computational load and memory usage by enhancing self-attention performance through low-rank estimation of binary residual polynomials.
BLADE: Box-Level Supervised Amodal Segmentation through Directed Expansion
Zhaochen Liu (Peking University), Tingting Jiang (Peking University)
Object DetectionSegmentationConvolutional Neural NetworkImage
🎯 What it does: A BLADE framework is proposed to achieve box-level supervised full-modal segmentation by using only the target's bounding box and category labels.
Blind Face Restoration under Extreme Conditions: Leveraging 3D-2D Prior Fusion for Superior Structural and Texture Recovery
Zhengrui Chen (Beijing University of Posts and Telecommunications), Weihong Deng (Beijing University of Posts and Telecommunications)
RestorationGenerative Adversarial NetworkImage
🎯 What it does: A blind facial restoration method named FREx is proposed, aimed at reconstructing high-quality facial images from extremely degraded inputs.
BLiRF: Bandlimited Radiance Fields for Dynamic Scene Modeling
Sameera Ramasinghe (Amazon), Anton van den Hengel (Amazon)
GenerationData SynthesisNeural Radiance FieldImage
🎯 What it does: A framework called BLiRF is proposed, which models the lighting field and density field of dynamic scenes as band-limited high-dimensional signals, achieving complete separation of lighting and density.
BLIVA: A Simple Multimodal LLM for Better Handling of Text-Rich Visual Questions
Wenbo Hu (University of California), Zhuowen Tu (University of California)
RecognitionGenerationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: We propose BLIVA, a multimodal large language model that integrates learned query embeddings with patch embeddings from an image encoder, enabling better understanding of textual information within images.
Block Image Compressive Sensing with Local and Global Information Interaction
Xiaoyu Kong (Harbin Institute of Technology), Zhenyu He (Southern University of Science and Technology)
RestorationCompressionConvolutional Neural NetworkTransformerImage
🎯 What it does: A block-level image compression sensing method called BRBCN is proposed, which combines CNN local reconstruction with Transformer global communication to achieve bidirectional information interaction both within and between blocks.
Block-Level Goal Recognition Design
Tsz-Chiu Au (Ulsan National Institute of Science and Technology)
RecognitionOptimizationTabular
🎯 What it does: This paper proposes a block-based Goal Recognition Design (GRD) framework and constructs a hierarchical design model to cluster multiple modifications; subsequently, an improved version of the pruned-reduce and design subtree pruning rules is designed, and a solver based on breadth-first search (BFS) and local search is implemented.
BOK-VQA: Bilingual outside Knowledge-Based Visual Question Answering via Graph Representation Pretraining
MinJun Kim (Hanbat National University), KyungTae Lim (Kakao Brain)
Convolutional Neural NetworkGraph Neural NetworkVision Language ModelImageTextGraph
🎯 What it does: A bilingual external knowledge visual question answering (BOK-VQA) dataset was constructed, and a multi-task model (GEL-VQA) was proposed to predict knowledge triples based on image + text and embed knowledge into VQA.
Boosting Adversarial Transferability across Model Genus by Deformation-Constrained Warping
Qinliang Lin (Shenzhen University), Siyang Song (University of Leicester)
Adversarial AttackConvolutional Neural NetworkTransformerImageVideoMultimodalityAudio
🎯 What it does: This paper proposes an input transformation attack method based on elastic constraint deformation (DeCoW) called DeCoWA, which utilizes max-min optimization combined with multi-deformation averaging to enhance the transferability of adversarial samples across model families (such as CNN and ViT) and across modalities (images, videos, audio).
Boosting Few-Shot Learning via Attentive Feature Regularization
Xingyu Zhu (University of Science and Technology of China), Xiangnan He (Brain-Inspired Technology Co., Ltd.)
ClassificationMeta LearningConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes an attention-based feature regularization method (AFR), which first selects relevant samples from the base class using semantic similarity, and then applies instance attention and channel attention at the feature level for adaptive mixing and recalibration of features, ultimately training a classifier to achieve few-shot learning.
Boosting Multiple Instance Learning Models for Whole Slide Image Classification: A Model-Agnostic Framework Based on Counterfactual Inference
Weiping Lin (Xiamen University), Liansheng Wang (University of Hong Kong)
ClassificationImage
🎯 What it does: A model-agnostic MIL enhancement framework CIMIL is proposed, which obtains high-quality pseudo-labels through counterfactual inference sub-bag evaluation and hierarchical instance search, trains an instance classifier, and uses its embeddings as prompts to improve the original features, thereby enhancing the bag-level and instance-level classification performance of WSI.
Boosting Neural Cognitive Diagnosis with Student’s Affective State Modeling
Shanshan Wang (Anhui University), Xingyi Zhang (Anhui University)
Contrastive LearningTabular
🎯 What it does: Affective Cognitive Diagnosis Model (ACD) is proposed, which enhances the accuracy of cognitive diagnosis by predicting students' emotional states during the answering process and converting them into personalized guessing and slipping parameters.
Boosting Residual Networks with Group Knowledge
Shengji Tang (Fudan University), Wanli Ouyang (Shanghai AI Laboratory)
ClassificationObject DetectionSegmentationKnowledge DistillationConvolutional Neural NetworkTransformerImage
🎯 What it does: A training framework based on grouped knowledge (GKT) is proposed to enhance the performance of residual networks through the aggregation and transfer of hierarchical subnetwork groups.
Bootstrapping Cognitive Agents with a Large Language Model
Feiyu Zhu (Carnegie Mellon University), Reid Simmons (Carnegie Mellon University)
Explainability and InterpretabilityRobotic IntelligenceTransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: Combining general knowledge generated by LLMs with cognitive architecture to quickly train interpretable cognitive agents for kitchen tasks from scratch.
Bootstrapping Large Language Models for Radiology Report Generation
Chang Liu (University of Science and Technology of China), Yongdong Zhang (University of Science and Technology of China)
GenerationDomain AdaptationTransformerLarge Language ModelContrastive LearningTextMultimodalityElectronic Health RecordsRetrieval-Augmented Generation
🎯 What it does: This paper proposes an adaptive approach to general large language models for generating radiology reports through domain-specific instance induction and a two-stage decoding process.
Box2Poly: Memory-Efficient Polygon Prediction of Arbitrarily Shaped and Rotated Text
Xuyang Chen (Huawei), Liqiu Meng (Technical University of Munich)
RecognitionObject DetectionComputational EfficiencyTransformerImage
🎯 What it does: This paper proposes an iterative polygon regression framework Box2Poly based on Sparse R‑CNN, which gradually regresses to text polygons using learnable box priors, and introduces Bézier curves as an intermediate representation along with PolyAlign for efficient RoI feature extraction.
Bridging the Gap between 2D and 3D Visual Question Answering: A Fusion Approach for 3D VQA
Wentao Mo (Peking University), Yang Liu (Peking University)
RetrievalRepresentation LearningTransformerVision Language ModelPoint Cloud
🎯 What it does: BridgeQA is proposed to enhance 3D Visual Question Answering (VQA) by combining 2D view selection based on question conditions and a dual-branch Transformer that integrates 2D and 3D visual information.
Bridging the Semantic Latent Space between Brain and Machine: Similarity Is All You Need
Jiaxuan Chen (Zhejiang University), Gang Pan (Zhejiang University)
ClassificationRetrievalTransformerContrastive LearningMultimodalityBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A self-supervised brain-CLIP semantic alignment network called BrainSem was designed and trained, which extracts a shared semantic latent space from a small amount of fMRI recordings through fuzzy one-to-many matching, achieving zero-shot brain image retrieval and classification.
Brush Your Text: Synthesize Any Scene Text on Images via Diffusion Model
Lingjun Zhang (East China Normal University), Yu Qiao (Shanghai Artificial Intelligence Laboratory)
GenerationData SynthesisDiffusion modelImageText
🎯 What it does: A training-free Diff-Text framework is proposed, capable of generating text images for any language scene by utilizing pre-rendered sketches and edge maps as control conditions, combined with text prompts to generate realistic scene images.
Building Minimal and Reusable Causal State Abstractions for Reinforcement Learning
Zizhao Wang (University of Texas at Austin), Peter Stone (George Mason University)
Robotic IntelligenceReinforcement LearningContrastive Learning
🎯 What it does: In multi-task reinforcement learning, a framework based on Causal Bijection Modeling (CBM) is proposed, which utilizes causal dynamics and a reward model to automatically learn minimal and task-specific state abstractions, significantly improving sample efficiency and generalization ability.
Building Variable-Sized Models via Learngene Pool
Boyu Shi (Southeast University), Xin Geng (Southeast University)
Computational EfficiencyKnowledge DistillationTransformerSupervised Fine-TuningImage
🎯 What it does: Proposes the Learngene Pool method, which constructs variable-sized networks using learngene instances distilled from a single large model, addressing issues of storage consumption, minimum model size limitations, and insufficient initialization of stitch layers in SN-Net.
BVT-IMA: Binary Vision Transformer with Information-Modified Attention
Zhenyu Wang (Xidian University), Guangming Shi (Alibaba Group)
ClassificationComputational EfficiencyKnowledge DistillationTransformerImage
🎯 What it does: An Information Modification Attention (IMA) is proposed and implemented to compensate for the attention shift and information loss caused by binarization in binary Vision Transformers.
Cached Transformers: Improving Transformers with Differentiable Memory Cachde
Zhaoyang Zhang (Chinese University of Hong Kong), Ping Luo (University of Hong Kong)
Object DetectionSegmentationGenerationTransformerImageText
🎯 What it does: This paper proposes the Cached Transformer, which incorporates a Gated Recurrent Cache (GRC) into self-attention with a differentiable memory cache to extend the attention range and capture long-range dependencies.
Calibrated One Round Federated Learning with Bayesian Inference in the Predictive Space
Mohsin Hasan (University of Waterloo), Pascal Poupart (University of Waterloo)
Federated LearningKnowledge DistillationTabular
🎯 What it does: A single-round federated learning algorithm is proposed, utilizing Bayesian inference to aggregate local predictive posteriors in the prediction space for model calibration and performance improvement under heterogeneous data.
CAMEL: Capturing Metaphorical Alignment with Context Disentangling for Multimodal Emotion Recognition
Linhao Zhang (Chinese Academy of Sciences), Haonan Liu (Tiangong University)
RecognitionTransformerContrastive LearningTextMultimodality
🎯 What it does: This paper proposes a multi-modal emotion recognition framework called CAMEL, which is based on conditional generation and context decoupling, specifically designed to capture the impact of implicit metaphors on emotions in multi-modal content.
CaMIL: Causal Multiple Instance Learning for Whole Slide Image Classification
Kaitao Chen (East China Normal University), Jing Zhao (East China Normal University)
ClassificationConvolutional Neural NetworkTransformerImageBiomedical Data
🎯 What it does: This paper proposes a causal inference-based multi-instance learning framework called CaMIL, which uses front-door adjustment methods to eliminate confounding associations in Whole Slide Image classification, and achieves pluggable predictions through instance buffering, clustering, cross-attention, and NWGM approximate expectations.
CamoDiffusion: Camouflaged Object Detection via Conditional Diffusion Models
Zhongxi Chen (Xiamen University), Xianming Lin (Xiamen University)
Object DetectionTransformerDiffusion modelImage
🎯 What it does: This paper proposes a covert target detection method called CamoDiffusion based on a conditional diffusion model, which addresses the issues of boundary blurriness and overconfidence misjudgment by generating masks through progressive denoising.
Can Large Language Models Serve as Rational Players in Game Theory? A Systematic Analysis
Caoyun Fan (Shanghai Jiao Tong University), Hao He (Shanghai Jiao Tong University)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper systematically analyzes the capability boundaries of large language models (LLMs) in game theory, assessing whether they can act as rational game players, and tests the desire construction, belief refinement, and optimal action capabilities of LLMs in three classic games (Dictator Game, Rock-Paper-Scissors, Ring-Network Game).
Can Large Language Models Understand Real-World Complex Instructions?
Qianyu He (Fudan University), Yanghua Xiao (Fudan University)
TransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: A benchmark named CELLO has been designed and released to systematically evaluate the understanding capabilities of large language models for complex instructions.
Can LLMs Fix Issues with Reasoning Models? Towards More Likely Models for AI Planning
Turgay Caglar (Colorado State University), Sarath Sreedharan (IBM Research)
TransformerLarge Language Model
🎯 What it does: This paper studies the use of large language models (LLMs) for the model space editing task in automated planning, exploring their effectiveness in addressing unsolvability and plan executability issues.
Can You Rely on Synthetic Labellers in Preference-Based Reinforcement Learning? It’s Complicated
Katherine Metcalf (Apple), Barry-John Theobald (Apple)
Reinforcement Learning
🎯 What it does: Compare the effects of human and synthetic labels in preference reinforcement learning, verify whether synthetic labels can replace human labels, and analyze the relationship between label consistency and policy performance.
CAR-Transformer: Cross-Attention Reinforcement Transformer for Cross-Lingual Summarization
Yuang Cai (Beijing University of Posts and Telecommunications), Yuyu Yuan (Beijing University of Posts and Telecommunications)
TransformerReinforcement LearningText
🎯 What it does: Designed and implemented the Cross-Attention Reinforcement (CAR) module, using the cross-attention of the Transformer as a policy to align and enhance cross-lingual summarization performance through reinforcement learning (policy gradient).
CARAT: Contrastive Feature Reconstruction and Aggregation for Multi-Modal Multi-Label Emotion Recognition
Cheng Peng (Zhejiang University), Gang Chen (Zhejiang University)
RecognitionTransformerContrastive LearningMultimodality
🎯 What it does: The CARAT framework is proposed, which achieves multi-label emotion recognition through the reconstruction fusion and feature aggregation of multimodal information.
Catalyst for Clustering-Based Unsupervised Object Re-identification: Feature Calibration
Huafeng Li (Kunming University of Science and Technology), Zhanxuan Hu (Yunnan Normal University)
RecognitionRepresentation LearningGraph Neural NetworkContrastive LearningImage
🎯 What it does: A framework is proposed and validated that inserts a Feature Calibration Module (FCM) into unsupervised object re-identification based on clustering, enhancing the synergy between pseudo-label generation and representation learning.
Catch-Up Mix: Catch-Up Class for Struggling Filters in CNN
Minsoo Kang (Korea Institute of Science and Technology), Suhyun Kim (Korea Institute of Science and Technology)
ClassificationAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Proposes the Catch-up Mix feature-level mixing method to address the issue of CNNs' over-reliance on a few filters, giving slow-learning filters more training opportunities.
CatFormer: Category-Level 6D Object Pose Estimation with Transformer
Sheng Yu (Beijing Institute of Technology), Yuanqing Xia (Beijing Institute of Technology)
Pose EstimationTransformerPoint Cloud
🎯 What it does: This paper proposes a category-level 6D object pose estimation framework based on Transformer, called CatFormer, which includes coarse and fine deformation modules, a graphical module, and a recursive refinement module, capable of accurately recovering the target's NOCS model and estimating its pose from RGB-D input;
CatmullRom Splines-Based Regression for Image Forgery Localization
Li Zhang (Hefei Institute of Physical Science, Chinese Academy of Sciences), Rujing Wang (Hefei Institute of Physical Science, Chinese Academy of Sciences)
SegmentationAnomaly DetectionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a Catmull-Rom spline regression-based image forgery localization network, CSR-Net, which combines Comprehensive Re-scoring (CRA) and Vertical Texture Interaction Perception (VTP) modules to achieve pixel-level precise localization of forged areas and suppress false positives.
Causal Adversarial Perturbations for Individual Fairness and Robustness in Heterogeneous Data Spaces
Ahmad-Reza Ehyaei (Max Planck Institute for Intelligent Systems), Golnoosh Farnadi (Mila Quebec AI Institute)
ClassificationOptimizationAdversarial AttackGenerative Adversarial NetworkTabular
🎯 What it does: A framework that combines individual fairness, adversarial robustness, and causal structure is proposed, defining CAPI fairness and causal adversarial perturbations (CAP), and achieving a classifier that satisfies all three through CAPIFY training.
Causal Discovery from Poisson Branching Structural Causal Model Using High-Order Cumulant with Path Analysis
Jie Qiao (Guangdong University of Technology), Zhifeng Hao (Shantou University)
Tabular
🎯 What it does: This paper proposes and studies a causal model based on Poisson branching structure, PB-SCM, and achieves identifiable causal structure learning for count data through higher-order cumulants and path analysis.
Causal Representation Learning via Counterfactual Intervention
Xiutian Li (Fudan University), Rui Feng (Fudan University)
Representation LearningAuto EncoderTabular
🎯 What it does: This paper proposes a VAE framework based on counterfactual interventions (CFI-VAE), which eliminates the induced bias and dataset bias introduced in causal graphs to learn unbiased causal effects and inject them into latent representations, achieving causally separable representations.
Causal Strategic Learning with Competitive Selection
Kiet Q. H. Vo (CISPA Helmholtz Center for Information Security), Krikamol Muandet (CISPA Helmholtz Center for Information Security)
Tabular
🎯 What it does: This study investigates causal policy learning in a multi-decision-maker environment, proposing a competitive selection framework and designing an optimal selection rule that balances optimization and incentives during the selection process. It also introduces a cooperation protocol and Mean Shift Linear Regression (MSLR) to achieve unbiased causal parameter estimation.
Causal Walk: Debiasing Multi-Hop Fact Verification with Front-Door Adjustment
Congzhi Zhang (Southeast University), Deyu Zhou (Southeast University)
Recurrent Neural NetworkGraph Neural NetworkTextGraph
🎯 What it does: This paper proposes a causal inference method based on front-door adjustment called Causal Walk, aimed at debiasing multi-hop fact verification.
Causal-Driven Skill Prerequisite Structure Discovery
Shenbao Yu (Xiamen University), Yinghui Pan (Shenzhen University)
Gaussian SplattingTabular
🎯 What it does: A two-stage framework CSPS based on causal structure learning is proposed to infer the prerequisite relationships of skills from students' practice scores.
Causality-Inspired Invariant Representation Learning for Text-Based Person Retrieval
Yu Liu (Jilin University), Xun Yang (University of Science and Technology of China)
RetrievalRepresentation LearningContrastive LearningTextMultimodality
🎯 What it does: A transferable and robust text-to-person retrieval method IRLT is proposed from a causal perspective. By incorporating a style interpolator and a scene simulator into the image encoder, it learns independent and sufficient causal visual representations, thereby achieving cross-modal alignment.
Cautiously-Optimistic Knowledge Sharing for Cooperative Multi-Agent Reinforcement Learning
Yanwen Ba (Hunan University), Shigeng Zhang (Central South University)
Reinforcement LearningSequential
🎯 What it does: This paper proposes a cautious optimistic knowledge sharing framework called CONS, which is used for sharing positive and negative experiences and performing soft policy updates and targeted exploration in decentralized training and execution of multi-agent reinforcement learning.
CAVEN: An Embodied Conversational Agent for Efficient Audio-Visual Navigation in Noisy Environments
Xiulong Liu (University of Washington), Anoop Cherian (Mitsubishi Electric Research Labs)
Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningMultimodalityAudio
🎯 What it does: A bidirectional natural language dialogue audio-visual navigation framework called CAVEN is proposed, capable of interacting with humans/oracles;
CcDPM: A Continuous Conditional Diffusion Probabilistic Model for Inverse Design
Yanxuan Zhao (China Aerodynamics Research and Development Center), Yueqing Wang (China Aerodynamics Research and Development Center)
GenerationOptimizationDiffusion modelTabular
🎯 What it does: A continuous condition diffusion probability model (CcDPM) is proposed, which combines continuous conditions with vicinal risk minimization (VRM), and accelerates the calculation of vicinal loss using k-d trees, while designing a multi-point sampling scheme to support inverse design under multiple working conditions.
CDPNet: Cross-Modal Dual Phases Network for Point Cloud Completion
Zhenjiang Du (University of Electronic Science and Technology of China), Yang Yang (University of Electronic Science and Technology of China)
RestorationGenerationConvolutional Neural NetworkGraph Neural NetworkImagePoint Cloud
🎯 What it does: This paper proposes CDPNet, which utilizes cross-modal information from single-view images and partial point clouds to achieve global completion and local detail reconstruction of 3D point clouds through a two-stage network.
Ced-NeRF: A Compact and Efficient Method for Dynamic Neural Radiance Fields
Youtian Lin (Nanjing University Harbin Institute of Technology)
GenerationData SynthesisComputational EfficiencyNeural Radiance FieldVideo
🎯 What it does: We propose Ced-NeRF, a dynamic NeRF framework that combines high-resolution meshes with deformation networks, capable of achieving fast training and high-quality rendering with only a minimal number of additional parameters.
CEDFlow: Latent Contour Enhancement for Dark Optical Flow Estimation
Fengyuan Zuo (Xi'an University of Technology), Haonan Su (Xi'an University of Technology)
Convolutional Neural NetworkOptical FlowImageVideo
🎯 What it does: Proposes the CEDFlow framework for achieving high-precision optical flow estimation under low-light conditions.
CEGAR-Based Approach for Solving Combinatorial Optimization Modulo Quantified Linear Arithmetics Problems
Kerian Thuillier (University of Rennes), Loïc Paulevé (University of Bordeaux)
OptimizationBiomedical Data
🎯 What it does: This paper proposes a method based on Counter-Example-Guided Abstraction Refinement (CEGAR) to solve combinatorial optimization problems (OPT+qLP) that involve both Boolean logic and quantified linear constraints, and implements an ASP solver extension named MERRINASP;
Cell Graph Transformer for Nuclei Classification
Wei Lou (Shenzhen Research Institute of Big Data), Haofeng Li (Shenzhen Research Institute of Big Data)
ClassificationSegmentationGraph Neural NetworkTransformerImageGraph
🎯 What it does: This paper proposes the Cell Graph Transformer (CGT), which classifies cell nucleus types by using both the cell nucleus and edges as input tokens and leveraging learnable adjacency relationships.
CF-NeRF: Camera Parameter Free Neural Radiance Fields with Incremental Learning
Qingsong Yan (Wuhan University), Fei Deng (Wuhan University)
GenerationOptimizationNeural Radiance FieldSimultaneous Localization and MappingOptical FlowImageVideo
🎯 What it does: A NeRF method called CF-NeRF is proposed, which does not require pre-defined camera parameters. It incrementally learns to estimate camera parameters frame by frame and constructs a 3D scene.
CFEVER: A Chinese Fact Extraction and VERification Dataset
Ying-Jia Lin (National Cheng Kung University), Hung-Yu Kao (National Cheng Kung University)
RetrievalTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Created the largest Chinese fact extraction and verification dataset CFEVER, which contains 30,012 manually annotated propositions based on Chinese Wikipedia and corresponding evidence.
CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation
Shoukun Sun (University of Idaho), Tiankai Yao (Idaho National Laboratory)
SegmentationTransformerSupervised Fine-TuningImage
🎯 What it does: An interactive image segmentation framework is proposed, which includes Iterative Click Loss (ICL), Cascade Forward Refinement (CFR), and SUEM Copy-Paste data augmentation.
CGMGM: A Cross-Gaussian Mixture Generative Model for Few-Shot Semantic Segmentation
Junao Shen (Zhejiang University), Wei Zhang (Zhejiang University)
SegmentationConvolutional Neural NetworkMixture of ExpertsImage
🎯 What it does: This paper proposes a Cross Gaussian Mixture Generative Model (CGMGM) that achieves few-shot semantic segmentation by constructing a joint distribution of pixels and categories on support images and query images.
CGS-Mask: Making Time Series Predictions Intuitive for All
Feng Lu (Huazhong University of Science and Technology), Albert Y. Zomaya (University of Sydney)
Explainability and InterpretabilityTime SeriesElectronic Health Records
🎯 What it does: This paper proposes a post-hoc, model-agnostic method based on cellular genetic strip masks (CGS-Mask) for interpreting feature importance in time series prediction models and generating easy-to-understand visual results.
Chain of Generation: Multi-Modal Gesture Synthesis via Cascaded Conditional Control
Zunnan Xu (Tsinghua University), Xiu Li (Tsinghua University)
GenerationData SynthesisPose EstimationConvolutional Neural NetworkContrastive LearningMultimodalityAudio
🎯 What it does: This paper proposes a cascade conditional control framework based on multimodal priors, which generates facial Blendshapes, body movements, and hand gestures using audio, text, and previous poses, achieving high-quality 3D pose synthesis even when some modalities are missing during the inference phase.
Chain-of-Thought Improves Text Generation with Citations in Large Language Models
Bin Ji (National University of Singapore), See-Kiong Ng (National University of Singapore)
GenerationRetrievalTransformerLarge Language ModelTextBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: This study investigates the enhancement of the correctness and citation quality of large language models in generating text with citations through Chain-of-Thought (CoT) prompts and a Citation Insurance Mechanism (CIM); systematic experiments were conducted on six major open-source LLMs.
Cheaper and Faster: Distributed Deep Reinforcement Learning with Serverless Computing
Hanfei Yu (Louisiana State University), Hao Wang (Louisiana State University)
OptimizationReinforcement LearningSequential
🎯 What it does: The MINIONSRL framework is proposed, utilizing Azure Serverless Computing (ACI) to achieve distributed DRL training, and dynamically adjusting the number of actors per round through a reinforcement learning scheduler to reduce training time and costs.
Chinese Spelling Correction as Rephrasing Language Model
Linfeng Liu (Shanghai Jiao Tong University), Hai Zhao (Shanghai Jiao Tong University)
GenerationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper studies Chinese spelling correction and proposes a rewritten language model (ReLM) that changes the model training objective to semantic rewriting of the input sentences, thereby replacing traditional character-level labeling methods.
ChromaFusionNet (CFNet): Natural Fusion of Fine-Grained Color Editing
Yi Dong (Alibaba Nanyang Technological University Singapore Joint Research Institute), Xuansong Xie (Alibaba Group)
Image HarmonizationRestorationTransformerGenerative Adversarial NetworkImage
🎯 What it does: A model specifically designed for fine-grained color editing fusion, CFNet, is proposed, which can automatically fix boundary inconsistencies after multi-region color editing, enhancing the overall visual effect.
Chronic Poisoning: Backdoor Attack against Split Learning
Fangchao Yu (Wuhan University), Lina Wang (Wuhan University)
Federated LearningAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: This study investigates backdoor attacks in Split Learning on the server side and proposes a three-stage framework (Steal, Finetune, Implant) called SFI, which injects backdoors without accessing client data and models.
CI-STHPAN: Pre-trained Attention Network for Stock Selection with Channel-Independent Spatio-Temporal Hypergraph
Hongjie Xia (Fudan University), Hongfeng Chai (Fudan University)
Recommendation SystemGraph Neural NetworkTransformerReinforcement LearningTime SeriesFinance Related
🎯 What it does: A two-stage framework called CI-STHPAN is proposed, which first uses Transformer and HGAT for self-supervised pre-training on stock time series, and then fine-tunes on stock ranking tasks to achieve quantitative stock selection.
CIDR: A Cooperative Integrated Dynamic Refining Method for Minimal Feature Removal Problem
Qian Chen (East China Normal University), Xiaofeng He (East China Normal University)
OptimizationExplainability and InterpretabilityTransformerLarge Language ModelText
🎯 What it does: The CIDR method is proposed to solve the minimum feature removal problem in NLP, utilizing cooperative integrated gradients to detect interactions between word pairs and transforming the problem into a knapsack problem to find the minimum feature set.
CK12: A Rounded K12 Knowledge Graph Based Benchmark for Chinese Holistic Cognition Evaluation
Weihao You (Tomorrow Advancing Life), Jinfeng Bai (Tomorrow Advancing Life)
TransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: A Chinese K12 assessment benchmark CK12 based on knowledge graphs has been constructed, covering 9 subjects, 584 primary knowledge points, and 1989 secondary knowledge points, totaling 39,452 questions. Additionally, 8 mainstream LLMs (including Chinese and English models) were evaluated under four prompting modes, and 7,334 multi-step reasoning questions were provided, with the CoT process assessed using interpretability metrics.
CL2CM: Improving Cross-Lingual Cross-Modal Retrieval via Cross-Lingual Knowledge Transfer
Yabing Wang (Zhejiang Gongshang University), Hao Luo (Alibaba Group)
RetrievalComputational EfficiencyKnowledge DistillationTransformerContrastive LearningImageTextMultimodality
🎯 What it does: A cross-language cross-modal retrieval framework CL2CM based on cross-language knowledge transfer is proposed, achieving alignment between visual and target language under the premise of using only source language annotations.
Clarifying the Behavior and the Difficulty of Adversarial Training
Xu Cheng (Nanjing University of Science and Technology), Quanshi Zhang (Shanghai Jiao Tong University)
OptimizationAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: This paper derives the approximate dynamics of multi-step adversarial attack perturbations based on a two-layer ReLU network under three simplified assumptions, and uses this theory to explain the difficulties in optimizing adversarial training, including phenomena such as gradient amplification, sample imbalance, Hessian reinforcement, and network parameter oscillation.
Class-Attribute Priors: Adapting Optimization to Heterogeneity and Fairness Objective
Xuechen Zhang (University of California), Samet Oymak (University of Michigan)
OptimizationMeta LearningImage
🎯 What it does: This paper proposes the Class-Attribute Priors (CAP) framework, which achieves personalized optimization strategies for class heterogeneity and fairness objectives by mapping class attributes to class-specific hyperparameters.
CLIM: Contrastive Language-Image Mosaic for Region Representation
Size Wu (Nanyang Technological University), Chen Change Loy (Nanyang Technological University)
Object DetectionRepresentation LearningVision Language ModelContrastive LearningImageText
🎯 What it does: By stitching multiple images into a mosaic and treating each image as a pseudo-region, contrastive learning is used to achieve region-text alignment, thereby enhancing open vocabulary object detection and visual-language model region representation without the need for box annotations.
CLIP-Gaze: Towards General Gaze Estimation via Visual-Linguistic Model
Pengwei Yin (Hikvision Research Institute), Di Xie (Hikvision Research Institute)
RecognitionDomain AdaptationTransformerPrompt EngineeringVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: Using the pre-trained vision-language model CLIP, diverse gaze-unrelated features are constructed, and gaze-related features are separated from these unrelated features in the feature space, enabling cross-domain gaze estimation. A Personalized Context Optimization (PCO) is proposed to automatically generate text prompts tailored for each individual and further adjust the feature distribution through feature ranking loss.