AAAI 2025 Papers — Page 28

Text and Image Are Mutually Beneficial: Enhancing Training-Free Few-Shot Classification with CLIP

Yayuan Li (Nanjing University), Yinghuan Shi (Southeast University)

🎯 What it does: This paper proposes the TIMO method, which utilizes the mutual guidance of image and text features from CLIP (Image-Guided-Text and Text-Guided-Image) to enhance training-free few-shot classification performance.

Text Proxy: Decomposing Retrieval from a 1-to-N Relationship into N 1-to-1 Relationships for Text-Video Retrieval

Jian Xiao (Hefei University of Technology), Richang Hong (Hefei University of Technology)

🎯 What it does: This paper proposes TV-ProxyNet, which decomposes the original 1-to-N text-video retrieval relationship into N 1-to-1 relationships by splitting the text query into a series of text proxies, thereby achieving finer semantic alignment.

Text to Point Cloud Localization with Multi-Level Negative Contrastive Learning

Dunqiang Liu (Xiamen University), Cheng Wang (Xiamen University)

🎯 What it does: This paper proposes a multi-layer negative contrastive learning framework that uses language as a filter rather than directly aligning text with point cloud features, significantly improving text-point cloud localization accuracy at the urban scale.

Text-Guided Fine-grained Counterfactual Inference for Short Video Fake News Detection

Linlin Zong (Dalian University of Technology), Shimin Wu (Dalian University of Technology)

🎯 What it does: The TGFC-SVFN model is proposed, which utilizes a text teacher model, fine-grained counterfactual reasoning, and multimodal fusion technology to detect fake news in short videos.

Text-Guided Nonverbal Enhancement Based on Modality-Invariant and -Specific Representations for Video Speaking Style Recognition

Beibei Zhang (Nanjing University), Gangshan Wu (Nanjing University)

🎯 What it does: This paper proposes a text-guided non-verbal enhancement method (TNvE) that selects key non-verbal segments based on textual information and combines modality-specific and modality-invariant representations to identify speaking styles in videos.

Text2Data: Low-Resource Data Generation with Textual Control

Shiyu Wang (Salesforce AI Research), Silvio Savarese (Salesforce AI Research)

🎯 What it does: This paper proposes a framework called Text2Data, which utilizes unlabeled data to pre-train a diffusion model and then fine-tunes it with a small amount of labeled data for text-to-data generation in low-resource scenarios.

Text2midi: Generating Symbolic Music from Captions

Keshav Bhandari (Queen Mary University of London), Dorien Herremans (Singapore University of Technology and Design)

🎯 What it does: An end-to-end text2midi model has been developed that can directly generate high-quality MIDI files based on natural language descriptions.

Text2Relight: Creative Portrait Relighting with Text Guidance

Junuk Cha (Ulsan National Institute of Science and Technology), Seungryul Baek (Ulsan National Institute of Science and Technology)

🎯 What it does: A text-driven single-image portrait relighting model named Text2Relight has been developed, capable of changing the lighting of the foreground and background while keeping the subject content unchanged based on natural language descriptions.

TextRefiner: Internal Visual Feature as Efficient Refiner for Vision-Language Models Prompt Tuning

Jingjing Xie (Xiamen University), Liujuan Cao (Xiamen University)

🎯 What it does: This paper proposes a plugin module called TextRefiner, which refines existing text prompts using the internal local token information of visual models, thereby enhancing the transfer performance of vision-language models in downstream tasks.

TextToucher: Fine-Grained Text-to-Touch Generation

Jiahang Tu (Zhejiang University), Hui Qian (Zhejiang University)

🎯 What it does: This paper proposes the TextToucher method, which generates tactile images from text conditions using fine-grained text.

Textualize Visual Prompt for Image Editing via Diffusion Bridge

Pengcheng Xu (Western University), Boyu Wang (Western University)

🎯 What it does: A diffusion bridge is constructed using a single text-to-image diffusion model, transforming visual prompts (before-and-after image pairs) into text embeddings suitable for image editing, avoiding the need for text-image-image triplets and large-scale retraining as in TI2I methods;

Textured Mesh Saliency: Bridging Geometry and Texture for Human Perception in 3D Graphics

Kaiwei Zhang (Shanghai Jiao Tong University), Guangtao Zhai (Shanghai Jiao Tong University)

🎯 What it does: A saliency prediction method for texture meshes is proposed, and the first texture mesh saliency dataset based on a 6DOF VR eye-tracking experiment is constructed.

TG-LLaVA: Text Guided LLaVA via Learnable Latent Embeddings

Dawei Yan (Northwestern Polytechnical University), Chunhua Shen (Zhejiang University)

🎯 What it does: This paper proposes TG-LLaVA, which optimizes the visual encoder through text-guided learnable latent embeddings to enhance the overall performance of visual language models.

TGBFormer: Transformer-GraphFormer Blender Network for Video Object Detection

Qiang Qi (Qingdao University of Science and Technology), Xiao Wang (Qingdao University of Science and Technology)

🎯 What it does: This paper proposes a TGBFormer network that combines Transformer and GraphFormer for video object detection, capable of simultaneously capturing global and local features and fusing them through a feature blender.

TGFormer: Transformer with Track Query Group for Multi-Object Tracking

Rui Zeng (Beijing University of Posts and Telecommunications), Songwei Pei (Beijing University of Posts and Telecommunications)

🎯 What it does: Designed and implemented TGFormer, a Transformer-based multi-object tracker that uses trajectory query groups to handle occlusion variations.

TGLsta: Low-resource Textual Graph Learning with Semantic and Topological Awareness via LLMs

Qin Zhang (Shenzhen University), Shirui Pan (Griffith University)

🎯 What it does: A low-resource text graph node classification method TGL-sta is proposed, utilizing large language models and multi-view contrastive learning, combining text, text summaries, topology, and graph structure information, and then achieving zero/few-shot inference through prompt tuning.

The (Exact) Price of Cardinality for Indivisible Goods: A Parametric Perspective

Alexander Lam (City University of Hong Kong), Ankang Sun (Hong Kong Polytechnic University)

🎯 What it does: This paper proposes and analyzes the fair allocation problem under given count constraints on assignable items (i.e., each agent can receive at most k items), defining and calculating the 'price' (worst-case social welfare loss) for utilitarian and egalitarian total utility.

The Adaptive Q-Network for Recommendation Tasks with Dynamic Item Space

Jianxiang Zhu (Shanghai University), Yaxin Peng (Shanghai University)

🎯 What it does: This paper proposes a dynamic recommendation task and designs the Adaptive Q-Network (AdaQN) algorithm to recommend items in a constantly changing item space.

The Bandit Whisperer: Communication Learning for Restless Bandits

Yunfan Zhao (Harvard University), Milind Tambe (Google Deepmind)

🎯 What it does: A method is proposed to improve decision-making in a restless multi-armed bandit (RMAB) with systemic noise using communication learning.

The Complexity of Extending Fair Allocations of Indivisible Goods

Argyrios Deligkas (Royal Holloway University of London), Stavros D. Ioannidis (Royal Holloway University of London)

🎯 What it does: The research explores how to extend to envy-free (EF) allocations under the condition of fixed partial allocations, proposing extension problems and conducting a parameterized complexity analysis.

The Cost Perspective of Liquid Democracy: Feasibility and Control

Shiri Alouf-Heffetz (Ben Gurion University), Georgios Papasotiropoulos (University of Warsaw)

🎯 What it does: Analyzes the feasible voting distribution and control issues of droplet democracy under budget constraints, and elaborates on its multi-dimensional optimization framework for minimizing costs, path lengths, and the concentration of voting power from the perspective of computational complexity.

The Distortion of Public-Spirited Participatory Budgeting

Mark Bedaywi (University of Toronto), Nisarg Shah (University of Toronto)

🎯 What it does: This paper studies the distortion problem of participatory budgeting (PB) systems under the assumption of public-spirited voting. It proves that the distortion of any deterministic rule under common ranking voting methods is at least Θ(m), and proposes a new 'rank-b-bundle' voting format, designing three multi-round protocols that achieve distortion upper bounds of O(√m), O(log m), and O(1), respectively; it also provides corresponding random rules and theoretical limits.

The Distributional Reward Critic Framework for Reinforcement Learning Under Perturbed Rewards

Xi Chen (Ohio State University), Andrew Perrault (Ohio State University)

🎯 What it does: This paper proposes the Distributional Reward Critic (DRC) framework and its general version (General Distributional Reward Critic, GDRC) for training reinforcement learning (RL) agents in environments where rewards are subject to unknown disturbances (Generalized Confusion Matrix, GCM) or continuous noise. By transforming the reward estimation into a classification task and using Ordinal Cross-Entropy (OCE), this method can estimate the reward distribution online and replace actual rewards with estimated rewards, enhancing learning robustness.

The Dynamic Duo of Collaborative Masking and Target for Advanced Masked Autoencoder Learning

Shentong Mo (Carnegie Mellon University)

🎯 What it does: This paper proposes a collaborative masking and self-supervised visual pre-training framework CMT-MAE that combines teacher and student attention.

The Gradient of Algebraic Model Counting

Jaron Maene (KU Leuven), Luc De Raedt (KU Leuven)

🎯 What it does: A general gradient definition ∇AMC based on Algebraic Model Counting (AMC) is proposed, and an optimized backpropagation algorithm that can be executed on various semirings is designed based on this.

The Illusion of Empathy: How AI Chatbots Shape Conversation Perception

Tingting Liu (National Institute on Drug Abuse), João Sedoc (New York University)

🎯 What it does: This study compares the differences in dialogue quality and user emotional perception between humans and the GPT-4o chatbot, exploring the impact of emotional perception on dialogue quality.

The Impact of Literal Sorting on Cardinality Constraint Encodings

Joseph E. Reeves (Carnegie Mellon University), Marijn J. H. Heule (University of Amsterdam)

🎯 What it does: This paper proposes various automated methods for sorting literals in cardinality constraint encoding to change the semantics of auxiliary variables and enhance the learning effectiveness of CDCL solvers.

The Indoor-Training Effect: Unexpected Gains from Distribution Shifts in the Transition Function

Serena Bono (Massachusetts Institute of Technology), Gabriel Kreiman (Harvard University)

🎯 What it does: This study investigates the phenomenon of training environments that are not completely identical to testing environments in reinforcement learning, known as the 'indoor training effect.' By adding controllable Gaussian noise to the transition function of the original MDP to generate various δ-environments, the performance differences between training in a noise-free environment and testing in a noisy environment (Generalization Agent) versus training and testing in the same noisy environment (Learnability Agent) are compared.

The Master Key Filters Hypothesis: Deep Filters Are General

Zahra Babaiee (Technische Universitat Wien), Radu Grosu (Massachusetts Institute of Technology)

🎯 What it does: This study investigates the generality of deep convolutional kernels in Deep Separable Convolutional Networks (DS-CNN) and proposes the Master Key Filters Hypothesis, suggesting that deep convolutional kernels maintain general features that can be transferred across domains and architectures.

The Parables of the Mustard Seed and the Yeast: Extremely Low-Budget, High-Performance Nighttime Semantic Segmentation

Shiqin Wang (Wuhan University of Science and Technology), Zheng Wang (Wuhan University)

🎯 What it does: For the nighttime semantic segmentation task with an extremely low annotation budget, a context-aware Region Active Query (CARQ) and Fragmented Collaborative Active Domain Adaptation (FS-ADA) framework is proposed, which can achieve performance close to fully supervised models with only 1% labeled data.

The Surprising Effectiveness of Infinite-Width NTKs for Characterizing and Improving Model Training

Joshua DeOliveira (Worcester Polytechnic Institute), Elke Rundensteiner (Massachusetts Institute of Technology)

🎯 What it does: By calculating the NTK Gram matrix of infinitely wide neural networks, predictions and improvements are made for various tasks such as architecture selection, pseudo-label validation, bias identification, and label correction without training any models.

The Unreasonable Effectiveness of Open Science in AI: A Replication Study

Odd Erik Gundersen (Norwegian University of Science and Technology), Nicklas Grimstad Nilsen (Aneo AS)

🎯 What it does: This paper conducts a systematic replication experiment on 30 highly cited artificial intelligence papers, documenting the issues encountered during the replication process and assessing reproducibility.

The Value of Recall in Extensive-Form Games

Ratip Emin Berker (Carnegie Mellon University), Vincent Conitzer (Carnegie Mellon University)

🎯 What it does: The concept of 'Value of Recall' (VoR) is proposed, quantifying the enhancement of players' earnings in extensive-form games due to perfect recall, and systematically studying the computational complexity, upper bounds, pathologies, and optimal allocation in cases of partial recall.

The VOROS: Lifting ROC Curves to 3D to Summarize Unbalanced Classifier Performance

Christopher Ratigan (Tufts University), Lenore Cowen (Tufts University)

🎯 What it does: A new evaluation metric called VOROS (Volume over the ROC Surface) is proposed, which elevates the traditional ROC curve to three-dimensional space to simultaneously consider the performance of classifiers under different costs and class imbalances.

Thermal-Aware Low-Light Image Enhancement: A Real-World Benchmark and a New Light-Weight Model

Zhen Wang (Institute of Science Tokyo), Zhishuai Yin (Wuhan University of Technology)

🎯 What it does: The first low-light image enhancement dataset TLIE combined with thermal imaging is proposed, and a lightweight multimodal network ThermNet is designed.

THESAURUS: Contrastive Graph Clustering by Swapping Fused Gromov-Wasserstein Couplings

Bowen Deng (Sun Yat-sen University), Tao Zhang (Sun Yat-sen University)

🎯 What it does: A graph node clustering framework THESAURUS is proposed, utilizing semantic prototypes, cross-view alignment, and fusion of Gromov-Wasserstein Optimal Transport (OT);

THGNets: Constrained Temporal Hypergraphs and Graph Neural Networks in Hyperbolic Space for Information Diffusion Prediction

Yanchao Liu (Communication University of China), Junpeng Gong (Communication University of China)

🎯 What it does: Designed the THGNets framework, which combines hyperbolic temporal hypergraph neural networks, hyperbolic graph neural networks, and hyperbolic GRU to predict the next infected user in information diffusion.

Thin-Plate Spline-based Interpolation for Animation Line Inbetweening

Tianyi Zhu (Harbin Institute of Technology), Dongwei Ren (Tianjin University)

🎯 What it does: Proposes an animation line drawing intermediate frame interpolation method based on Thin Plate Spline (TPS) and a motion refinement module;

Thinking in Granularity: Dynamic Quantization for Image Super-Resolution by Intriguing Multi-Granularity Clues

Mingshen Wang (Hefei University of Technology), Meng Wang (Anhui University)

🎯 What it does: The Granular-DQ framework is proposed to achieve layer-invariant dynamic quantization based on multi-granularity features and entropy statistics for single image super-resolution networks.

Thinking Racial Bias in Fair Forgery Detection: Models, Datasets and Evaluations

Decheng Liu (Xidian University), Xinbo Gao (Chongqing University of Posts and Telecommunications)

🎯 What it does: This paper proposes a comprehensive benchmark for evaluating the fairness of face forgery detection, including a new face forgery dataset, two categories of improved fairness metrics, and a weight pruning method that requires no training.

Thought-Path Contrastive Learning via Premise-Oriented Data Augmentation for Logical Reading Comprehension

Chenxu Wang (Beijing Institute of Technology), Zhen Yang (Beijing Institute of Technology)

🎯 What it does: A premise-based Contrastive Generation and Thought Comparison Learning (PODA-TPCL) framework is proposed to enhance logical reading comprehension abilities.

Threshold UCT: Cost-Constrained Monte Carlo Tree Search with Pareto Curves

Martin Kurečka (Masaryk University), Vít Unčovský (Masaryk University)

🎯 What it does: The Threshold UCT algorithm is proposed for safe planning in constrained Markov decision processes.

Through the Dual-Prism: A Spectral Perspective on Graph Data Augmentation for Graph Classifications

Yutong Xia (National University of Singapore), Roger Zimmermann (National University of Singapore)

🎯 What it does: This paper proposes a graph data augmentation method based on spectral theory called Dual-Prism (DP), which achieves global deformation and attribute preservation of graph structures by modifying only high-frequency features while keeping low-frequency spectra unchanged.

Till the Layers Collapse: Compressing a Deep Neural Network Through the Lenses of Batch Normalization Layers.

Zhu Liao (Telecom Paris), Enzo Tartaglione (Telecom Paris)

🎯 What it does: A deep neural network layer pruning method called TLC based on batch normalization parameters is proposed, which removes unimportant layers by evaluating the ON/OFF status of each layer, thereby reducing the network depth.

Tilted Quantile Gradient Updates for Quantile-Constrained Reinforcement Learning

Chenglin Li (Tsinghua University), Hua Geng (Tsinghua University)

🎯 What it does: This paper proposes a quantile-constrained reinforcement learning method TQPO based on tilted quantile gradient updates, which directly samples to estimate the quantile gradient and addresses the overly conservative problem through tilted updates.

Time Series Supplier Allocation via Deep Black-Litterman Model

Xinke Jiang (University of Electronic Science and Technology of China), Jiayuan Luo (Zhongnan University of Economics and Law)

🎯 What it does: This paper proposes a Deep Black-Litterman Model (DBLM) for Time Series Supplier Allocation (TSSA), which achieves automatic learning and allocation decisions for the future supplier perspective matrix by combining the traditional Black-Litterman portfolio model with spatiotemporal graph neural networks.

TIME-FS: Joint Learning of Tensorial Incomplete Multi-View Unsupervised Feature Selection and Missing-View Imputation

Yanyong Huang (Southwestern University of Finance and Economics), Tianrui Li (Southwest Jiaotong University)

🎯 What it does: A joint learning framework is proposed that can simultaneously complete missing views, perform unsupervised feature selection, and learn low-dimensional representations on incomplete multi-view data.

TimeCAP: Learning to Contextualize, Augment, and Predict Time Series Events with Large Language Model Agents

Geon Lee (Korea Advanced Institute of Science and Technology), Haifeng Chen (NEC Labs)

🎯 What it does: A framework named TIMECAP is proposed, utilizing two LLM agents (one for textual time series context and one for context-based predictions), and combining a multimodal encoder for dual enhancement of input and prompts, ultimately achieving event prediction.

TimeCHEAT: A Channel Harmony Strategy for Irregularly Sampled Multivariate Time Series Analysis

Jiexi Liu (Nanjing University of Aeronautics and Astronautics), Songcan Chen (Nanjing University of Aeronautics and Astronautics)

🎯 What it does: A new temporal analysis framework called TimeCHEAT is proposed, specifically designed for irregularly sampled multivariate time series (ISMTS), which can simultaneously consider local channel dependence (CD) and global channel independence (CI) strategies.

TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment

Chenxi Liu (Nanyang Technological University), Rui Zhao (SenseTime Research)

🎯 What it does: This paper proposes the TimeCMA framework, which utilizes cross-modal alignment combined with LLM and time series encoders to achieve multivariate time series forecasting.

TimeDP: Learning to Generate Multi-Domain Time Series with Domain Prompts

Yu-Hao Huang (Nanjing University), Jiang Bian (Microsoft Research Asia)

🎯 What it does: A multi-domain time series diffusion model, TimeDP, is proposed, which learns domain prompts using time series prototypes and a prototype allocation module, and generates time series for different domains with a small number of samples.

TimePFN: Effective Multivariate Time Series Forecasting with Synthetic Data

Ege Onur Taga (University of Michigan), Samet Oymak (University of Michigan)

🎯 What it does: A multivariate time series zero/few-shot prediction framework called TimePFN based on synthetic data and Transformer is proposed, along with a unified model training and fine-tuning process.

Times2D: Multi-Period Decomposition and Derivative Mapping for General Time Series Forecasting

Reza Nematirad (Kansas State University), Balasubramaniam Natarajan (Kansas State University)

🎯 What it does: This paper proposes a time series forecasting framework named Times2D, which transforms 1D sequences into 2D representations through frequency domain periodic decomposition and derivative heatmap mapping, and then utilizes 2D convolution and self-attention networks for prediction.

TinyFoA: Memory Efficient Forward-Only Algorithm for On-Device Learning

Baichuan Huang (Lund University), Amir Aminifar (Lund University)

🎯 What it does: A forward training algorithm for device learning, TinyFoA, is proposed for efficiently training neural networks on mobile/IoT devices.

TinySAM: Pushing the Envelope for Efficient Segment Anything Model

Han Shu (University of Science and Technology of China), Xinghao Chen (Huawei Noah's Ark Lab)

🎯 What it does: A lightweight TinySAM model was constructed using techniques such as full-stage knowledge distillation, hard mask weighting and hard prompt sampling, post-training quantization, and hierarchical everything reasoning, achieving a significant reduction in computational load while maintaining the zero-shot segmentation performance of SAM.

TinySubNets: An Efficient and Low Capacity Continual Learning Strategy

Marcin Pietron, Roberto Corizzo (American University)

🎯 What it does: This paper proposes TinySubNets (TSN), a no-forgetting continual learning framework based on pruning, variable sparsity, adaptive quantization, and weight sharing.

TIV-Diffusion: Towards Object-Centric Movement for Text-driven Image to Video Generation

Xingrui Wang (University of Science and Technology of China), Zhibo Chen (University of Science and Technology of China)

🎯 What it does: A text-image-to-video generation framework TIV-Diffusion based on diffusion models is proposed, utilizing object separation for motion control.

TNCSE: Tensor Norm Constraints for Unsupervised Contrastive Learning of Sentence Embeddings

Tianyu Zong (University of Chinese Academy of Sciences), Jungang Xu (University of Chinese Academy of Sciences)

🎯 What it does: An unsupervised contrastive learning framework TNCSE based on tensor norm constraints is proposed, and sentence embeddings are achieved through a dual-encoder integration.

To Predict or Not to Predict? Proportionally Masked Autoencoders for Tabular Data Imputation

Jungkyu Kim (Yonsei University), Taeyoung Park (Yonsei University)

🎯 What it does: Proposes the Proportional Masked Autoencoder (PMAE) to improve missing value imputation for tabular data.

Tokenization, Fusion, and Augmentation: Towards Fine-grained Multi-modal Entity Representation

Yichi Zhang (Zhejiang University), Huajun Chen (Zhejiang University)

🎯 What it does: The MYGO framework is proposed, which completes multimodal knowledge graphs through modal tokenization, hierarchical Triple Modeling, and fine-grained contrastive learning.

TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification

Xiao Wang (Wuhan University of Science and Technology), Xin Xu (Wuhan University)

🎯 What it does: A TokenMatcher framework is designed to achieve unsupervised visible-infrared person re-identification through multi-class token matching, neighborhood learning, and homogeneous fusion.

Tokenphormer: Structure-aware Multi-token Graph Transformer for Node Classification

Zijie Zhou (University of Macau), Leong Hou U (University of Macau)

🎯 What it does: A graph transformer named Tokenphormer is proposed, which learns node representations through various fine-grained and global token generation mechanisms (walk-token, SGPM-token, hop-token) to achieve efficient node classification.

Told You That Will Not Work: Optimal Corrections to Planning Domains Using Counter-Example Plans

Songtuan Lin (Australian National University), Pascal Bercher (Paris-Saclay University)

🎯 What it does: This paper proposes an optimal repair method for planning domains based on positive and negative plans, which can output the smallest set of atomic repairs under the premise of given positive and negative example plans, making all positive examples feasible and all negative examples infeasible.

ToMATO: Verbalizing the Mental States of Role-Playing LLMs for Benchmarking Theory of Mind

Kazutoshi Shinoda (NTT Corporation), Kuniko Saito (NTT Corporation)

🎯 What it does: The ToMATO benchmark dataset was constructed to generate psychological state question-answering through multi-turn LLM-LLM dialogues, assessing the Theory of Mind (ToM) capabilities of LLMs.

Topo2Seq: Enhanced Topology Reasoning via Topology Sequence Learning

Yiming Yang (Chinese University of Hong Kong Shenzhen), Zhen Li (Tencent)

🎯 What it does: This paper proposes a topology reasoning framework called Topo2Seq, enhanced by learning topological sequences, for real-time perception of lane centerlines and their topological relationships from perspective images captured by multi-view cameras.

Topology-Aware 3D Gaussian Splatting: Leveraging Persistent Homology for Optimized Structural Integrity

Tianqi Shen (City University of Hong Kong), Ning An (China Coal Research Institute)

🎯 What it does: Based on 3D Gaussian Splatting, we propose Topology-Aware 3D Gaussian Splatting (Topology-GS), which enhances view synthesis quality through two new techniques: 1) Local Voronoi Interpolation (LPVI) based on persistent homology to fill in the sparsity of point clouds in low-curvature areas; 2) Persistent Homology Loss (PersLoss) constrains the topological features of images during training, ensuring the integrity of feature layer structures.

Toward Adaptive Large Language Models Structured Pruning via Hybrid-grained Weight Importance Assessment

Jun Liu (Northeastern University), Yanzhi Wang (Northeastern University)

🎯 What it does: A hybrid granularity weight importance assessment (HyWIA) framework is proposed, which utilizes attention to fuse fine-grained and coarse-grained importance scores to achieve structured pruning of LLMs, followed by LoRA fine-tuning to restore performance.

Toward Efficient Data-Free Unlearning

Chenhao Zhang (University of Queensland), Miao Xu (University of Queensland)

🎯 What it does: This paper proposes a data-free unlearning method based on data-free knowledge distillation (ISPF), which addresses the issue of low distillation efficiency caused by the generator producing too many forgotten class samples in traditional methods.

Toward Falsifying Causal Graphs Using a Permutation-Based Test

Elias Eulig (German Cancer Research Center), Dominik Janzing (Amazon Research)

🎯 What it does: This paper proposes a baseline based on node permutation to evaluate the consistency of a given directed acyclic graph in observational data.

Toward Improving Robustness and Accuracy in Unsupervised Domain Adaptation

Aishwarya Soni (Indian Institute of Technology Benares Hindu University), Tanima Dutta (Indian Institute of Technology Benares Hindu University)

🎯 What it does: A two-step SGD training scheme called Consistent Attention Mapping and Self Pseudo Label Refinement (CAM+SPLR) is proposed to enhance the robustness and accuracy of unsupervised domain transfer models against adversarial examples.

Toward Modality Gap: Vision Prototype Learning for Weakly-supervised Semantic Segmentation with CLIP

Zhongxing Xu (Monash University), Zongyuan Ge

🎯 What it does: This paper proposes the Vision Prototype Learning framework, which learns category prototypes in the visual space to bridge the gap between the text-visual modalities of CLIP and improve the localization quality of weakly supervised semantic segmentation.

Toward Verifiable Instruction-Following Alignment for Retrieval Augmented Generation

Guanting Dong (Renmin University of China), Ji-Rong Wen (Renmin University of China)

🎯 What it does: Generate high-quality instruction-following data through an automated and verifiable synthesis process, enhancing the instruction adherence capability of retrieval-augmented generation systems.

Towards a Comprehensive, Efficient and Promptable Anatomic Structure Segmentation Model Using 3D Whole-Body CT Scans

Heng Guo (DAMO Academy Alibaba Group), Minfeng Xu (DAMO Academy Alibaba Group)

🎯 What it does: CT-SAM3D is proposed, a three-dimensional interactive segmentation model based on whole-body CT, capable of achieving high-precision segmentation of nearly a hundred anatomical structures.

Towards a Multimodal Large Language Model with Pixel-Level Insight for Biomedicine

Xiaoshuang Huang (Baidu Inc), Yehui Yang (Peking University)

Towards Accurate Binary Spiking Neural Networks: Learning with Adaptive Gradient Modulation Mechanism

Yu Liang (University of Electronic Science and Technology of China), Yang Yang (University of Electronic Science and Technology of China)

🎯 What it does: This paper addresses the issue of frequent weight sign flips during the learning process of Binary Spiking Neural Networks (BSNN) by proposing an Adaptive Gradient Modulation Mechanism (AGMM) to dynamically reduce gradient magnitude, thereby decreasing flip frequency and enhancing network performance.

Towards Adversarially Robust Dataset Distillation by Curvature Regularization

Eric Xue (University of Toronto), Haohan Wang (University of Illinois Urbana-Champaign)

🎯 What it does: The GUARD method is proposed, which incorporates curvature regularization during the dataset distillation process to enhance the adversarial robustness of the distilled model.

Towards Audio-Visual Navigation in Noisy Environments: A Large-Scale Benchmark Dataset and an Architecture Considering Multiple Sound-Sources

Zhanbo Shi (Tongji University), Ying Shen (Tongji University)

🎯 What it does: A large-scale audio-visual navigation benchmark dataset BeDAViN and a navigation framework ENMuS3 for multi-source noise environments are proposed, aiming to address the challenges of audio-visual navigation in the presence of multiple sound sources.

Towards Better Robustness Against Natural Corruptions in Document Tampering Localization

Huiru Shao (Xi'an Jiaotong-Liverpool University), Qiufeng Wang (Duke Kunshan University)

🎯 What it does: Proposes adversarial forensic regularization (AFR) based on mutual information and distribution discrepancy regularization (DDR) to enhance the robustness of document tampering localization under natural degradation (noise, JPEG, social network transmission) and adversarial attacks; simultaneously constructs the TSroie-CRP dataset.

Towards Better Spherical Sliced-Wasserstein Distance Learning with Data-Adaptive Discriminative Projection Direction

Hongliang Zhang (Nanjing University of Science and Technology), Jian Yang (Nanjing University of Science and Technology)

🎯 What it does: A data-adaptive spherical sliced Wasserstein distance (DSSW) is proposed, which better measures the differences in spherical distributions by assigning weights to different projection directions.

Towards Effective, Efficient and Unsupervised Social Event Detection in the Hyperbolic Space

Xiaoyan Yu (Beijing Institute of Technology), Philip S. Yu (University of Illinois at Chicago)

🎯 What it does: An unsupervised social event detection framework called HyperSED is designed, which first aggregates massive social messages into semantic anchor nodes to construct a Semantic Anchor Graph (SAMG). It then learns structure and geometry-aware anchor representations in hyperbolic space and utilizes differentiable structural entropy to build a hierarchical tree for event detection.

Towards Efficient Collaboration via Graph Modeling in Reinforcement Learning

Wenzhe Fan (University of Illinois Chicago), Xinhua Zhang (Peking University)

🎯 What it does: A factorized multi-agent transformer (f-MAT) based on a graph model is proposed, achieving efficient collaborative training and execution under the CTDE framework.

Towards Efficient Low-Order Hybrid Optimizer for Language Model Fine-Tuning

Minping Chen (Hong Kong University of Science and Technology), Zeyi Wen (Hong Kong University of Science and Technology)

🎯 What it does: A low-order hybrid optimizer, LoHO, is proposed, which integrates zero-order (MeZO) and first-order (SGD/Adam) optimizers in language model fine-tuning, achieving memory efficiency while improving accuracy and convergence speed.

Towards Efficient Object Re-Identification with a Novel Cloud-Edge Collaborative Framework

Chuanming Wang (Beijing University of Posts and Telecommunications), Huadong Ma (Beijing University of Posts and Telecommunications)

🎯 What it does: A cloud-edge collaborative ReID framework based on Distribution-Aware Correlation Modeling Network (DaCM) is proposed, achieving efficient inference and dynamic bandwidth allocation.

Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering

Hao Lu (Hong Kong University of Science and Technology), Ying-Cong Chen

🎯 What it does: This paper proposes a general multi-camera 3D object detection framework PR-BEV based on perspective rendering, which reconstructs semantic maps from different viewpoints in BEV using implicit foreground volumes and corrects object positions through 3D boxes or pre-trained 2D detectors, thereby enhancing cross-domain generalization performance.

Towards Global-Topology Relation Graph for Inductive Knowledge Graph Completion

Ling Ding (Tianjin University), Dongxiao He (Tianjin University)

🎯 What it does: The TARGI model is proposed, which constructs a relational graph for each type of topological relationship from a global perspective, and uses this graph to learn embeddings for new relationships and entities, thereby achieving complete inductive knowledge graph completion.

Towards Learnable Anchor for Deep Multi-View Clustering

Bocheng Wang (Northwestern Polytechnical University), Xuelong Li (China Telecom)

🎯 What it does: This paper proposes a learnable anchor point deep multi-view clustering model DMAC, which dynamically updates anchor points using positive incentive noise and achieves cross-view anchor point clustering consistency through anchor graph convolution and mutual information maximization, thereby obtaining better clustering results while maintaining linear time complexity.

Towards Loss-Resilient Image Coding for Unstable Satellite Networks

Hongwei Sha (Nanjing University), Zhan Ma (Nanjing University)

🎯 What it does: A loss-robust learning-based image compression method for unstable satellite networks is proposed, achieving an evolvable multi-level bitstream;

Towards Macro-AUC Oriented Imbalanced Multi-Label Continual Learning

Yan Zhang (Shandong University), Yilong Yin (Shandong University)

🎯 What it does: This paper proposes a macro AUC optimization method for multi-label continual learning (MLCL), mainly achieved by introducing a reweighted label distribution-aware margin loss (RLDAM) and a weight retention update (WRU) mechanism;

Towards More Discriminative Feature Learning in SNNs with Temporal-Self-Erasing Supervision

Wei Liu (Chinese Academy of Sciences), Weiming Hu (Chinese Academy of Sciences)

🎯 What it does: This paper studies a Temporal-Self-Erasing (TSE) supervision method to enhance the feature discrimination ability of Spiking Neural Networks (SNN) at different time steps.

Towards Multimodal Sentiment Analysis via Hierarchical Correlation Modeling with Semantic Distribution Constraints

Qinfu Xu (China University of Petroleum), Hengyang Zhou (China University of Petroleum)

🎯 What it does: This paper studies a hierarchical correlation modeling network called HCMNet for multimodal sentiment analysis.

Towards Open-Vocabulary Remote Sensing Image Semantic Segmentation

Chengyang Ye (Dalian University of Technology), Pingping Zhang (Dalian University of Technology)

🎯 What it does: This paper proposes the Open Vocabulary Remote Sensing Image Semantic Segmentation task (OVRSISS), constructs the LandDiscover50K dataset with 51,846 images and 40 categories, and designs the GSNet model to achieve pixel-level segmentation for arbitrary semantic categories.

Towards Practical Classical Planning Compilations of Numeric Planning

Luigi Bonassi (University of Brescia), Enrico Scala (University of Brescia)

🎯 What it does: Three methods are proposed to compile finite field numerical planning into classical planning, and their feasibility is verified.

Towards Precise Prediction Uncertainty in GNNs: Refining GNNs with Topology-grouping Strategy

Hyunjin Seo (Korea Advanced Institute of Science and Technology), Eunho Yang (Korea Advanced Institute of Science and Technology)

🎯 What it does: A post-calibration method based on neighborhood similarity and confidence grouping, SIMI-MAILBOX, is proposed to enhance the predictive uncertainty estimation of graph neural networks.

Towards Projected and Incremental Pseudo-Boolean Model Counting

Suwei Yang (Grabtaxi Holdings), Kuldeep S. Meel (Georgia Institute of Technology)

🎯 What it does: PBCount2 is proposed, an exact pseudo-Boolean model counter that supports projection counting and incremental counting.

Towards Real-Time Approximate Counting

Yash Pote (National University of Singapore), Jiong Yang (Georgia Institute of Technology)

🎯 What it does: A new approximate model counting algorithm, ApproxMC7, is proposed, aimed at quickly estimating the number of satisfying solutions for Boolean formulas under real-time or low-precision requirements.

Towards Realistic Semi-supervised Medical Image Classification

Wenxue Li (Monash University), Zongyuan Ge (Monash University)

🎯 What it does: This paper proposes a self-calibrating semantic training framework to address the issues of open set and class imbalance in semi-supervised learning for medical images.

Towards Robust Knowledge Unlearning: An Adversarial Framework for Assessing and Improving Unlearning Robustness in Large Language Models

Hongbang Yuan (Chinese Academy of Sciences), Jun Zhao (Chinese Academy of Sciences)

🎯 What it does: This paper proposes a dynamic automated attack framework called DUA to evaluate the robustness of large language models during the unlearning process. It also introduces the Latent Adversarial Unlearning (LAU) framework based on latent space adversarial training, along with its two variants, AdvGA and AdvNPO, aimed at significantly enhancing the model's robustness in unlearning tasks.

Towards Robust Visual Question Answering via Prompt-Driven Geometric Harmonization

Yishu Liu (China Academy of Electronics and Information Technology), Bingzhi Chen (Beijing Institute of Technology)

🎯 What it does: The paper proposes a Prompt-Driven Geometric Harmonization (PDGH) framework, which enhances VQA robustness by generating question-guided visual prompts and embedding visual, questions, and prompts onto the unit sphere, integrating orthogonal entropy constraints, geometric contrast constraints, and semantic redundancy correction, particularly in scenarios of class imbalance and language bias.

Towards Runtime Analysis of Population-Based Co-evolutionary Algorithms on Sparse Binary Zero-Sum Game

Per Kristian Lehre (University of Birmingham), Shishen Lin (University of Birmingham)

🎯 What it does: This paper conducts a runtime analysis of the binary zero-sum game DIAGONAL, proving that single-pair CoEAs (such as RLS-PD and (1+1)-CoEA) cannot find the optimal solution in polynomial time, while population-based CoEAs (PDCoEA) can solve it with high probability in polynomial time under low mutation rates and sufficiently large population sizes.

Towards S²-Challenges Underlying LLM-Based Augmentation for Personalized News Recommendation

Shicheng Wang (Institute of Information Engineering), Lihong Wang

🎯 What it does: This paper proposes a news recommendation method based on large language models, S LENR 2, which expands the dataset by generating synthetic news and introduces a structure-aware refinement module and a semantic-aware denoising module to address the issues of structural loss and semantic noise caused by traditional LLM-generated data.