NeurIPS 2024 Papers — Page 11
Conference on Neural Information Processing Systems · 4035 papers
DMNet: Self-comparison Driven Model for Subject-independent Seizure Detection
Shihao Tu (Zhejiang University), Yang Yang (Zhejiang University)
Domain AdaptationAnomaly DetectionConvolutional Neural NetworkTime SeriesBiomedical Data
🎯 What it does: A differential matrix network DMNet based on self-comparison is proposed for inter-subject epilepsy seizure detection.
DMPlug: A Plug-in Method for Solving Inverse Problems with Diffusion Models
Hengkang Wang (University of Minnesota), Ju Sun (University of Minnesota)
RestorationSuper ResolutionDiffusion modelImage
🎯 What it does: This paper proposes a new plugin method DMPlug, which uses a pre-trained diffusion model to solve inverse problems, addressing issues such as insufficient manifold feasibility and measurement feasibility in traditional methods, as well as lack of robustness to unknown noise.
DN-4DGS: Denoised Deformable Network with Temporal-Spatial Aggregation for Dynamic Scene Rendering
Jiahao Lu (University of Science and Technology of China), Tianzhu Zhang (University of Science and Technology of China)
RestorationGenerationData SynthesisGaussian SplattingVideo
🎯 What it does: This paper proposes a dynamic scene rendering method DN-4DGS based on a denoising deformation network.
Do causal predictors generalize better to new domains?
Vivian Yvonne Nastl, Moritz Hardt (Max Planck Institute for Intelligent Systems)
ClassificationDomain AdaptationTabularBiomedical DataBenchmark
🎯 What it does: Evaluate the performance of models trained with causal features versus all features on 16 tabular datasets in cross-domain generalization.
Do Finetti: On Causal Effects for Exchangeable Data
Siyuan Guo (Max Planck Institute for Intelligent Systems), Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
Tabular
🎯 What it does: In a non-i.i.d. exchangeable data environment, the study investigates the problem of causal effect estimation, proposing a causal interpretation of exchangeable generative processes, a truncated factorization formula, conditional intervention distributions, and a simultaneous graph structure and effect estimation algorithm based on multi-environment data called Do-Finetti.
Do LLMs Build World Representations? Probing Through the Lens of State Abstraction
Zichao Li (Mila), Jackie CK Cheung
Representation LearningTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: This paper studies how large language models construct world representations in text planning tasks and proposes a detection framework based on reinforcement learning state abstraction.
Do LLMs dream of elephants (when told not to)? Latent concept association and associative memory in transformers
Yibo Jiang (University of Chicago), Bryon Aragam (University of Chicago)
RetrievalTransformerLarge Language ModelText
🎯 What it does: This paper studies the vulnerability of large language models to context hijacking in fact retrieval tasks, interpreting it as a potential associative memory problem related to conceptual associations, followed by theoretical and experimental analysis on a single-layer Transformer.
Do's and Don'ts: Learning Desirable Skills with Instruction Videos
Hyunseung Kim (Krafton), Jaegul Choo (Korea Advanced Institute of Science and Technology)
Robotic IntelligenceReinforcement LearningVideo
🎯 What it does: This paper proposes the DoDont algorithm, which uses a small number of 'Do' and 'Don't' instruction videos to train an instruction network. This network is then embedded as a distance metric into the unsupervised skill discovery (DSD) framework, allowing the agent to learn desirable complex behaviors without explicit rewards and to avoid unsafe or undesirable behaviors.
Does Egalitarian Fairness Lead to Instability? The Fairness Bounds in Stable Federated Learning Under Altruistic Behaviors
Jiashi Gao (Southern University of Science and Technology), Xuetao Wei (Southern University of Science and Technology)
Federated LearningTabular
🎯 What it does: This study investigates whether pursuing balanced fairness in federated learning leads to coalition instability and provides a core-stable upper bound for fairness.
Does Reasoning Emerge? Examining the Probabilities of Causation in Large Language Models
Javier Gonzalez, Aditya V. Nori (Microsoft Research)
Large Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: A causal probability-based evaluation framework (Necessary PN and Sufficient PS) is proposed to detect the performance of large language models (LLMs) in real reasoning, especially counterfactual reasoning.
Does Video-Text Pretraining Help Open-Vocabulary Online Action Detection?
Qingsong Zhao (Tongji University), Cairong Zhao (Tongji University)
RecognitionObject DetectionTransformerContrastive LearningVideoText
🎯 What it does: A zero-shot online action detection framework, OV-OAD, is proposed, which utilizes large-scale video-text pairs for pre-training and achieves real-time, open-vocabulary action recognition solely based on text supervision.
Does Worst-Performing Agent Lead the Pack? Analyzing Agent Dynamics in Unified Distributed SGD
Jie Hu (North Carolina State University), Do Young Eun (North Carolina State University)
Convolutional Neural NetworkImageTabular
🎯 What it does: This paper studies the asymptotic convergence of Unified Distributed SGD (UD-SGD) under heterogeneous agents, Markov sampling, and various communication modes, providing a central limit theorem and mean square error analysis.
DOFEN: Deep Oblivious Forest ENsemble
Kuan-Yu Chen (Sinopac Holdings), Tien-Hao Chang
ClassificationOptimizationTabularBenchmark
🎯 What it does: The DOFEN model is proposed, a deep differentiable forgettable decision tree forest designed for handling tabular data;
DoFIT: Domain-aware Federated Instruction Tuning with Alleviated Catastrophic Forgetting
Binqian Xu (Nanjing University of Science and Technology), Jinhui Tang (National University of Singapore)
Federated LearningSupervised Fine-TuningTextFinance Related
🎯 What it does: Proposes the DoFIT framework to address the catastrophic forgetting problem caused by cross-domain data heterogeneity in federated instruction tuning;
DOGS: Distributed-Oriented Gaussian Splatting for Large-Scale 3D Reconstruction Via Gaussian Consensus
Yu Chen (National University of Singapore), Gim Hee Lee (National University of Singapore)
Gaussian SplattingPoint Cloud
🎯 What it does: This paper proposes DOGS, a distributed 3D Gaussian splatting training framework that utilizes the ADMM consensus mechanism.
Doing Experiments and Revising Rules with Natural Language and Probabilistic Reasoning
Wasu Top Piriyakulkij (Cornell University), Kevin Ellis (Google)
TransformerLarge Language ModelReinforcement LearningText
🎯 What it does: This study explores how to combine large language models (LLMs) with Monte Carlo methods for online Bayesian inference through experimental design and natural language rule inference.
Domain Adaptation for Large-Vocabulary Object Detectors
Kai Jiang (Xidian University), Shijian Lu
Object DetectionDomain AdaptationKnowledge DistillationGraph Neural NetworkContrastive LearningImage
🎯 What it does: Proposes Knowledge Graph Distillation (KGD) to perform unsupervised domain adaptation for large vocabulary object detectors using the implicit knowledge graph from CLIP.
DomainGallery: Few-shot Domain-driven Image Generation by Attribute-centric Finetuning
Yuxuan Duan (Shanghai Jiao Tong University), Liqing Zhang (Shanghai Jiao Tong University)
GenerationData SynthesisSupervised Fine-TuningDiffusion modelImage
🎯 What it does: Fine-tuning the pre-trained Stable Diffusion in a fine-grained, attribute-driven manner under the condition of a very small number (10 images) of target domain images to achieve cross-category, cross-attribute, and even personalized image generation;
Don't Compress Gradients in Random Reshuffling: Compress Gradient Differences
Abdurakhmon Sadiev (King Abdullah University of Science and Technology), Peter Richtárik (King Abdullah University of Science and Technology)
OptimizationFederated LearningConvolutional Neural NetworkImageTabular
🎯 What it does: This paper studies the combination of distributed random reshuffling (Random Reshuffling, RR) and gradient compression, proposing four new algorithms: Q-RR, DIANA-RR, Q-NASTYA, and DIANA-NASTYA, along with their theoretical convergence analysis.
Don't Look Twice: Faster Video Transformers with Run-Length Tokenization
Rohan Choudhury (Carnegie Mellon University), Laszlo Attila Jeni
RecognitionComputational EfficiencyTransformerVideo
🎯 What it does: A video segmentation method based on run-length encoding, RLT, is proposed, which can significantly reduce the number of input tokens for the Transformer while maintaining performance, thus improving training and inference speed;
Doob's Lagrangian: A Sample-Efficient Variational Approach to Transition Path Sampling
Yuanqi Du (Cornell University), Kirill Neklyudov (Université de Montréal)
OptimizationSequentialPhysics RelatedStochastic Differential Equation
🎯 What it does: A variational framework based on Doob h-transform is proposed for efficient Transition Path Sampling.
DOPPLER: Differentially Private Optimizers with Low-pass Filter for Privacy Noise Reduction
Xinwei Zhang (University of Southern California), Meisam Razaviyayn (University of Southern California)
OptimizationSafty and PrivacyConvolutional Neural NetworkTransformerImage
🎯 What it does: The DOPPLER low-pass filtering module is proposed, which can suppress gradient noise in the frequency domain during differential privacy optimization, thereby improving model performance.
Double-Ended Synthesis Planning with Goal-Constrained Bidirectional Search
Kevin Yu (Massachusetts Institute of Technology), Connor W. Coley (Massachusetts Institute of Technology)
OptimizationDrug DiscoveryReinforcement LearningTabular
🎯 What it does: A bidirectional synthetic planning algorithm DESP is proposed to address the computer-aided synthetic planning problem with starting material constraints.
Doubly Hierarchical Geometric Representations for Strand-based Human Hairstyle Generation
Yunlu Chen (Carnegie Mellon University), Fernando De la Torre (Carnegie Mellon University)
GenerationData SynthesisAuto EncoderPoint CloudMesh
🎯 What it does: This paper designs and trains a dual-layer hierarchical variational autoencoder model for generating complete 3D human hair styles from low-frequency guiding strands to high-frequency detailed strands.
Doubly Mild Generalization for Offline Reinforcement Learning
Yixiu Mao (Tsinghua University), Xiangyang Ji (Tsinghua University)
Reinforcement LearningTabular
🎯 What it does: This paper studies how to moderately utilize generalization capabilities in offline reinforcement learning, proposing the Doubly Mild Generalization (DMG) method, which balances mild action generalization and mild generalization propagation to enhance offline RL performance and achieve seamless transfer to online learning.
DPIC: Decoupling Prompt and Intrinsic Characteristics for LLM Generated Text Detection
Xiao Yu (University of Science and Technology of China), Nenghai Yu (University of Science and Technology of China)
ClassificationGenerationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: A DPIC method has been developed, which reconstructs prompts and regenerates text through an auxiliary LLM, decoupling prompts from internal features to detect LLM-generated text.
DRACO: A Denoising-Reconstruction Autoencoder for Cryo-EM
YingJun Shen, Jingyi Yu (ShanghaiTech University)
RestorationSegmentationTransformerAuto EncoderImageVideo
🎯 What it does: A denoising-reconstruction autoencoder named DRACO is proposed, specifically designed for low signal-to-noise ratio cryo-EM micrographs, which can serve as a foundational model for downstream task transfer.
Drago: Primal-Dual Coupled Variance Reduction for Faster Distributionally Robust Optimization
Ronak Mehta (University of Washington), Zaid Harchaoui (University of Washington)
OptimizationTabularBenchmark
🎯 What it does: A random primal-dual algorithm named DRAGO is proposed to solve the penalized distributionally robust optimization (DRO) problem with closed, convex uncertainty sets.
DreamClear: High-Capacity Real-World Image Restoration with Privacy-Safe Dataset Curation
Yuang Ai (Institute of Automation, Chinese Academy of Sciences), Hongxia Yang (ByteDance, Inc)
RestorationData SynthesisTransformerLarge Language ModelMixture of ExpertsDiffusion modelImage
🎯 What it does: This paper proposes the GenIR unified dataset construction pipeline and two key components, DreamClear, achieving a large-scale, privacy-preserving real image restoration system.
DreamMesh4D: Video-to-4D Generation with Sparse-Controlled Gaussian-Mesh Hybrid Representation
Zhiqi Li (Zhejiang University), Peidong Liu (Westlake University)
GenerationDiffusion modelVideoMesh
🎯 What it does: A new framework called DreamMesh4D is proposed, which generates high-quality 4D dynamic objects from monocular videos through a sparse-controlled Gaussian mesh hybrid representation.
DreamScene4D: Dynamic Multi-Object Scene Generation from Monocular Videos
Wen-Hsuan Chu (Carnegie Mellon University), Katerina Fragkiadaki (Carnegie Mellon University)
GenerationData SynthesisDepth EstimationDiffusion modelGaussian SplattingOptical FlowVideo
🎯 What it does: Proposes DreamScene4D, a method for generating dynamic 4D scenes with multiple objects from monocular video;
DreamSteerer: Enhancing Source Image Conditioned Editability using Personalized Diffusion Models
Zhengyang Yu (Australian National University), Jing Zhang (Australian National University)
Image TranslationGenerationData SynthesisOptimizationDiffusion modelScore-based ModelImage
🎯 What it does: This paper presents DreamSteerer, a pluggable method that significantly enhances the editability of source images under personalized concept conditions by fine-tuning existing text-to-image personalization diffusion models.
Drift-Resilient TabPFN: In-Context Learning Temporal Distribution Shifts on Tabular Data
Kai Helli (University of Freiburg), Frank Hutter (University of Freiburg)
ClassificationDomain AdaptationExplainability and InterpretabilityTransformerTabularFinance Related
🎯 What it does: Proposes Drift-Resilient TabPFN, which utilizes In-Context Learning to achieve adaptive prediction for temporal distribution drift in tabular data;
DRIP: Unleashing Diffusion Priors for Joint Foreground and Alpha Prediction in Image Matting
Xiaodi Li (Zhejiang University), Yi Yang (Zhejiang University)
SegmentationGenerationDiffusion modelImage
🎯 What it does: The DRIP method is proposed, utilizing a pre-trained latent diffusion model to jointly predict the foreground color and alpha mask of a single image, achieving high-quality image matting.
Drones Help Drones: A Collaborative Framework for Multi-Drone Object Trajectory Prediction and Beyond
Zhechao Wang (Chinese Academy of Sciences), Xian Sun (Chinese Academy of Sciences)
Object DetectionObject TrackingConvolutional Neural NetworkGaussian SplattingImage
🎯 What it does: Proposed the DHD framework for trajectory prediction in multi-drone collaboration, which includes ground-prior-based BEV generation and sliding window sparse interaction.
DropBP: Accelerating Fine-Tuning of Large Language Models by Dropping Backward Propagation
Sunghyeon Woo (Seoul National University), Dongsoo Lee (Seoul National University)
OptimizationComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: During the fine-tuning process of LLMs, certain layers of backpropagation are randomly dropped to reduce computational and activation storage overhead.
DropEdge not Foolproof: Effective Augmentation Method for Signed Graph Neural Networks
Zeyu Zhang (Huazhong Agricultural University), Wanli.Li
Graph Neural NetworkGraph
🎯 What it does: Proposes the Signed Graph Augmentation framework (SGA), which addresses the sparsity and imbalance triangle problem in SGNN through structural augmentation, candidate edge selection, and difficulty-based curriculum learning, significantly improving link sign prediction performance.
Du-IN: Discrete units-guided mask modeling for decoding speech from Intracranial Neural signals
Hui Zheng (Peking University), Yunzhe Liu (Beijing Normal University)
ClassificationRecognitionConvolutional Neural NetworkTransformerAuto EncoderBiomedical DataAudio
🎯 What it does: A speech decoding framework called Du-IN based on sEEG has been developed, which achieves speech decoding through mask modeling guided by region-level discrete coding.
DU-Shapley: A Shapley Value Proxy for Efficient Dataset Valuation
Felipe Garrido, Vianney Perchet (ENSAE)
Data-Centric LearningTabular
🎯 What it does: This paper studies the problem of dataset valuation, using the Shapley value to measure the incremental value of each data owner in a joint machine learning task, and proposes a new approximation method - DU-Shapley.
Dual Cone Gradient Descent for Training Physics-Informed Neural Networks
Youngsik Hwang (Ulsan National Institute of Science and Technology), Dongyoung Lim
OptimizationPhysics Related
🎯 What it does: This paper proposes the Dual Cone Gradient Descent (DCGD) framework to address the pathologies caused by gradient imbalance and conflicts during the training of Physics-Informed Neural Networks (PINNs), ensuring that each gradient update falls within the dual cone, thereby achieving multi-objective optimization.
Dual Critic Reinforcement Learning under Partial Observability
Jinqiu Li (Chinese Academy of Sciences), Shiming Xiang (Chinese Academy of Sciences)
Recurrent Neural NetworkTransformerReinforcement LearningSequential
🎯 What it does: A Dual Critic Reinforcement Learning (DCRL) framework is proposed in partially observable environments, which trains the Actor by simultaneously using a standard Critic (observing only history) and an Oracle Critic (having access to the full state), thereby reducing variance and improving learning efficiency while ensuring unbiasedness.
Dual Defense: Enhancing Privacy and Mitigating Poisoning Attacks in Federated Learning
Runhua Xu (Beihang University), Jianxin Li (Beihang University)
Anomaly DetectionFederated LearningSafty and PrivacyConvolutional Neural NetworkImage
🎯 What it does: A dual defense framework DDFed is proposed, which can simultaneously enhance privacy protection and resist model poisoning attacks in federated learning.
Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images
Bahri Batuhan Bilecen (Bilkent University), Aysegul Dundar (Bilkent University)
GenerationData SynthesisNeural Radiance FieldAuto EncoderGenerative Adversarial NetworkImageMesh
🎯 What it does: Utilizing a dual-encoder framework to project a single facial image into the latent space of the PanoHead's 3D generative network, achieving high-fidelity 360-degree 3D head reconstruction and editing;
Dual Lagrangian Learning for Conic Optimization
Mathieu Tanneau (Georgia Institute of Technology), Pascal Van Hentenryck (Georgia Institute of Technology)
OptimizationTabular
🎯 What it does: A Dual Lagrangian Learning (DLL) method is proposed, utilizing dual cone theory and machine learning to predict the dual feasible solutions of parametric cone optimization problems, thereby providing effective Lagrangian dual bounds.
Dual Prototype Evolving for Test-Time Generalization of Vision-Language Models
Ce Zhang (Carnegie Mellon University), Yaqi Xie (Carnegie Mellon University)
ClassificationDomain AdaptationOptimizationTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Proposes the Dual Prototype Evolving (DPE) method, which achieves adaptive generalization of VLM through dual-modal prototype evolution during testing.
Dual Risk Minimization: Towards Next-Level Robustness in Fine-tuning Zero-Shot Models
Kaican Li (Hong Kong University of Science and Technology), Nevin L. Zhang (Hong Kong University of Science and Technology)
Domain AdaptationOptimizationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringImage
🎯 What it does: This paper proposes Dual Risk Minimization (DRM), which balances expected performance and worst-case performance when fine-tuning zero-shot pre-trained models to enhance robustness against distribution shifts.
Dual-Diffusion for Binocular 3D Human Pose Estimation
Xiaoyue Wan (Shanghai Jiao Tong University), Xu Zhao (Shanghai Jiao Tong University)
Pose EstimationGraph Neural NetworkTransformerDiffusion modelImageVideo
🎯 What it does: A Dual-Diffusion framework is proposed, utilizing diffusion models to simultaneously denoise binocular 2D keypoints and their corresponding 3D joints, thereby improving the accuracy of 3D human pose estimation.
Dual-frame Fluid Motion Estimation with Test-time Optimization and Zero-divergence Loss
Yifei Zhang (University of Chinese Academy of Sciences), Hao Zhao (Tsinghua University)
Object TrackingOptimizationGraph Neural NetworkOptical FlowVideo
🎯 What it does: This paper studies a fully self-supervised three-dimensional particle tracking velocimetry (PTV) framework that implements dual-frame flow estimation through optimization during testing.
Dual-Personalizing Adapter for Federated Foundation Models
yiyuan yang, Michael Blumenstein (University of Technology Sydney)
Federated LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes FedDPA—a dual personalized adapter designed to simultaneously address client personalization and testing distribution shift issues within the Federated Foundation Model (FedFM) framework.
Dual-Perspective Activation: Efficient Channel Denoising via Joint Forward-Backward Criterion for Artificial Neural Networks
Tian Qiu (Zhejiang University), Mingli Song (Zhejiang University)
ClassificationOptimizationComputational EfficiencyConvolutional Neural NetworkTransformerImageTextGraph
🎯 What it does: This paper proposes an end-to-end trainable Dual-Perspective Activation (DPA) mechanism, which identifies and suppresses the activation of irrelevant channels in neural networks through online forward and backward criteria, thereby achieving channel denoising and sparse representation.
Dueling over Dessert, Mastering the Art of Repeated Cake Cutting
Simina Branzei, Kun Wang (Purdue University)
🎯 What it does: Analyzed the fairness problem of two players repeatedly distributing a cake, proposed a strategy for Alice to utilize Bob's greedy strategy to achieve fair outcomes, and proved that fair outcomes can be reached in both sequential and parallel settings. Further discussed the convergence properties of fictitious player learning dynamics.
DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs
Haokun Lin (University of Chinese Academy of Sciences), Ying Wei (Zhejiang University)
Anomaly DetectionOptimizationComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: A low-bit quantization method for LLM named DuQuant is proposed, which effectively suppresses large-scale outliers (including ordinary and large-scale outliers) in activations through rotation and permutation transformations, thereby improving the performance of 4/6-bit quantized models.
Dynamic 3D Gaussian Fields for Urban Areas
Tobias Fischer (ETH Zurich), Peter Kontschieder (Meta Reality Labs)
Data SynthesisAutonomous DrivingComputational EfficiencyGaussian SplattingPoint Cloud
🎯 What it does: An efficient renderable neural field representation for dynamic urban environments has been constructed—4D Gaussian Fields (4DGF), which can generate high-quality interactive view synthesis from multi-sequence heterogeneous data.
Dynamic Conditional Optimal Transport through Simulation-Free Flows
Gavin Kerrigan (University of California, Irvine), Padhraic Smyth (University of California, Irvine)
GenerationOptimizationFlow-based ModelTabularTime Series
🎯 What it does: A theory of dynamic conditional optimal transport is proposed, and based on this, a conditional generative model (COT-FM) without simulation flow matching is constructed.
Dynamic Model Predictive Shielding for Provably Safe Reinforcement Learning
Arko Banerjee (University of Texas at Austin), Isil Dillig (University of Texas at Austin)
OptimizationSafty and PrivacyReinforcement Learning
🎯 What it does: This paper proposes a Dynamic Model Predictive Shielding (DMPS) method that combines local planning with deep learning strategies to achieve safe and efficient reinforcement learning.
Dynamic Neural Regeneration: Enhancing Deep Learning Generalization on Small Datasets
Vijaya Raghavan T Ramkumar (Eindhoven University of Technology), Bahram Zonooz (Eindhoven University of Technology)
ClassificationData-Centric LearningConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: A dynamic neural regeneration (DNR) iterative training framework is proposed, which enhances the generalization ability of small sample datasets by reinitializing part of the parameters based on a data-aware significance mask at the end of each training generation.
Dynamic Rescaling for Training GNNs
Nimrah Mustafa (CISPA Helmholtz Center for Information Security), Rebekka Burkholz (CISPA Helmholtz Center for Information Security)
Graph Neural NetworkGraph
🎯 What it does: This paper proposes dynamically scaling parameters (dynamic re-calibration) during the training process of GAT networks to change the gradient scale while keeping the network function unchanged, thereby controlling the training dynamics.
Dynamic Service Fee Pricing under Strategic Behavior: Actions as Instruments and Phase Transition
Rui Ai (Massachusetts Institute of Technology), Feng Zhu (Massachusetts Institute of Technology)
Recommendation SystemOptimizationReinforcement LearningTabular
🎯 What it does: This paper addresses the dynamic pricing problem of service fees on third-party platforms, focusing on how to maximize revenue in the face of strategic and forward-looking consumers when there is a lack of demand information, only equilibrium prices and quantities can be observed, and consumers may conceal their preferences.
Dynamic Subgroup Identification in Covariate-adjusted Response-adaptive Randomization Experiments
Yanping Li (Nankai University), Waverly Wei (University of Southern California)
OptimizationDrug DiscoveryReinforcement LearningTabularBiomedical Data
🎯 What it does: A dynamic subgroup identification method is proposed within the CARA experimental framework to identify the best-performing subgroups in real-time during randomized clinical trials and adjust treatment allocation.
Dynamic Tuning Towards Parameter and Inference Efficiency for ViT Adaptation
Wangbo Zhao (National University of Singapore), Yang You (National University of Singapore)
ClassificationSegmentationOptimizationComputational EfficiencyTransformerMixture of ExpertsImageVideo
🎯 What it does: A dynamic tuning framework named DyT is proposed to simultaneously enhance parameter efficiency and inference efficiency during the adaptation process of Vision Transformers (ViT).
Dynamics of Supervised and Reinforcement Learning in the Non-Linear Perceptron
Christian Schmid (Institute of Neuroscience University of Oregon), James M Murray
Reinforcement LearningImage
🎯 What it does: Analyze the learning dynamics of nonlinear perceptrons under supervised learning and reinforcement learning, and establish a flow equation model.
DynaMITE-RL: A Dynamic Model for Improved Temporal Meta-Reinforcement Learning
Anthony Liang (University of Southern California), Craig Boutilier (Google Research)
Meta LearningRecurrent Neural NetworkTransformerReinforcement LearningSequential
🎯 What it does: Proposed the DynaMITE-RL method, designed the DLCMDP framework for the meta-RL problem where latent states change over time, and implemented it in both online and offline training;
DynaMo: In-Domain Dynamics Pretraining for Visuo-Motor Control
Zichen Jeff Cui (New York University), Lerrel Pinto (New York University)
Representation LearningRobotic IntelligenceConvolutional Neural NetworkTransformerReinforcement LearningVideo
🎯 What it does: Learning visual representations for visual motion control through self-supervised dynamic pre-training (DynaMo) in the domain.
e-COP : Episodic Constrained Optimization of Policies
Akhil Agnihotri (University of Southern California), Sahil Singla (Google DeepMind)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes a policy optimization algorithm for finite-horizon constrained reinforcement learning, e-COP, which achieves precise control over safety constraints for the first time in an episodic setting.
E-Motion: Future Motion Simulation via Event Sequence Diffusion
Song Wu (Xidian University), Jinjian Wu (Xidian University)
GenerationData SynthesisReinforcement LearningDiffusion modelVideoSequentialStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: This paper proposes an Event-Sequence Diffusion Network, which utilizes high spatiotemporal resolution data from event cameras as conditional input. It generates future event sequences through pre-training, reinforcement learning alignment, and multi-frame enhancement during testing to predict object motion.
E2E-MFD: Towards End-to-End Synchronous Multimodal Fusion Detection
Jiaqing Zhang (Xidian University), Xue Yang (Southeast University)
Object DetectionConvolutional Neural NetworkDiffusion modelImageMultimodality
🎯 What it does: An end-to-end multimodal fusion detection framework E2E-MFD is proposed, capable of completing the training and inference of visible-infrared image fusion and object detection in one go.
E2ENet: Dynamic Sparse Feature Fusion for Accurate and Efficient 3D Medical Image Segmentation
Boqian Wu (University of Twente), Elena Mocanu (University of Twente)
SegmentationConvolutional Neural NetworkBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: A 3D medical image segmentation network called E2ENet is proposed, which mainly enhances segmentation accuracy while reducing model size and computational load through Dynamic Sparse Feature Fusion (DSFF) and Restricted Depth-Shift.
EAGLE: Efficient Adaptive Geometry-based Learning in Cross-view Understanding
Thanh-Dat Truong (University of Arkansas), Khoa Luu (University of Arkansas)
SegmentationDomain AdaptationAutonomous DrivingPrompt EngineeringImage
🎯 What it does: This paper proposes an unsupervised cross-view adaptation method called EAGLE to enhance semantic segmentation performance from ground vehicle perspectives to drone perspectives.
EAI: Emotional Decision-Making of LLMs in Strategic Games and Ethical Dilemmas
Mikhail Mozikov (AIRI), Ilya Makarov
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: The paper proposes the EAI framework, which injects emotions into LLMs and evaluates their performance in ethical judgment and game decision-making.
EASI: Evolutionary Adversarial Simulator Identification for Sim-to-Real Transfer
Haoyu Dong (Nanjing University), Chunlin Chen (Nanjing University)
Domain AdaptationOptimizationRobotic IntelligenceReinforcement LearningGenerative Adversarial NetworkSequential
🎯 What it does: A method called EASI based on GAN and evolutionary strategies is proposed to identify the physical parameter distribution that is most similar to reality, thereby achieving efficient sim-to-real transfer.
Easy Regional Contrastive Learning of Expressive Fashion Representations
Daiqing Qi (University of Virginia), Sheng Li (University of Virginia)
RetrievalRepresentation LearningTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a lightweight regional contrastive learning framework E², based on a simple structure of CLIP. By inserting selection tokens and a fusion block into the visual encoder and introducing regional contrastive loss, it significantly enhances the fine-grained representation and cross-modal retrieval performance of fashion domain visual-language models.
Easy-to-Hard Generalization: Scalable Alignment Beyond Human Supervision
Zhiqing Sun (Carnegie Mellon University), Chuang Gan
Supervised Fine-TuningReinforcement LearningText
🎯 What it does: This paper proposes the 'easy-to-difficult generalization' of transferring human supervision from easy tasks to hard tasks, by first training an evaluation model and then using it to assess or guide the generator through reinforcement learning, thereby improving model performance without human supervision on hard tasks.
ECLipsE: Efficient Compositional Lipschitz Constant Estimation for Deep Neural Networks
Yuezhu Xu (Purdue University), S Sivaranjani
OptimizationComputational EfficiencyConvolutional Neural NetworkTabularStochastic Differential Equation
🎯 What it does: This paper proposes a scalable Lipschitz constant estimation framework, which breaks down large-scale SDP verification problems into smaller subproblems layer by layer, introducing two algorithms: ECLipsE (using small-scale SDP) and ECLipsE-Fast (using closed-form solutions);
ECMamba: Consolidating Selective State Space Model with Retinex Guidance for Efficient Multiple Exposure Correction
Wei Dong (McMaster University), Jun Chen (McMaster University)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Proposes a dual-branch Mamba network based on the Retinex theory for exposure correction of underexposed/overexposed images;
Edit Distance Robust Watermarks via Indexing Pseudorandom Codes
Noah Golowich (Massachusetts Institute of Technology), Ankur Moitra (Massachusetts Institute of Technology)
Text
🎯 What it does: A theoretically feasible watermarking scheme is proposed, which can resist attacks involving constant proportion insertion, deletion, and replacement of text generated by language models while maintaining undetectability.
EDT: An Efficient Diffusion Transformer Framework Inspired by Human-like Sketching
Xinwang Chen (Midea Group), Jian Tang (Beijing Innovation Center of Humanoid Robotics)
GenerationData SynthesisComputational EfficiencyTransformerDiffusion modelImage
🎯 What it does: An efficient diffusion Transformer framework EDT is proposed, balancing speed and quality;
EEG2Video: Towards Decoding Dynamic Visual Perception from EEG Signals
Xuanhao Liu, Wei-Long Zheng (Shanghai Jiao Tong University)
ClassificationGenerationData SynthesisTransformerDiffusion modelVideoMultimodalityBenchmark
🎯 What it does: The first large-scale EEG-video alignment dataset SEED-DV (20 subjects, 1400 two-second videos) was constructed, and EEG visual perception classification benchmarks and video reconstruction benchmarks were designed on it, proposing a Seq2Seq-based EEG2Video video reconstruction framework.
EEGPT: Pretrained Transformer for Universal and Reliable Representation of EEG Signals
Guangyu Wang, Haifeng Li (Harbin Institute of Technology)
Representation LearningTransformerContrastive LearningTime SeriesBiomedical DataElectrocardiogram
🎯 What it does: Developed a 10M parameter EEGPT pre-trained Transformer model to learn general and robust EEG representations from multi-task, multi-modal EEG data, achieving state-of-the-art performance on downstream tasks through linear probing.
Effective Exploration Based on the Structural Information Principles
Xianghua Zeng (Beihang University), Angsheng Li (Zhongguancun Laboratory)
Reinforcement LearningTabular
🎯 What it does: A framework called SI2E based on the principle of structural information is proposed for efficient exploration in reinforcement learning.
Effective Rank Analysis and Regularization for Enhanced 3D Gaussian Splatting
Junha Hyung (Korea Advanced Institute of Science and Technology), Jin-Hwa Kim (NAVER AI Lab)
Gaussian SplattingPoint Cloud
🎯 What it does: An effective rank analysis and regularization method is proposed to improve the issues of thin voxels and surface roughness in 3D Gaussian Splatting.
Efficiency for Free: Ideal Data Are Transportable Representations
Peng Sun (Westlake University), Tao Lin (Westlake University)
Computational EfficiencyKnowledge DistillationRepresentation LearningConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes the Representation Learning Accelerator (RELA) framework, which utilizes publicly available pre-trained models as labelers to generate high-quality efficient data, thereby accelerating the representation learning process.
Efficiency of the First-Price Auction in the Autobidding World
Yuan Deng (Google Research), Song Zuo (Google Research)
Optimization
🎯 What it does: This study investigates the efficiency of first-price auctions in environments with fully automated bidders and hybrid bidders, providing precise lower and upper bounds for the Price of Anarchy (PoA).
Efficient $\Phi$-Regret Minimization with Low-Degree Swap Deviations in Extensive-Form Games
Brian Hu Zhang (Carnegie Mellon University), Tuomas Sandholm (Carnegie Mellon University)
Optimization
🎯 What it does: This paper studies effective algorithms for minimizing Φ-regret for low-order exchange biases (k-mediator and low-degree polynomial biases) in Extensive-Form Games, and provides a multi-parameter complexity analysis.
Efficient Adaptation of Pre-trained Vision Transformer via Householder Transformation
Wei Dong (Xi'an University of Architecture and Technology), Heng Tao Shen
ClassificationOptimizationTransformerSupervised Fine-TuningImage
🎯 What it does: A parameter-efficient fine-tuning method based on Householder transformation, HTA, is proposed, which can achieve task transfer on pre-trained Vision Transformers with very few trainable parameters.
Efficient Adversarial Training in LLMs with Continuous Attacks
Sophie Xhonneux (Mila Université de Montréal), Leo Schwinn (Technical University of Munich)
Computational EfficiencyAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: An efficient continuous adversarial training algorithm CAT and CAPO is proposed, enhancing the robustness of LLM against discrete adversarial attacks while maintaining practicality.
Efficient and Private Marginal Reconstruction with Local Non-Negativity
Brett Mullins (University of Massachusetts), Daniel Sheldon (University of Massachusetts)
Safty and PrivacyComputational EfficiencyTabular
🎯 What it does: This paper proposes a post-processing method based on residual queries, called ReM, for efficiently reconstructing multi-dimensional marginal queries in a differentially private (DP) environment, and extends to Gaussian noise and local non-negativity constraints.
Efficient and Sharp Off-Policy Evaluation in Robust Markov Decision Processes
Andrew Bennett (Morgan Stanley), Kaiwen Wang (Cornell University)
Reinforcement LearningBiomedical DataElectronic Health Records
🎯 What it does: A precise evaluation framework for the best/worst policy value in model-free reinforcement learning under unknown environmental disturbances is proposed.
Efficient Availability Attacks against Supervised and Contrastive Learning Simultaneously
Yihan Wang (Academy of Mathematics and Systems Science, Chinese Academy of Sciences), Xiao-Shan Gao (Academy of Mathematics and Systems Science, Chinese Academy of Sciences)
Adversarial AttackContrastive LearningImage
🎯 What it does: A usability attack method that is effective for both supervised learning and contrastive learning is proposed, introducing two enhanced non-learning example attacks: AUE and AAP;
Efficient Centroid-Linkage Clustering
Mohammadhossein Bateni, Jakub Lacki
OptimizationComputational EfficiencyTabular
🎯 What it does: This paper proposes an approximate Centroid-Linkage Hierarchical Clustering (HAC) algorithm that can complete clustering in sub-quadratic time.
Efficient Combinatorial Optimization via Heat Diffusion
Hengyuan Ma (Institute of Science and Technology for Brain-inspired Intelligence Fudan University), Jianfeng Feng (Institute of Science and Technology for Brain-inspired Intelligence Fudan University)
OptimizationGraphOrdinary Differential Equation
🎯 What it does: A combination optimization framework based on thermal diffusion, HeO, is proposed, which utilizes the heat equation to smooth the objective function and introduces a thermal diffusion parameter during the gradient descent process to expand the search receptive field and improve the efficiency of global optimal search.
Efficient Contextual LLM Cascades through Budget-Constrained Policy Learning
Xuechen Zhang (University of Michigan), Jiasi Chen (University of Michigan)
OptimizationComputational EfficiencyTransformerLarge Language ModelReinforcement LearningPrompt EngineeringText
🎯 What it does: The TREACLE framework is proposed, which uses reinforcement learning to automatically select suitable LLMs and prompting strategies for each problem under budget constraints, and can perform re-querying based on query context and historical answers to improve accuracy.
Efficient Discrepancy Testing for Learning with Distribution Shift
Gautam Chandrasekaran (University of Texas at Austin), Arsen Vasilyan (University of California Berkeley)
Domain AdaptationComputational Efficiency
🎯 What it does: This paper proposes a novel decision-making method for distribution shift learning, primarily by efficiently testing the 'localized discrepancy' to construct a series of unified, efficient, and testable learning algorithms.
Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability
Ming Xiang (Northeastern University), Lili Su (Northeastern University)
OptimizationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: This paper proposes the FedAWE algorithm, which addresses the bias problem caused by client availability heterogeneity and non-stationarity in federated learning. The core idea is to adaptively compensate for missing updates through innovative echo compensation and achieve information balance using implicit graph theory mixing.
Efficient Graph Matching for Correlated Stochastic Block Models
Shuwen Chai (Northwestern University), Miklos Z. Racz
Graph Neural NetworkGraph
🎯 What it does: A polynomial-time algorithm is proposed that can perform graph matching in two types of balanced community correlated stochastic block models (correlated SBM), achieving nearly complete matching or even complete matching when the average degree is logarithmic; subsequently, this algorithm is applied to community recovery, providing an efficient solution within the information-theoretic feasible region.
Efficient Large Multi-modal Models via Visual Context Compression
Jieneng Chen (Johns Hopkins University), Alan Yuille (Johns Hopkins University)
CompressionComputational EfficiencyTransformerLarge Language ModelVision Language ModelImageVideoMultimodality
🎯 What it does: This paper addresses the issue of visual token redundancy in multimodal large language models (MLLMs) by proposing the Visual Context Compressor and the staged training framework LLaVolta, significantly improving training and inference efficiency while maintaining or even enhancing model performance.
Efficient Leverage Score Sampling for Tensor Train Decomposition
Vivek Bharadwaj (University of California Berkeley), Guillaume Rabusseau (Mila and Université de Montréal)
OptimizationComputational EfficiencyTabular
🎯 What it does: A randomized ALS algorithm based on precise leverage score sampling (rTT-ALS) is proposed for Tensor Train decomposition of high-dimensional tensors.
Efficient Lifelong Model Evaluation in an Era of Rapid Progress
Ameya Prabhu (University of Tübingen), Samuel Albanie (University of Cambridge)
Computational EfficiencyImage
🎯 What it does: A framework called Sort & Search is proposed, which utilizes the evaluation results of existing models for sample sorting and subset sampling, achieving efficiency in lifelong model evaluation.
Efficient LLM Jailbreak via Adaptive Dense-to-sparse Constrained Optimization
Kai Hu (Carnegie Mellon University), Matt Fredrikson (Carnegie Mellon University)
OptimizationAdversarial AttackTransformerLarge Language ModelText
🎯 What it does: This paper proposes a new token-level attack method called Adaptive Dense-to-Sparse Constrained Optimization (ADC) to break the security defenses of large language models.
Efficient LLM Scheduling by Learning to Rank
Yichao Fu (University of California San Diego), Hao Zhang
GenerationData SynthesisOptimizationComputational EfficiencyTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Construct a lightweight generative length ranking predictor and combine it with the vLLM service to dynamically schedule LLM requests based on learned ranking information, significantly reducing HOL blocking, improving chat latency, and increasing synthetic data generation throughput.
Efficient Minimum Bayes Risk Decoding using Low-Rank Matrix Completion Algorithms
Firas Trabelsi (Google), Markus Freitag (Google)
OptimizationComputational EfficiencyText
🎯 What it does: This paper proposes an MBR decoding approximation method based on low-rank matrix completion, significantly reducing computational load;