Ctrl+K
- $C^2M^3$_ Cycle-Consistent Multi-Model Merging
- $E^3$_ Exploring Embodied Emotion Through A Large-Scale Egocentric Video Dataset
- $SE(3)$ Equivariant Ray Embeddings for Implicit Multi-View Depth Estimation
- $_beta$-DPO_ Direct Preference Optimization with Dynamic $_beta$
- $_boldsymbol{_mu}_mathbf{P^2}$_ Effective Sharpness Aware Minimization Requires Layerwise Perturbation Scaling
- $_epsilon$-Softmax_ Approximating One-Hot Vectors for Mitigating Label Noise
- $_nabla^2$DFT_ A Universal Quantum Chemistry Dataset of Drug-Like Molecules and a Benchmark for Neural Network Potentials
- $_textit{Bifr__ost}$_ 3D-Aware Image Compositing with Language Instructions
- $_textit{NeuroPath}$_ A Neural Pathway Transformer for Joining the Dots of Human Connectomes
- $_textit{Read-ME}$_ Refactorizing LLMs as Router-Decoupled Mixture of Experts with System Co-Design
- $_textit{Trans-LoRA}$_ towards data-free Transferable Parameter Efficient Finetuning
- $_texttt{ConflictBank}$_ A Benchmark for Evaluating the Influence of Knowledge Conflicts in LLMs
- $_texttt{Model-GLUE}$_ Democratized LLM Scaling for A Large Model Zoo in the Wild
- $_texttt{dattri}$_ A Library for Efficient Data Attribution
- $_texttt{pfl-research}$_ simulation framework for accelerating research in Private Federated Learning
- $_text{Di}^2_text{Pose}$_ Discrete Diffusion Model for Occluded 3D Human Pose Estimation
- $_text{ID}^3$_ Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition
- (FL)$^2$_ Overcoming Few Labels in Federated Semi-Supervised Learning
- 2D-OOB_ Attributing Data Contribution Through Joint Valuation Framework
- 2DQuant_ Low-bit Post-Training Quantization for Image Super-Resolution
- 3-in-1_ 2D Rotary Adaptation for Efficient Finetuning, Efficient Batching and Composability
- 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction
- 3D Focusing-and-Matching Network for Multi-Instance Point Cloud Registration
- 3D Gaussian Rendering Can Be Sparser_ Efficient Rendering via Learned Fragment Pruning
- 3D Gaussian Splatting as Markov Chain Monte Carlo
- 3D Structure Prediction of Atomic Systems with Flow-based Direct Preference Optimization
- 3DCoMPaT200_ Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition
- 3DET-Mamba_ Causal Sequence Modelling for End-to-End 3D Object Detection
- 3DGS-Enhancer_ Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors
- 4+3 Phases of Compute-Optimal Neural Scaling Laws
- 4-bit Shampoo for Memory-Efficient Network Training
- 4D Gaussian Splatting in the Wild with Uncertainty-Aware Regularization
- 4DBInfer_ A 4D Benchmarking Toolbox for Graph-Centric Predictive Modeling on RDBs
- 4Diffusion_ Multi-view Video Diffusion Model for 4D Generation
- 4M-21_ An Any-to-Any Vision Model for Tens of Tasks and Modalities
- 4Real_ Towards Photorealistic 4D Scene Generation via Video Diffusion Models
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings
- A Bayesian Approach to Data Point Selection
- A Benchmark Dataset for Event-Guided Human Pose Estimation and Tracking in Extreme Conditions
- A Benchmark Suite for Evaluating Neural Mutual Information Estimators on Unstructured Datasets
- A Best-of-both-worlds Algorithm for Bandits with Delayed Feedback with Robustness to Excessive Delays
- A Boosting-Type Convergence Result for AdaBoost.MH with Factorized Multi-Class Classifiers
- A Canonicalization Perspective on Invariant and Equivariant Learning
- A Careful Examination of Large Language Model Performance on Grade School Arithmetic
- A Cat Is A Cat (Not A Dog!)_ Unraveling Information Mix-ups in Text-to-Image Encoders through Causal Analysis and Embedding Optimization
- A Closer Look at AUROC and AUPRC under Class Imbalance
- A Closer Look at the CLS Token for Cross-Domain Few-Shot Learning
- A Combinatorial Algorithm for the Semi-Discrete Optimal Transport Problem
- A Compositional Atlas for Algebraic Circuits
- A Comprehensive Analysis on the Learning Curve in Kernel Ridge Regression