๐ข KAIST
GTA: Generative Trajectory Augmentation with Guidance for Offline Reinforcement Learning
ยท3982 wordsยท19 minsยท
loading
ยท
loading
Machine Learning
Reinforcement Learning
๐ข KAIST
Generative Trajectory Augmentation (GTA) significantly boosts offline reinforcement learning by generating high-reward trajectories using a conditional diffusion model, enhancing algorithm performanceโฆ
GrounDiT: Grounding Diffusion Transformers via Noisy Patch Transplantation
ยท2589 wordsยท13 minsยท
loading
ยท
loading
Multimodal Learning
Vision-Language Models
๐ข KAIST
GrounDiT: Training-free spatial grounding for text-to-image generation using Diffusion Transformers and a novel noisy patch transplantation technique for precise object placement.
Generalizable Person Re-identification via Balancing Alignment and Uniformity
ยท3010 wordsยท15 minsยท
loading
ยท
loading
AI Generated
Computer Vision
Face Recognition
๐ข KAIST
Balancing Alignment and Uniformity (BAU) framework improves generalizable person re-identification by mitigating the polarized effects of data augmentation, achieving state-of-the-art performance.
Exploiting Representation Curvature for Boundary Detection in Time Series
ยท2189 wordsยท11 minsยท
loading
ยท
loading
Machine Learning
Self-Supervised Learning
๐ข KAIST
RECURVE: A novel boundary detection method leveraging representation trajectory curvature, surpassing state-of-the-art techniques by accommodating both gradual and abrupt changes in time series.
Exactly Minimax-Optimal Locally Differentially Private Sampling
ยท1615 wordsยท8 minsยท
loading
ยท
loading
AI Theory
Privacy
๐ข KAIST
This paper provides the first exact minimax-optimal mechanisms for locally differentially private sampling, applicable across all f-divergences.
EPIC: Effective Prompting for Imbalanced-Class Data Synthesis in Tabular Data Classification via Large Language Models
ยท5652 wordsยท27 minsยท
loading
ยท
loading
AI Generated
Machine Learning
Few-Shot Learning
๐ข KAIST
EPIC: Effective prompting makes LLMs generate high-quality synthetic tabular data, significantly boosting imbalanced-class classification.
Effective Rank Analysis and Regularization for Enhanced 3D Gaussian Splatting
ยท2803 wordsยท14 minsยท
loading
ยท
loading
AI Generated
Computer Vision
3D Vision
๐ข KAIST
Effective rank regularization enhances 3D Gaussian splatting, resolving needle-like artifacts and improving 3D model quality.
Do's and Don'ts: Learning Desirable Skills with Instruction Videos
ยท2781 wordsยท14 minsยท
loading
ยท
loading
AI Generated
Machine Learning
Reinforcement Learning
๐ข KAIST
DoDont, a novel algorithm, uses instruction videos to guide unsupervised skill discovery, effectively learning desirable behaviors while avoiding undesirable ones in complex continuous control tasks.
Direct Consistency Optimization for Robust Customization of Text-to-Image Diffusion models
ยท3011 wordsยท15 minsยท
loading
ยท
loading
Computer Vision
Image Generation
๐ข KAIST
Boosting personalized image generation! Direct Consistency Optimization (DCO) fine-tunes text-to-image models, ensuring subject consistency and prompt fidelity, even when merging separately customizedโฆ
Differential Privacy in Scalable General Kernel Learning via $K$-means Nystr{"o}m Random Features
ยท1468 wordsยท7 minsยท
loading
ยท
loading
AI Generated
AI Theory
Privacy
๐ข KAIST
Differentially private scalable kernel learning is achieved via a novel DP K-means Nystrรถm method, enabling efficient and accurate model training for general kernels while safeguarding privacy.
Adaptive $Q$-Aid for Conditional Supervised Learning in Offline Reinforcement Learning
ยท3193 wordsยท15 minsยท
loading
ยท
loading
Machine Learning
Reinforcement Learning
๐ข KAIST
Q-Aided Conditional Supervised Learning (QCS) effectively combines the stability of return-conditioned supervised learning with the stitching ability of Q-functions, achieving superior offline reinforโฆ
A Unified Confidence Sequence for Generalized Linear Models, with Applications to Bandits
ยท1965 wordsยท10 minsยท
loading
ยท
loading
AI Generated
Machine Learning
Reinforcement Learning
๐ข KAIST
A unified confidence sequence (CS) construction for generalized linear models (GLMs) achieves state-of-the-art regret bounds for contextual bandits, notably a poly(S)-free regret for logistic bandits.
(FL)$^2$: Overcoming Few Labels in Federated Semi-Supervised Learning
ยท2049 wordsยท10 minsยท
loading
ยท
loading
AI Generated
Machine Learning
Federated Learning
๐ข KAIST
Federated Semi-Supervised Learning (FSSL) struggles with limited labeled data. (FL)ยฒ bridges this gap using adaptive thresholding, sharpness-aware consistency regularization, and learning status-awarโฆ