Deep Learning
Chain of Agents: Large Language Models Collaborating on Long-Context Tasks
Input reduction reduces the length of the input context before feeding to downstream LLMs.
SDformer: Similarity-driven Discrete Transformer For Time Series Generation
Comprehensive experiments show that our method significantly outperforms competing approaches in terms of the generated time series quality while also ensuring a short inference time.
Hierarchical Object-Aware Dual-Level Contrastive Learning for Domain Generalized Stereo Matching
Stereo matching aims to find horizontal pixel-wise displacement, i.e .disparity, between a rectified stereo image pair to recover depth for applications including autonomous driving, robotics, and augmented reality.
Transformers as Game Players: Provable In-context Game-playing Capabilities of Pre-trained Models
The in-context learning (ICL) capability of pre-trained models based on the transformer architecture has received growing interest in recent years.
A Label is Worth a Thousand Images in Dataset Distillation
Understanding how and why data distillation methods work is vital not only for improving these methods but also for revealing fundamental characteristics of "good" training data.
Neglected Hessian component explains mysteries in sharpness regularization
SAM can improve generalization in deep learning. Seemingly similar methods like weight noise and gradient penalties often fail to provide such benefits. We investigate this inconsistency and reveal its connection to the the structure of the Hessian of the loss.
Energy-based Hopfield Boosting for Out-of-Distribution Detection
Hopfield Boosting, a boosting approach, which leverages modern Hopfield energy to sharpen the decision boundary between the in-distribution and OOD data.