← Search

Dennis Duan

2 accepted papers

2025

Restructuring Vector Quantization with the Rotation Trick

ICLR 2025oral

Vector Quantized Variational AutoEncoders (VQ-VAEs) are designed to compress a continuous input to a discrete latent space and reconstruct it with minimal distortion. They operate by maintaining a set of vectors---often referred to as the codebook---and quantizing each encoder output to the nearest…

Cited by 3SourcePDFScholar
2024

Context-Aware Meta-Learning

ICLR 2024poster

Large Language Models like ChatGPT demonstrate a remarkable capacity to learn new concepts during inference without any fine-tuning. However, visual models trained to detect new objects during inference have been unable to replicate this ability, and instead either perform poorly or require meta-tra…