Qi Wu

Achieving higher compression rates without sacrificing quality, this method eliminates the need for mesh storage in unstructured volumes.

Automated architecture search can enhance embodied agent performance, but it also reveals critical challenges that could hinder optimization.

Learning policies for distributional outcomes reveals that regret rates are intricately linked to policy class complexity, challenging conventional wisdom in offline policy learning.

AEGIS blocks all known malicious-router attacks by confining plaintext handling to a secure enclave, making it a game-changer for LLM API security.

By integrating amplitude-phase factorization with bespoke implicit representations, FDF sets a new benchmark in screen content image super-resolution, outperforming existing methods.

A real-time generative world model can synthesize complex driving scenarios that traditional simulators struggle to capture, enabling safer and more effective evaluation of autonomous vehicle policies.

Text-to-image models can now generate megapixel images 6x faster and with better quality by replacing traditional decoders with a pixel diffusion-based upsampler.

LLM agents often say one thing, believe another, and do something completely different, especially when interacting with other agents.

Video diffusion models already contain implicit multi-view knowledge, making them surprisingly effective for novel view synthesis when adapted to ignore temporal coherence.

VLNVerse tackles the sim-to-real gap in vision-language navigation by providing a unified, large-scale benchmark with realistic physics simulation and full-kinematics embodied agents.