About
Research & Development
StreamME: Creating 3D Avatars Instantly from Live Streams
TLDR: StreamME is a novel method for rapidly reconstructing high-quality 3D head avatars directly from live video streams. It uses an "on-the-fly training" approach...
Research & Development
Boosting LLM Teamwork: A New Framework for Adaptive Document Understanding
TLDR: A new framework for multi-agent LLM systems improves document understanding by enabling dynamic task routing, bidirectional feedback, and parallel agent evaluation. This approach,...
Research & Development
Targeted Adjustments: How Reinforcement Learning Sparsely Fine-Tunes Large Language Models
TLDR: Reinforcement learning (RL) fine-tuning of large language models (LLMs) surprisingly updates only a small fraction (5-30%) of parameters, leaving most unchanged. This sparse...
Research & Development
Adaptive AI: How ARIA Agents Learn and Evolve with Human Guidance
TLDR: ARIA (Adaptive Reflective Interactive Agent) is a new LLM agent framework that learns continuously at test time by interacting with human experts. It...
Research & Development
SADA: Enhancing Diffusion Model Efficiency Through Adaptive Acceleration
TLDR: SADA (Stability-guided Adaptive Diffusion Acceleration) is a novel training-free framework that significantly speeds up ODE-based generative models like diffusion models. It achieves this...
Research & Development
JAM: A Two-Stage Framework for Enhanced Multi-Agent Motion Prediction in Autonomous Driving
TLDR: JAM is a novel two-stage framework for multi-agent motion prediction in autonomous driving. It first generates diverse marginal trajectory proposals by classifying trajectory...