Na Li

TUMIX: Multi-Agent Test-Time Scaling with Tool-Use Mixture

R1-Code-Interpreter: LLMs Reason with Code via Supervised and Multi-stage Reinforcement Learning

Model-Based Diffusion Sampling for Predictive Control in Offline Decision Making

Reinforcement Learning with Discrete Diffusion Policies for Combinatorial Action Spaces

One-Step Flow Policy Mirror Descent

Zeroth-Order Constrained Optimization from a Control Perspective via Feedback Linearization

RODS: Robust Optimization Inspired Diffusion Sampling for Detecting and Reducing Hallucination in Generative Models

Constrained Optimization From a Control Perspective via Feedback Linearization

Efficient Online Reinforcement Learning for Diffusion Policy

Constrained Optimization From a Control Perspective via Feedback Linearization