PPL
Qian Wang
Papers (5)
Efficient On-policy Visual-RL via Stochastic Decoupled Policy GradientTuning-free Instruction-based Video Editing Via Structural Noise Initialization and GuidanceTurning Stale Gradients into Stable Gradients: Coherent Coordinate Descent with Implicit Landscape Smoothing for Lightweight Zeroth-Order OptimizationMulti-Token Residual PredictionAn Efficient Hybrid Sparse Attention with CPU-GPU Parallelism for Long-Context Inference