Mingsheng Long

Associate Professor@Tsinghua University·中国·OpenReview

研究方向

Foundation model · World model · Scientific machine learning · Reinforcement learning · Sustainable AI (weather · climate) · Time series and dynamical systems · Deep learning · Transfer learning

Mingsheng Long

Mitigating Noise Shift in Denoising Generative Models with Noise Awareness Guidance

Vid2World: Crafting Video Diffusion Models to Interactive World Models

POET: Partially Observed Earth Transformer with High-Dimensional Position Embedding

Foresight Diffusion: Improving Sampling Consistency in Predictive Diffusion Models

CoRA: Covariate-Aware Adaptation of Time Series Foundation Models

Accuracy Law for the Future of Deep Time Series Forecasting

FlashBias: Fast Computation of Attention with Bias

PhySense: Sensor Placement Optimization for Accurate Physics Sensing

Sundial: A Family of Highly Capable Time Series Foundation Models

RLVR-World: Training World Models with Reinforcement Learning

Domain Guidance: A Simple Transfer Approach for a Pre-trained Diffusion Model

Dynamical Diffusion: Learning Temporal Dynamics with Diffusion Models

Transolver++: An Accurate Neural Solver for PDEs on Million-Scale Geometries

Timer-XL: Long-Context Transformers for Unified Time Series Forecasting

Unisolver: PDE-Conditional Transformers Are Universal PDE Solvers

Trajectory World Models for Heterogeneous Environments

Long-Sequence Recommendation Models Need Decoupled Embeddings

Unisolver: PDE-Conditional Transformers Towards Universal Neural PDE Solvers

Metadata Matters for Time Series: Informative Forecasting with Transformers