Haotian Ye

InfoTok: Adaptive Discrete Video Tokenizer via Information-Theoretic Compression

DiffusionNFT: Online Diffusion Reinforcement with Forward Process

Can Language Models Discover Scaling Laws?

Inference-time scaling of diffusion models through classical search

reAR: Rethinking Visual Autoregressive Models via Token-wise Consistency Regularization

Fractional Reasoning via Latent Steering Vectors Improves Inference Time Compute

Discrete Diffusion Trajectory Alignment via Stepwise Decomposition

NFT: Bridging Supervised Learning and Reinforcement Learning in Math Reasoning

A Snapshot of Influence: A Local Data Attribution Framework for Online Reinforcement Learning

Reducing Hallucinations in Large Vision-Language Models via Latent Space Steering

Weak-for-Strong: Training Weak Meta-Agent to Harness Strong Executors

TFG-Flow: Training-free Guidance in Multimodal Generative Flow