10 papers
MixFlow Training: Alleviating Exposure Bias with Slowed Interpolation Mixture
Hui Li, Fu-Yun Wang, Haoyuan Xia +4
This paper studies the training-testing discrepancy (a.k.a. exposure bias) problem for improving the diffusion models. During training, the input of a prediction network at one tra…
Prompt Reinjection: Alleviating Prompt Forgetting in Multimodal Diffusion Transformers for Text-to-Image Generation
Yuxuan Yao, Yuxuan Chen, Hui Li +6
Multimodal Diffusion Transformers (MMDiTs) for text-to-image generation maintain separate text and image branches, with bidirectional information flow between text tokens and visua…
SlotMemory: Object-Centric KV Memory for Streaming Long-Video Generation
Weijia Dou, Hui Li, Jiahao Cui +3
Streaming video generation models typically rely on temporal-centric memory, which organizes historical context as raw frames, chunk segments, or unclustered tokens. This organizat…
The Thinking Pixel: Recursive Sparse Reasoning in Multimodal Diffusion Latents
Yuwei Sun, Yuxuan Yao, Hui Li +1
Diffusion models have achieved success in high-fidelity data synthesis, yet their capacity for more complex, structured reasoning like text following tasks remains constrained. Whi…
Head-wise Adaptive Rotary Positional Encoding for Fine-Grained Image Generation
Jiaye Li, Baoyou Chen, Hui Li +3
Transformers rely on explicit positional encoding to model structure in data. While Rotary Position Embedding (RoPE) excels in 1D domains, its application to image generation revea…
Pyramidal Patchification Flow for Visual Generation
Hui Li, Baoyou Chen, Liwei Zhang +3
Diffusion transformers (DiTs) adopt Patchify, mapping patch representations to token representations through linear projections, to adjust the number of tokens input to DiT blocks…