19 papers
Direct Forcing: Single-Stage Training with Representation-Space Supervision for Few-Step Autoregressive Video Generation
Haojun Xu, Jie Huang, Chuyang Zhao +6
Few-step autoregressive video generation enables low-latency streaming, yet prevailing training pipelines separate causal adaptation from distribution matching. Recent representati…
REEL: Learning to Retrieve, Evaluate, and Evolve Visual Memory for Any Long-Horizon Multi-Shot Video Generation
Hang Xu, Guohui Zhang, Xiaoxiao Ma +7
Memory-based long-video generation aims to combine user control with long-range visual consistency. We organize this process into three functional components: a planner, an adapter…
GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
Qize Yu, Lianrui Fan, Bowen Ping +19
Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidenc…
Poly-OPD: Heterogeneous Multi-Teacher On-Policy Distillation for Capability-Selectable Flow Models
Siming Fu, Haojun Xu, Ruizhe He +9
Leading open text-to-image models often carry complementary strengths: one may lead on preference-aligned aesthetics while another follows compositional instructions more faithfull…
JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion
Yicheng Xiao, Wenxun Dai, Xinran Qin +22
Real-time video editing requires low-latency causal generation with bounded computational resources while preserving source fidelity and long-term temporal consistency. We present…
Any-OPD: Heterogeneous On-Policy Distillation for Flow-Matching Models via Representation-Space Bridging
Siming Fu, Zheming Fu, Ruizhe He +7
On-policy distillation, in which a teacher corrects samples that the student itself generates, presupposes that the two models speak the same language: identical VAE latents, match…
Ultra Flash: Scaling Real-Time Streaming Video Generation to High Resolutions
Xin Lu, Luxury, Jie Huang +27
While recent autoregressive video diffusion models achieve remarkable streaming quality, they remain confined to low resolutions (e.g., 480P), leaving efficient, scalable, real-tim…
HPSv3++: Scaling Reward Models Across the Full Spectrum of Diffusion Model Capabilities
Yijun Liu, Jie Huang, Zeyue Xue +5
Reward models guide text-to-image (T2I) systems toward outputs aligned with human preferences. However, typical reward models such as HPSv3 are trained on pre-annotated data from e…
Echo-Memory: A Controlled Study of Memory in Action World Models
Wayne King, Zeyue Xue, Yuxuan Bian +13
We present \textbf{Echo-Memory}, a controlled study of memory mechanisms in action-conditioned world models. These models generate multi-segment videos from a first frame, text pro…
Echo-Infinity: Learning Evolving Memory for Real-Time Infinite Video Generation
Yuxuan Bian, Zeyue Xue, Songchun Zhang +9
We present Echo Infinity, an autoregressive (AR) framework towards real-time infinite video generation that employs a learnable evolving memory to dynamically filter, abstract, and…
Sharp description of local minima in the loss landscape of high-dimensional two-layer ReLU neural networks
Jie Huang, Bruno Loureiro, Stefano Sarao Mannelli
We study the population loss landscape of two-layer ReLU networks of the form in a realisable teacher-student setting with Gaussian covaria…
Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusion
Peiliang Cai, Evelyn Zhang, Jiacheng Liu +8
Recent advances in autoregressive video diffusion have enabled sequential and streaming video generation. However, long-horizon generation requires increasingly large KV caches, ma…
OmniNFT: Modality-wise Omni Diffusion Reinforcement for Joint Audio-Video Generation
Guohui Zhang, XiaoXiao Ma, Jie Huang +9
Recent advances in joint audio-video generation have been remarkable, yet real-world applications demand strong per-modality fidelity, cross-modal alignment, and fine-grained synch…
A Systematic Post-Train Framework for Video Generation
Zeyue Xue, Siming Fu, Jie Huang +9
While large-scale video diffusion models have demonstrated impressive capabilities in generating high-resolution and semantically rich content, a significant gap remains between th…
RAE-AR: Taming Autoregressive Models with Representation Autoencoders
Hu Yu, Hang Xu, Jie Huang +4
The latent space of generative modeling is long dominated by the VAE encoder. The latents from the pretrained representation encoders (e.g., DINO, SigLIP, MAE) are previously consi…
Astrolabe: Steering Forward-Process Reinforcement Learning for Distilled Autoregressive Video Models
Songchun Zhang, Zeyue Xue, Siming Fu +6
Distilled autoregressive (AR) video models enable efficient streaming generation but frequently misalign with human visual preferences. Existing reinforcement learning (RL) framewo…
OmniForcing: Unleashing Real-time Joint Audio-Visual Generation
Yaofeng Su, Yuming Li, Zeyue Xue +7
Recent joint audio-visual diffusion models achieve remarkable generation quality but suffer from high latency due to their bidirectional attention dependencies, hindering real-time…
INT v.s. FP: A Comprehensive Study of Fine-Grained Low-bit Quantization Formats
Mengzhao Chen, Meng Wu, Hui Jin +10
Modern AI hardware, such as Nvidia's Blackwell architecture, is increasingly embracing low-precision floating-point (FP) formats to handle the pervasive activation outliers in Larg…
Scaling Law for Quantization-Aware Training
Mengzhao Chen, Chaoyi Zhang, Jing Liu +8
Large language models (LLMs) demand substantial computational and memory resources, creating deployment challenges. Quantization-aware training (QAT) addresses these challenges by…