26 papers
Dreaming in Flow: Bidirectional Self-Refinement for Unified Multimodal Models
Ke Hao, Yuanzhi Liang, Tingxi Chen +5
Unified multimodal models can generate images and interpret them, but their two capabilities are usually optimized with separate supervision. We introduce Bidirectional Self-Refine…
LLMODE: Aligning ODEs with LLMs via Gated Token Injection for Irregular Spatio-Temporal Forecasting
Di Zhang, Jingyang Zhang, Ziqian Wang +4
Large language models (LLMs) have shown promise for spatio-temporal forecasting, but existing approaches often rely on regularly sampled token sequences and struggle with irregular…
Sample-Adaptive Latent Rewards for Uncertainty-Guided Diffusion Post-Training
Rui Li, Yuanzhi Liang, Ke Hao +4
Latent reward models can supervise visual diffusion models without decoding intermediate states into pixel space. This makes alignment with human preferences more efficient. Howeve…
From World Action Models to Embodied Brains: A Roadmap for Open-World Physical Intelligence
Yuanzhi Liang, Xufeng Zhan, Haibin Huang +2
Artificial general intelligence ultimately requires agents that can reason and act in the physical world. Action models, vision-language-action policies, and world models have adva…
VideoWeave: Unlocking Geometric Consistency in Video Generation via Joint Geometry-Video Modeling
Xunzhi Xiang, Zixuan Duan, Yabo Chen +8
Large-scale video diffusion models often fail to preserve 3D structure over time, causing geometric drift and implausible motion under viewpoint changes. Existing methods usually e…
Reinforcing Few-step Generators via Reward-Tilted Distribution Matching
Yushi Huang, Xiangxin Zhou, Ruoyu Wang +3
Recent advances in few-step diffusion distillation have enabled efficient image generation, yet aligning these models with human preferences remains challenging. We propose Reward-…
Full-4D: Generating Full-Scope 4D Scenes from a Single-View Video
Tingxi Chen, Ke Hao, Yabo Chen +6
Generating 4D scenes from a single-view video is inherently ill-posed: a single viewpoint lacks the information needed to recover a complete, dynamic scene with full coverage. Exis…
PhysOmni: Physics-Grounded Multi-Object Scene Generation from a Single Image with Real-Time Interaction
Xin Zhang, Yabo Chen, Yijie Fang +5
Recent generative video models achieve impressive visual quality but remain constrained by limited physical consistency and controllability. Existing video generation methods provi…
Learning to Credit the Right Steps: Objective-aware Process Optimization for Visual Generation
Rui Li, Ke Hao, Yuanzhi Liang +4
Reinforcement learning, particularly Group Relative Policy Optimization (GRPO), has emerged as an effective framework for post-training visual generative models with human preferen…
Reward-Aware Trajectory Shaping for Few-step Visual Generation
Rui Li, Bingyu Li, Yuanzhi Liang +3
Achieving high-fidelity generation in extremely few sampling steps has long been a central goal of generative modeling. Existing approaches largely rely on distillation-based frame…
CRAFT-LoRA: Content-Style Personalization via Rank-Constrained Adaptation and Training-Free Fusion
Yu Li, Yujun Cai, Chi Zhang
Personalized image generation requires effectively balancing content fidelity with stylistic consistency when synthesizing images based on text and reference examples. Low-Rank Ada…
Uncovering Context Reliance in Unstructured Knowledge Editing
Zisheng Zhou, Mengqi Zhang, Shiguang Wu +4
Editing Large language models (LLMs) with real-world, unstructured knowledge is essential for correcting and updating their internal parametric knowledge. In this work, we revisit…
Spatial-Temporal State Propagation Autoregressive Model for 4D Object Generation
Liying Yang, Jialun Liu, Jiakui Hu +5
Generating high-quality 4D objects with spatial-temporal consistency is still formidable. Existing diffusion-based methods often struggle with spatial-temporal inconsistency, as th…
TeleBoost: A Systematic Alignment Framework for High-Fidelity, Controllable, and Robust Video Generation
Yuanzhi Liang, Xuan'er Wu, Yirui Liu +12
Post-training is the decisive step for converting a pretrained video generator into a production-oriented model that is instruction-following, controllable, and robust over long te…
CtrlVDiff: Controllable Video Generation via Unified Multimodal Video Diffusion
Dianbing Xi, Jiepeng Wang, Yuanzhi Liang +8
We tackle the dual challenges of video understanding and controllable video generation within a unified diffusion framework. Our key insights are two-fold: geometry-only cues (e.g.…
Learning What to Trust: Bayesian Prior-Guided Optimization for Visual Generation
Ruiying Liu, Yuanzhi Liang, Haibin Huang +2
Group Relative Policy Optimization (GRPO) has emerged as an effective and lightweight framework for post-training visual generative models. However, its performance is fundamentall…
UniModel: A Visual-Only Framework for Unified Multimodal Understanding and Generation
Chi Zhang, Jiepeng Wang, Youming Wang +5
We present UniModel, a unified generative model that jointly supports visual understanding and visual generation within a single pixel-to-pixel diffusion framework. Our goal is to…
Rethinking Reward Signals in Video GRPO: When Scores Become Targets
Rui Li, Yuanzhi Liang, Ziqi Ni +3
Group Relative Policy Optimization (GRPO) enables stable and preference-oriented updates via group-wise comparisons for post-training video generation. However, GRPO directly optim…
Seeing What Matters: Visual Preference Policy Optimization for Visual Generation
Ziqi Ni, Yuanzhi Liang, Rui Li +4
Reinforcement learning (RL) has become a powerful tool for post-training visual generative models, with Group Relative Policy Optimization (GRPO) increasingly used to align generat…
Uni-Inter: Unifying 3D Human Motion Synthesis Across Diverse Interaction Contexts
Sheng Liu, Yuanzhi Liang, Jiepeng Wang +3
We present Uni-Inter, a unified framework for human motion generation that supports a wide range of interaction scenarios: including human-human, human-object, and human-scene-with…
Information Capacity: Evaluating the Efficiency of Large Language Models via Text Compression
Cheng Yuan, Jiawei Shao, Xuelong Li
Recent years have witnessed the rapid advancements of large language models (LLMs) and their expanding applications, leading to soaring demands for computational resources. The wid…
ScRPO: From Errors to Insights
Lianrui Li, Dakuan Lu, Jiawei Shao +1
We introduce Self-correction Relative Policy Optimization (ScRPO), a novel reinforcement learning framework designed to empower large language models with advanced mathematical rea…
CAS-Spec: Cascade Adaptive Self-Speculative Decoding for On-the-Fly Lossless Inference Acceleration of LLMs
Zhiyuan Ning, Jiawei Shao, Ruge Xu +4
Speculative decoding has become a widely adopted as an effective technique for lossless inference acceleration when deploying large language models (LLMs). While on-the-fly self-sp…
Pipeline Parallelism is All You Need for Optimized Early-Exit Based Self-Speculative Decoding
Ruanjun Li, Ziheng Liu, Yuanming Shi +3
Large language models (LLMs) deliver impressive generation quality, but incur very high inference cost because each output token is generated auto-regressively through all model la…