arXiv subjects

Wanyuan Pang

Publications and source records attributed to Wanyuan Pang.

3 recordsLinked to original sources

arXiv2026

TempAct: Advancing Temporal Plausibility in Autoregressive Video Generation via Planner-Executor RL

Autoregressive (AR) video diffusion models enable low-latency streaming generation by synthesizing videos chunk by chunk with cached visual context, but this chunk-wise formulation makes temporal instruction following ambiguous. A single global prompt does not specify which sub-event should be realized in each chunk, while naively switching to step-wise prompts often leads to delayed reactions, blended step semantics, and error propagation across prompt transitions. These failures are difficult to address with supervised fine-tuning or distillation alone: SFT suffers from exposure bias, while rollout-based distillation still optimizes low-level denoising or teacher-distribution matching rather than directly enforcing action ordering and prompt-transition correctness. We address these challenges with TempAct, a planner--executor reinforcement learning framework that jointly optimizes temporal decomposition and step-conditioned execution for temporally plausible AR video generation. TempAct uses an LLM planner to explore span-aware step prompts that are executable by the video model, and trains an AR diffusion executor to follow these prompts under its own generated histories. Its key mechanism is hierarchical group exploration: candidate plans form planning groups, and each plan induces an execution group of multiple continuations from a shared visual context, enabling plan-level credit assignment for long-horizon temporal outcomes and executor-level credit assignment for prompt-switch behavior. We further design hierarchical rewards that combine plan-quality and full-video temporal feedback for the planner with local transition-level step-following rewards, aesthetic regularization, and KL constraints for the executor. Experiments on Self-Forcing and LongLive show that TempAct improves temporal consistency while preserving overall visual quality.

cs.CV↗

arXiv2025

GRPO-Guard: Mitigating Implicit Over-Optimization in Flow Matching via Regulated Clipping

Recently, GRPO-based reinforcement learning has shown remarkable progress in optimizing flow-matching models, effectively improving their alignment with task-specific rewards. Within these frameworks, the policy update relies on importance-ratio clipping to constrain overconfident positive and negative gradients. However, in practice, we observe a systematic shift in the importance-ratio distribution-its mean falls below 1 and its variance differs substantially across timesteps. This left-shifted and inconsistent distribution prevents positive-advantage samples from entering the clipped region, causing the mechanism to fail in constraining overconfident positive updates. As a result, the policy model inevitably enters an implicit over-optimization stage-while the proxy reward continues to increase, essential metrics such as image quality and text-prompt alignment deteriorate sharply, ultimately making the learned policy impractical for real-world use. To address this issue, we introduce GRPO-Guard, a simple yet effective enhancement to existing GRPO frameworks. Our method incorporates ratio normalization, which restores a balanced and step-consistent importance ratio, ensuring that PPO clipping properly constrains harmful updates across denoising timesteps. In addition, a gradient reweighting strategy equalizes policy gradients over noise conditions, preventing excessive updates from particular timestep regions. Together, these designs act as a regulated clipping mechanism, stabilizing optimization and substantially mitigating implicit over-optimization without relying on heavy KL regularization. Extensive experiments on multiple diffusion backbones (e.g., SD3.5M, Flux.1-dev) and diverse proxy tasks demonstrate that GRPO-Guard significantly reduces over-optimization while maintaining or even improving generation quality.

cs.CV↗

arXiv2025

VOILA: Complexity-Aware Universal Segmentation of CT images by Voxel Interacting with Language

Satisfactory progress has been achieved recently in universal segmentation of CT images. Following the success of vision-language methods, there is a growing trend towards utilizing text prompts and contrastive learning to develop universal segmentation models. However, there exists a significant imbalance in information density between 3D images and text prompts. Moreover, the standard fully connected layer segmentation approach faces significant challenges in handling multiple classes and exhibits poor generalizability. To address these challenges, we propose the VOxel Interacting with LAnguage method (VOILA) for universal CT image segmentation. Initially, we align voxels and language into a shared representation space and classify voxels on the basis of cosine similarity. Subsequently, we develop the Voxel-Language Interaction framework to mitigate the impact of class imbalance caused by foreground-background discrepancies and variations in target volumes. Furthermore, a Complexity-Aware Sampling method is proposed to focus on region hard to segment, achieved by generating pseudo-heatmaps from a trainable Gaussian mixture distribution. Our results indicate the proposed VOILA is capable to achieve improved performance with reduced parameters and computational cost during training. Furthermore, it demonstrates significant generalizability across diverse datasets without additional fine-tuning.

cs.CV↗