Hunyuan3D 2.1 generates high-fidelity 3D assets using Hunyuan3D-DiT and Hunyuan3D-Paint, enhancing geometric detail and material quality.

Team Hunyuan3D, Shuhui Yang, Mingxin Yang et al.

Proposes PeRL, a permutation-enhanced RL framework, achieving state-of-the-art on multi-image reasoning benchmarks with significant margin.

Yizhen Zhang, Yang Ding, Shuoshuo Zhang et al.

Proposes CSVC, a causal prompt optimization framework guiding diffusion models to generate causally consistent counterfactual videos, improving effectiveness and quality.

Nikos Spyrou, Athanasios Vlontzos, Paraskevas Pegios et al.

Integrating LLMs like GPT-3 with vision models enhances traffic scene segmentation, achieving 85% mIoU and 78% zero-shot accuracy on Cityscapes and BDD100K.

Sanjeda Akter, Ibne Farabi Shihab, Anuj Sharma

Vid-CamEdit enables video camera trajectory editing via generative rendering and geometry estimation, enhancing novel view video synthesis quality.

Junyoung Seo, Jisang Han, Jaewoo Jung et al.

TR2M leverages image and text to predict pixel-wise scale maps, converting relative to absolute depth with high cross-domain generalization.

Beilei Cui, Yiming Huang, Long Bai et al.

Proposes a triplane-based multi-camera tokenization, reducing tokens by 72% and doubling inference speed, for end-to-end autonomous driving.

Boris Ivanovic, Cristiano Saltori, Yurong You et al.

DAVID-XR1 integrates fine-grained defect annotations and chain-of-thought reasoning, enabling interpretable AI-generated video detection with strong cross-generator generalization.

Yifeng Gao, Yifan Ding, Hongyu Su et al.

Proposed occlusion-aware bilayer mask decoder with SAM integration achieves 94.4% mAP on prohibited item segmentation in X-ray images.

Yunhan Ren, Ruihuang Li, Lingbo Liu et al.

QuadricFormer leverages superquadrics for efficient 3D occupancy prediction, achieving 21.11 mIoU with fewer primitives and lower computational cost.

Sicheng Zuo, Wenzhao Zheng, Xiaoyong Han et al.

VINCIE model learns in-context image editing from videos, achieving state-of-the-art results on multi-turn editing benchmarks.

Leigang Qu, Feng Cheng, Ziyan Yang et al.

CreatiPoster employs a multi-layer protocol model to generate editable graphic compositions, with background synthesis ensuring visual harmony and flexibility.

Dexiang Hong, Zhao Zhang, Weidong Chen et al.

VRBench evaluates multi-step reasoning in long videos, with 960 videos and 8,243 QA pairs.

Jiashuo Yu, Yue Wu, Meng Chu et al.

VideoExplorer employs dynamic reasoning and temporal grounding to outperform baselines in long-video understanding, achieving 54.3% accuracy.

Huaying Yuan, Zheng Liu, Junjie Zhou et al.

Rectified flow turns 3D hand trajectories into interaction sounds that are often indistinguishable from real audio.

Yiming Dou, Wonseok Oh, Yuqing Luo et al.

Proposes dual volume packing for efficient part-level 3D generation from a single image, avoiding segmentation priors.

Jiaxiang Tang, Ruijie Lu, Zhaoshuo Li et al.

Proposes 4D Gaussian Transformer (4DGT) for real-time dynamic scene reconstruction from monocular videos, significantly improving speed and accuracy.

Zhen Xu, Zhengqin Li, Zhao Dong et al.

Self Forcing introduces autoregressive training with distribution matching, enabling real-time video generation with high quality and low latency, outperforming traditional slow diffusion models.

Xun Huang, Zhengqi Li, Guande He et al.

ZeroVO achieves zero-shot generalization across environments, improving performance by over 30%.

Lei Lai, Zekai Yin, Eshed Ohn-Bar

SpatialLM integrates point cloud encoding with large language models, achieving state-of-the-art indoor scene layout estimation and 3D detection.

Yongsen Mao, Junhao Zhong, Chuan Fang et al.