Proposed a global pose control method for generative view synthesis in Normalized Object Coordinate Space, significantly improving image quality and consistency.

Zhibing Li, Amogh Gupta, Behnoosh Parsa et al.

LongEgoRefer benchmark challenges existing models with long-form egocentric video referring expression comprehension.

Shunya Kato, Taiki Miyanishi, Shuhei Kurita et al.

PWM-ArtGen uses diffusion-based joint modeling of visual dynamics and kinematic parameters, outperforming baselines with strong zero-shot generalization.

Wentao Zheng, Ancong Wu

ReQuest improves long-video QA accuracy through uncertainty-driven keyframe selection.

Minkuk Kim, Suyong Yun, Young Tae Kim et al.

LASER regulates visual attention via grounding and sink suppression rewards, effectively mitigating visual forgetting in LVLMs.

Bowen Yuan, Zijian Wang, Yadan Luo et al.

ICDepth uses In-Context Conditioning and SAND-Attention for video depth estimation, achieving 6-13x data efficiency improvement.

Xuanhua He, Jiaxin Xie, Mingzhe Zheng et al.

Proposes PanoGaussian, combining panoramic trajectories and Gaussian distillation for consistent 4D scene synthesis from monocular videos, even in unseen regions.

Yuankun Yang, Yi Wei, Wenyang Zhou et al.

DriveTeach-VLA enhances VLA models using DVD and 2D-TGP, achieving 90.4 PDMS on Navsim.

Yuguang Yang, Canyu Chen, Zhewen Tan et al.

MVFusion-GS integrates motion-variance guided features and Transformer-based temporal attention to enhance dynamic Gaussian scene reconstruction, outperforming SOTA by 1.2dB PSNR.

Jianwei Hu, Tingxuan Huang, Hengyu Zhou et al.

Introduced O3-D dataset using depth ordering tasks, revealing models' shallow utilization of depth cues with 47-56% accuracy, highlighting significant visual-linguistic biases.

Yiqian Liu, Iuliia Kotseruba, John K. Tsotsos

This study evaluates four motion-based AI video detectors, revealing biases and shortcut learning, with frequency-based methods showing better robustness.

Joren Michels, Lode Jorissen, Nick Michiels

DeWorldSG generates robust 3D semantic scene graphs via depth-guided filtering, improving triplet recall by 77.4%.

Seok-Young Kim, Abdelrahman Elskhawy, Taewook Ha et al.

SoftMoR enhances Vision Transformers via recursion, boosting SR-ViT accuracy to 82.48% on ImageNet-1K.

Sang In Lee, Jihun Park

Proposes ISPA, compressing KV cache by 50% via parametric absorption, with near-lossless visual quality.

Xiaomeng Fu, Jia Li, Yiming Hu et al.

CaBM replaces fixed concept sets with natural language captions, enabling leakage-free, interpretable image recognition with competitive accuracy.

Seref Baris Cagliyan, Umut Ozdemir, Merve Tapli et al.

HOMER employs hierarchical memory and agentic reasoning, achieving +10.8 performance on long video benchmarks.

Yixin Ji, Fanghua Ye, Juntao Li et al.

AMVL framework resolves train-inference mismatch via bidirectional calibration, improving BLINK benchmark average score by +10.83.

Shijie Li, Yilin Gao, Siyuan Yang et al.

DriveVer is a lightweight trajectory verifier that enhances planning models on the NAVSIM benchmark.

Chong He, Yuechen Luo, Fang Li et al.

Vitality-guided compression reduces 3D Diffusion Transformer models by 66%, maintaining geometric fidelity while significantly decreasing size.

Jaeah Lee, Hyunjin Kim, Jaewoong Cho et al.

MEPA introduces a scale-aware MoE framework with semantic guidance, achieving faster training and superior image quality (FID 2.32 vs. 4.10).

Nuoyan Zhou, Zhijun Tu, Lei Yu et al.