Proposes AdaSlot, a dynamic slot attention framework that adjusts object slot numbers based on scene complexity, improving unsupervised object discovery.

Ke Fan, Zechen Bai, Tianjun Xiao et al.

EgoExo-Fitness dataset combines synchronized egocentric and exocentric videos, enabling multi-task full-body action understanding with rich annotations.

Yuan-Ming Li, Wei-Jin Huang, An-Lan Wang et al.

Real2Code employs image segmentation and fine-tuned LLM code generation to reconstruct complex articulated objects with up to 10 joints, outperforming state-of-the-art methods.

Zhao Mandi, Yijia Weng, Dominik Bauer et al.

Flash-VStream enables real-time long video stream understanding, significantly reducing inference latency and VRAM usage.

Haoji Zhang, Yiqin Wang, Yansong Tang et al.

LVBench evaluates long video understanding, focusing on long-term memory and reasoning, with a dataset averaging 4101 seconds, covering six core tasks.

Weihan Wang, Zehai He, Wenyi Hong et al.

Hydra-MDP employs multi-teacher knowledge distillation with multi-head decoding for end-to-end multimodal planning, achieving first place in Navsim with significant generalization improvements.

Zhenxin Li, Kailin Li, Shihao Wang et al.

LlamaGen applies large language model's next-token prediction to image generation, achieving 2.18 FID on ImageNet with a 3.1B parameter model, surpassing diffusion models.

Peize Sun, Yi Jiang, Shoufa Chen et al.

FPN-IAIA-BL model improves mass margin classification accuracy in mammography, achieving AUROC of 0.88.

Julia Yang, Alina Jade Barnett, Jon Donnelly et al.

Proposes a four-stage diffusion-based data augmentation framework for Earth Observation images, significantly enhancing semantic diversity and model accuracy.

Tiago Sousa, Benoît Ries, Nicolas Guelfi

Proposes ShareGPT4Video with differential captioning, achieving SOTA on three benchmarks, using 40K GPT-4V annotated videos for enhanced video understanding and generation.

Lin Chen, Xilin Wei, Jinsong Li et al.

MLVU benchmark evaluates 23 multimodal models on long videos (up to 2 hours) across 9 tasks, revealing significant room for improvement in long video understanding.

Junjie Zhou, Yan Shu, Bo Zhao et al.

Proposes self-guidance using weaker models to enhance diffusion model image quality, achieving a record FID of 1.25 on ImageNet-512.

Tero Karras, Miika Aittala, Tuomas Kynkäänniemi et al.

SpatialRGPT enhances 3D spatial reasoning in VLMs via scene graph and depth integration, achieving 91.78% success on SpatialRGBT-Bench.

An-Chieh Cheng, Hongxu Yin, Yang Fu et al.

This paper introduces a novel evaluation framework for aerial Visual Place Recognition (VPR), emphasizing multi-scale map construction and overlap optimization.

Ivan Moskalenko, Anastasiia Kornilova, Gonzalo Ferrer

Video-MME benchmarks multi-modal video understanding with 900 videos across 6 domains, achieving 75% accuracy with Gemini 1.5 Pro.

Chaoyou Fu, Yuhan Dai, Yongdong Luo et al.

Proposes CV-VAE, a video VAE with a latent space compatible with Stable Diffusion's image VAE, enabling fourfold increase in frame generation.

Sijie Zhao, Yong Zhang, Xiaodong Cun et al.

EMAG method improves hand forecasting accuracy by considering ego-motion, achieving 1.7% and 7.0% improvements on Ego4D and EPIC-Kitchens 55 datasets respectively.

Masashi Hatano, Ryo Hachiuma, Hideo Saito

SparseDrive achieves end-to-end autonomous driving via sparse scene representation, reducing collision rate by 71.4%.

Wenchao Sun, Xuewu Lin, Yining Shi et al.

VideoTree constructs adaptive hierarchical video representations, achieving 61.1% accuracy for long video reasoning without training.

Ziyang Wang, Shoubin Yu, Elias Stengel-Eskin et al.

Proposes ICTM, an efficient MAP approximation for high-dimensional linear inverse problems using flow models, outperforming existing flow matching methods.

Yasi Zhang, Peiyu Yu, Yaxuan Zhu et al.