The 'Agentic Self-Improvement' framework employs a two-stage goal-driven optimization, significantly improving semantic adherence and control in image-to-video generation, outperforming unguided methods with up to 69% preference.

Aman Tyagi, Hemanth Boinpally, Jonathan Chen et al.

HAMP-LIC introduces Hessian-based mixed-precision post-training quantization, achieving up to 4.85× compression with only 0.59% BD-rate loss.

Yuefeng Zhang

GAS framework uses decoupled Transformer architecture with NEP for visual understanding, achieving zero inference overhead and significant improvements in spatial perception.

Zhongbin Guo, Jiahao Xie, Dongling Xiao et al.

Meta-detection with verifiable temporal grounding and reward redistribution boosts AI-generated video detection F1 to 75.39%, outperforming existing methods.

Bowei Liu, Zheng Lu, Yuhan Bian et al.

Proposes a reflection-guided self-distillation framework for test-time GUI grounding, achieving a 7.4% accuracy boost without human annotations.

Shiyu Xuan, Zechao Li

Proposes EGVOR with evidence chain and hierarchical diagnosis, improving robustness by 15% in complex urban scenes.

Zhaoyang Wei, Bowen Jiang, Xumeng Han et al.

MVTrack uses H.264 motion vectors for ultrafast, appearance-free moving object tracking; achieves high accuracy with minimal parameters and FLOPs.

Iñaki Erregue, Kamal Nasrollahi, Sergio Escalera

VGGD leverages pretrained geometric priors with dual-path architecture to enhance single-frame surround-view 3D reconstruction, achieving state-of-the-art results.

Junhong Lin, Jinlong Wang, Xianda Guo et al.

Gaussian Sculpting employs end-to-end field optimization with bi-level training for high-quality surface reconstruction, achieving PSNR of 32.5dB on object scenes.

Ke Jiaxin, Juncheng Liu, Yi Wang et al.

AlignBlock module significantly enhances ViT data efficiency, boosting terminal scene mAP@50 to 0.973.

Haoran Sui, Yaoyuan Jia

PCFlow achieves efficient image restoration through flow matching, significantly improving visual quality.

Sangwoo Jo, Donggeun Ko, Jayeon Kang et al.

This study reveals that frozen brain MRI foundation models encode site information as a linearly decodable fingerprint, surpassing clinical variables, rooted in low-level image statistics.

Saman Rahbar

Introduces the Space-Creation Index (SCI) combined with event-based junk possession index to evaluate spatial utilization quality in football, distinguishing dead possession from space creation.

Seongjin Choi

MSP-Net uses manifold-guided spectral prompts to achieve AUC > 0.80 and Precision > 0.96 on HOT2020 and HOT2023 datasets.

Juliu Li, Hanlin Qin, Shuowen Yang et al.

RecoverFly introduces a failure-aware reinforcement learning post-training framework, boosting success rate by 3.12-8.37% on UAV VLN tasks.

Boxiong Wang, Hui Kang, Geng Sun et al.

Proposed view-adaptive neural renderer improves 3D reconstruction from a single image by correcting view-dependent errors using self-attention, achieving near state-of-the-art results.

U-Chae Jun, Jaeeun Ko, Jiwoo Kang

AeroReformer2 employs a bilateral network with boundary-preserving visual paths and kernel linear cross-modal attention, achieving 62.09% mIoU in spoken-query remote sensing segmentation.

Rui Li, Chenxi Duan, Haoyang Yang

FiRe refines images at a fixed noise level, achieving 3× faster inference and 8× fewer FLOPs while maintaining high-quality counterfactuals.

Yan Zeng, Changlu Guo, Oskar Kristoffersen et al.

RenderMatte fine-tunes FLUX.1 with alpha-edge loss and group-relative alignment, achieving state-of-the-art image matting.

Zecheng Ren, Yafei Hu, Jianing Zhao et al.

SUMI model enhances point cloud detail recovery with diffusion module, improving CD by 16.1%.

Yanlong Li, Kanchana Thilakarathna