ReCogDrive combines VLM and diffusion planning to generate smooth, safe trajectories, achieving SOTA on NAVSIM with PDMS 90.8.

Yongkang Li, Kaixin Xiong, Xiangyu Guo et al.

Proposes VLM4TS with ViT4TS for zero-shot time series anomaly detection, achieving 24.6% F1 improvement and 36x token efficiency.

Zelin He, Sarah Alnegheimish, Matthew Reimherr

Proposes a real-time large-scale scene reconstruction method combining fast initial pose estimation and direct Gaussian primitive sampling, completing scene modeling within 30 minutes with PSNR 21.7dB.

Andreas Meuleman, Ishaan Shah, Alexandre Lanvin et al.

Proposes 3D-aware pseudo-label learning for semantic correspondence, achieving over 4% improvement on SPair-71k.

Olaf Dünkel, Thomas Wimmer, Christian Theobalt et al.

PAM combines SAM 2 and LLMs for efficient, multi-task regional understanding in images and videos, achieving 1.2-2.4× faster performance.

Weifeng Lin, Xinyu Wei, Ruichuan An et al.

Follow-Your-Creation employs a video inpainting framework to generate and edit 4D content from monocular videos, achieving multi-view consistency and flexible editing.

Yue Ma, Kunyu Feng, Xinhua Zhang et al.

Struct2D enables 3D spatial reasoning using perception-guided structured 2D inputs, achieving high performance without explicit 3D data.

Fangrui Zhu, Hanhui Wang, Yiming Xie et al.

Proposes Contour Errors (CE), an ego-centric matching metric based on Hausdorff distance, improving 3D MOT evaluation over IoU and CPD.

Sharang Kaul, Simon Bultmann, Mario Berk et al.

Rex-Thinker uses Chain-of-Thought reasoning for object referring, enhancing precision and interpretability.

Qing Jiang, Xingyu Chen, Zhaoyang Zeng et al.

Temporal-RLT framework enhances video understanding efficiency through reward design and data selection, significantly reducing training data.

Hongyu Li, Songhao Han, Yue Liao et al.

ShapeLLM-Omni employs 3D VQVAE and autoregressive transformers to enable unified 3D understanding and generation, supported by the 3D-Alpaca dataset.

Junliang Ye, Zhengyi Wang, Ruowen Zhao et al.

UMA introduces multi-level surface alignment guided by 2D point tracking, achieving 30% reduction in geometric error and 20% detail enhancement at 4K resolution.

Heming Zhu, Guoxing Sun, Christian Theobalt et al.

ArtiScene leverages image intermediary, enabling training-free, text-driven 3D scene generation with high aesthetic quality.

Zeqi Gu, Yin Cui, Zhaoshuo Li et al.

TW-GRPO framework enhances video reasoning with focused thinking, achieving 50.4% accuracy on CLEVRER.

Jisheng Dang, Jingze Wu, Teng Wang et al.

Proposes RT-X Net, a cross-attention transformer that fuses RGB and thermal images, achieving superior low-light enhancement with PSNR of 27.75dB.

Raman Jha, Adithya Lenka, Mani Ramanagopal et al.

Proposes VG LLM, leveraging a pre-trained 3D geometry encoder to learn spatial priors from videos, significantly improving 3D scene understanding and reasoning.

Duo Zheng, Shijia Huang, Yanyang Li et al.

ScaleLong evaluates long video understanding across timescales, revealing a U-shaped performance curve.

David Ma, Huaqing Yuan, Xingjian Wang et al.

Proposed GenCAD-Self-Repairing enhances CAD generation feasibility from 10% to 97% using diffusion guidance and self-repair, fixing 65% of infeasible designs.

Chikaha Tsuji, Enrique Flores Medina, Harshit Gupta et al.

LODGE introduces a hierarchical LOD framework for large-scale 3D Gaussian Splatting, reducing GPU memory by 50% and increasing rendering speed to 200 FPS.

Jonas Kulhanek, Marie-Julie Rakotosaona, Fabian Manhardt et al.

VScan accelerates LVLM inference by 2.91× with only 4.6% performance loss, using a two-stage token pruning strategy.

Ce Zhang, Kaixin Ma, Tianqing Fang et al.