Using sparse 3D hand joints with occlusion-aware features, this method enhances egocentric video generation fidelity and control.

Chenyangguang Zhang, Botao Ye, Boqi Chen et al.

Shape-of-You (SoY) employs fused Gromov-Wasserstein optimal transport with 3D geometric priors, achieving state-of-the-art semantic correspondence without explicit annotations.

Jiin Im, Sisung Liu, Je Hyeong Hong

WeEdit employs HTML-generated 330K training pairs, glyph-guided fine-tuning, and multi-objective reinforcement learning for multi-language text editing.

Hui Zhang, Juntao Liu, Zongkai Liu et al.

ShotVerse employs Vision-Language Model-based planning and diffusion control to generate cinematic multi-shot videos from text, aligning (caption, trajectory, video) triplets for high fidelity.

Songlin Yang, Zhe Wang, Xuyi Yang et al.

AssistMimic employs multi-agent reinforcement learning to achieve 83% success in physics-grounded human-human assistive motion imitation.

Yuto Shibata, Kashu Yamazaki, Lalit Jayanti et al.

COMIC system uses LLM critics to generate sketch comedy videos near professional quality.

Susung Hong, Brian Curless, Ira Kemelmacher-Shlizerman et al.

V2M-Zero generates time-aligned music from video using event curves, achieving significant improvements in audio quality and beat alignment across datasets.

Yan-Bo Lin, Jonah Casebeer, Long Mai et al.

DynVLA uses Dynamics CoT to predict compact world dynamics, excelling on datasets like NAVSIM.

Shuyao Shang, Bing Zhan, Yunfei Yan et al.

Proposed BiLaLoRA with CLIP-based H2C loss achieves superior real-world image dehazing with efficient parameter tuning.

Yan Zhang, Long Ma, Yuxin Feng et al.

Proposes World2Act, a latent-space post-training framework that improves success rates by +2.5% using contrastive alignment without pixel supervision.

An Dinh Vuong, Tuan Van Vo, Abdullah Sohail et al.

InternVL-U is a 4B-parameter multimodal model achieving state-of-the-art understanding, reasoning, generation, and editing, outperforming larger models.

Changyao Tian, Danni Yang, Guanzhou Chen et al.

OddGridBench reveals MLLMs' deficiency in visual discrepancy detection, proposing OddGrid-GRPO to enhance performance.

Tengjin Weng, Wenhao Jiang, Jingyi Wang et al.

RoomTour3D-IGR learns implicit geometry from web videos, improving NaviLLM by over 6% across four VLN benchmarks.

Mingfei Han, Haihong Hao, Liang Ma et al.

WS-Net addresses weak signal collapse via state-space and weak signal attention fusion, achieving up to 55% RMSE and 63% SAD reductions.

Zekun Long, Ali Zia, Guanyiman Fu et al.

HDR-NSFF introduces 4D spatio-temporal neural scene flow fields for high-quality dynamic HDR scene reconstruction, outperforming 2D fusion methods.

Shin Dong-Yeon, Kim Jun-Seong, Kwon Byung-Ki et al.

QualiTeacher enhances image restoration by conditioning on pseudo-label quality, significantly improving model generalization.

Fengyang Xiao, Jingjia Feng, Peng Hu et al.

Proposed Masked Motion Diffusion Model (MMDM) with Kinematic Attention Aggregation for robust motion reconstruction.

Junkun Jiang, Jie Chen, Ho Yin Au et al.

UniLongGen enhances long-sequence image generation stability via dynamic memory management.

Haoyu Chen, Qing Liu, Yuqian Zhou et al.

UnSCAR employs a multi-expert MoE architecture, supporting over 16 degradation types, enhancing scalability and controllability.

Debabrata Mandal, Soumitri Chattopadhyay, Yujie Wang et al.

LiveWorld addresses 'out-of-sight dynamics' by introducing a monitor-based mechanism for persistent world evolution.

Zicheng Duan, Jiatong Xia, Zeyu Zhang et al.