PhyGDPO enhances text-to-video generation's physical consistency via physics-guided preference optimization, outperforming existing methods.

Yuanhao Cai, Kunpeng Li, Menglin Jia et al.

Proposes a rubric-guided reinforcement learning framework to enhance AI research plan generation, achieving 70% expert preference and 12-22% improvements across domains.

Shashwat Goel, Rishi Hazra, Dulhan Jayalath et al.

Proposes CreativeDC, a two-stage prompting framework, boosting Python task diversity by 1.6× with structured divergent-convergent reasoning.

Manh Hung Nguyen, Sebastian Tschiatschek, Adish Singla

HY-Motion 1.0 scales DiT-based flow matching to 1B+ parameters, enabling high-fidelity text-to-3D human motion generation.

Yuxin Wen, Qing Shuai, Di Kang et al.

RealX3D benchmark shows significant performance drop in multi-view visual restoration under physical degradations.

Shuhong Liu, Chenyu Bao, Ziteng Cui et al.

CME-CAD introduces a multi-expert reinforcement learning framework for high-precision CAD code generation, trained on 17,300+ industrial-grade instances.

Ke Niu, Haiyang Yu, Zhuofan Chen et al.

MFT method surpasses LoRA in VLMs without updating weights, enhancing performance.

Mingyuan Zhang, Yue Bai, Yifan Wang et al.

AutoForge automates environment synthesis and ERPO enhances multi-environment RL stability, achieving 73.1 in tau-bench.

Shihao Cai, Runnan Fang, Jialong Wu et al.

DA360 leverages ViT-based shift learning for scale-invariant panoramic depth estimation, producing high-quality point clouds with seamless boundaries.

Hualie Jiang, Ziyang Song, Zhiqiang Lou et al.

Proposed a monocular depth estimation method using visual autoregressive (VAR) priors, achieving state-of-the-art indoor performance with only 74K samples.

Amir El-Ghoussani, André Kaup, Nassir Navab et al.

Dream-VL and Dream-VLA leverage diffusion language models for breakthroughs in vision-language and vision-language-action tasks.

Jiacheng Ye, Shansan Gong, Jiahui Gao et al.

EnFlow combines flow-based conformer generation with explicit energy modeling, enabling low-energy molecular structure discovery with 1-2 ODE steps.

Guikun Xu, Xiaohan Yi, Ziqiao Meng et al.

CoAgent employs a plan-synthesize-verify-edit loop with explicit entity memory, achieving significant improvements in long video coherence and visual consistency, with Subject Consistency at 94.70%.

Qinglin Zeng, Kaitong Cai, Ruiqi Chen et al.

Multi-source diverse pretraining enables human-to-robot skill transfer via vision-language-action models, nearly doubling generalization performance.

Simar Kareer, Karl Pertsch, James Darpinian et al.

OxygenREC employs a Fast-Slow thinking architecture integrating world knowledge, enhancing deep reasoning and multi-scenario recommendation efficiency.

Xuegang Hao, Ming Zhang, Alex Li et al.

VULCAN uses tool-augmented multi-agent collaboration with recursive backtracking to achieve robust multi-step 3D object arrangement, outperforming baselines.

Zhengfei Kuang, Rui Lin, Long Zhao et al.

Yume-1.5 employs joint spatiotemporal-channel modeling (TSCM) to enable real-time long video generation with improved speed and stability, supporting interactive world creation.

Xiaofeng Mao, Zhen Li, Chuanhao Li et al.

Proposes CAT, a structured context management framework achieving 57.6% success on SWE tasks, significantly outperforming baseline methods.

Shukai Liu, Jian Yang, Bo Jiang et al.

Introduces Time-integrated Optimal Transport (TiOT) framework, enhancing time series comparison accuracy and stability.

Thai P. D. Nguyen, Hong T. M. Chu, Kim-Chuan Toh

Proposes multiple tokenization penalty functions to quantify the impact of word splitting on LLM performance; experiments show significant correlation.

Sachin Pawar, Manoj Apte, Kshitij Jadhav et al.