SAM2Act integrates large-scale visual foundation models with multi-view transformers, achieving 86.8% success and enhanced spatial memory.

Haoquan Fang, Markus Grotz, Wilbert Pumacay et al.

iRe-VLA alternates frozen-backbone online RL with full-model imitation, raising MetaWorld unseen-task success to 0.80.

Yanjiang Guo, Jianke Zhang, Xiaoyu Chen et al.

CuriousBot uses an actionable 3D relational graph for mobile interaction, achieving 82% average task success and outperforming 2D VLM baselines.

Yixuan Wang, Leonor Fermoselle, Tarik Kelestemur et al.

FAST uses DCT-based compression to encode robot action sequences, enabling efficient training of autoregressive VLAs on high-frequency, dexterous tasks.

Karl Pertsch, Kyle Stachowicz, Brian Ichter et al.

Using attribute learning and data-efficient adaptation, robotic grasping achieves over 81% success rate.

Yang Yang, Houjian Yu, Xibai Lou et al.

VinT-6D dataset integrates vision, touch, and proprioception to enhance robotic manipulation precision.

Zhaoliang Wan, Yonggen Ling, Senlin Yi et al.

Proposes Multimodal Chain-of-Thought Co-Navigation (MCoCoNav) for multi-robot semantic navigation, achieving over 87% success rate and 46% SPL on HM3D.

Zhixuan Shen, Haonan Luo, Kexun Chen et al.

VLABench benchmark evaluates language-conditioned robotic manipulation with 100 long-horizon tasks, testing generalization and reasoning capabilities of VLAs.

Shiduo Zhang, Zhe Xu, Peiju Liu et al.

BODex uses bilevel optimization for efficient dexterous grasp synthesis, achieving over 75% success rate.

Jiayi Chen, Yubin Ke, He Wang

STRAP combines pre-trained vision models and subsequence dynamic time warping for sub-trajectory retrieval, significantly enhancing robot policy generalization and robustness.

Marius Memmel, Jacob Berg, Bingqing Chen et al.

Introduces Policy Decorator, a model-agnostic online residual policy framework, boosting large imitation policies’ success rates by 20-35% in robotic tasks.

Xiu Yuan, Tongzhou Mu, Stone Tao et al.

Exbody2 employs automated data filtering and teacher-student reinforcement learning to enable humanoid robots to perform expressive, dynamic motions with superior tracking accuracy, surpassing prior methods.

Mazeyu Ji, Xuanbin Peng, Fangchen Liu et al.

Proposes SARA shield, using reachability analysis to classify contact types and ensure robot kinetic energy stays below injury thresholds in human environments.

Jakob Thumm, Julian Balletshofer, Leonardo Maglanoc et al.

Score and Distribution Matching transforms diffusion policies into single-step generators, achieving 6× inference speedup with state-of-the-art action quality.

Bofang Jia, Pengxiang Ding, Can Cui et al.

Multi-GraspLLM generates multi-hand semantic-guided grasp poses using multimodal LLM, significantly improving accuracy.

Haosheng Li, Weixin Mao, Weipeng Deng et al.

FLIP employs flow-based generative planning with pixel-level flow and video models, achieving 100% success in long-horizon manipulation tasks.

Chongkai Gao, Haozhuo Zhang, Zhixuan Xu et al.

Introducing Predictive Motion Priors (PMP) based on CVAE for decoupled upper-limb prediction and robust lower-limb control, significantly improving humanoid robot manipulation.

Chenhao Lu, Xuxin Cheng, Jialong Li et al.

RoboTron-Mani integrates multimodal data fusion and 3D perception, surpassing expert models with improved generalization and task performance.

Feng Yan, Fanfan Liu, Liming Zheng et al.

P3-PO leverages human-prescribed semantic points and vision models to enhance robot policy generalization, achieving 43% overall improvement in real-world tasks.

Mara Levy, Siddhant Haldar, Lerrel Pinto et al.

Proposes neural Riemannian PDE solver (NES) for high-dimensional geodesic distance fields, enabling real-time, global path optimization in robotics.

Yiming Li, Jiacheng Qiu, Sylvain Calinon