RoboBrain 2.5: Depth in Sight, Time in Mind.

BAAI RoboBrain Team

Co-First Author, Project Lead, Technical Report 2026

Project | Paper | Code | Checkpoints

TL;DR: RoboBrain 2.5 is a next-generation embodied AI foundation model designed for complex robotic manipulation, driven by two major upgrades: (1) precise 3D spatial reasoning that generates depth-aware, physically-constrained manipulation traces, and (2) dense temporal value estimation that provides step-by-step execution feedback. Together, these advancements enable a more physically grounded and execution-aware intelligence for downstream learning tasks.

RoboBrain 2.0: See Better. Think Harder. Do Smarter.

BAAI RoboBrain Team

Co-First Author, Core Contributor, Technical Report 2025

Project | Paper | Code | Checkpoints

TL;DR: RoboBrain 2.0 is the previously most powerful embodied brain model designed to unify perception, reasoning, and planning for complex physical tasks. Achieving top performance across diverse spatial and temporal benchmarks, it excels in critical real world capabilities including spatial understanding, trajectory forecasting, and multiple agent long horizon planning, marking a significant step toward developing generalist embodied agents.

LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training

LLaVA-OneVision Community Contributors

Core Contributor, Technical Report 2025

Paper | Code | Datasets | Checkpoints

TL;DR: LLaVA-OneVision-1.5 is an open and highly efficient family of Large Multimodal Models that achieves state of the art performance with significantly reduced training costs. Powered by massive curated datasets and an optimized training framework developed under a $16,000 budget, the models consistently outperform competitors like Qwen2.5-VL across numerous benchmarks.

Robo-Dopamine: General Process Reward Modeling for High-Precision Robotic Manipulation

Huajie Tan*, Sixiang Chen*, Yijie Xu*, Zixiao Wang, Yuheng Ji, Cheng Chi, Yaoxu Lyu, Zhongxia Zhao, Xiansheng Chen, Peterson Co, Shaoxuan Xie, Guocai Yao, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang

Co-First Author, Project Lead, CVPR 2026

Project | Paper | Code | Checkpoints

TL;DR: Joy is dopamine’s handiwork—whether in humans or in robotics.

Action-Sketcher: From Reasoning to Action via Visual Sketches for Long-Horizon Robotic Manipulation

Huajie Tan*, Peterson Co*, Yijie Xu*, Shanyu Rong, Yuheng Ji, Cheng Chi, Xiansheng Chen, Qiongyu Zhang, Zhongxia Zhao, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang

Co-First Author, Project Lead, CVPR 2026

Project | Paper | Code

TL;DR: Make intent visible. Make action reliable.

Reason-RFT: Reinforcement Fine-Tuning for Visual Reasoning of Vision Language Models

Huajie Tan*, Yuheng Ji*, Xiaoshuai Hao*, Xiansheng Chen, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang

Co-First Author, NeurIPS 2025

Project | Paper | Code | Datasets | Checkpoints

TL;DR: Reason-RFT is a pioneering two-stage RFT framework that enhances visual reasoning in VLMs, delivering superior performance, data efficiency, and robust generalization under domain shifts across various complex tasks.

RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete

Yuheng Ji*, Huajie Tan*, Jiayu Shi*, Xiaoshuai Hao*, Yuan Zhang, Hengyuan Zhang, Pengwei Wang, Mengdi Zhao, Yao Mu, Pengju An, Xinda Xue, Qinghang Su, Huaihai Lyu, Xiaolong Zheng, Jiaming Liu, Zhongyuan Wang, Shanghang Zhang

Co-First Author, CVPR 2025

Project | Paper | Code | Datasets | Checkpoints

TL;DR: RoboBrain is the first unified embodied brain model, equiped with planning capability, affordance perception and trajectory prediction.

RoboOS-NeXT: A Unified Memory-based Framework for Lifelong, Scalable, and Robust Multi-Robot Collaboration

Huajie Tan*, Cheng Chi*, Xiansheng Chen*, Yuheng Ji*, Zhongxia Zhao, Xiaoshuai Hao, Yaoxu Lyu, Mingyu Cao, Junkai Zhao, Huaihai Lyu, Enshen Zhou, Ning Chen, Yankai Fu, Cheng Peng, Wei Guo, Dong Liang, Zhuo Chen, Mengsi Lyu, Chenrui He, Yulong Ao, Yonghua Lin, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang

Co-First Author, Project Lead, ArXiv 2025

Project | Paper | Code

TL;DR: RoboOS-NeXT is a unified memory-based framework for lifelong, scalable, and robust multi-robot collaboration.