| 姓名 | 职位 | 研究方向 | 学术成就 & 荣誉 | 核心论文汇总(论文名、何处发表、获奖情况) |
|---|---|---|---|---|
| 杨耀东 | 灵初联创、灵初-北大联合实验室首席科学家 北京大学人工智能研究院助理教授;UCL (深度强化学习策源地) 博士 |
强化学习、AI对齐、多智能体学习、具身智能 | - 发表论文100+,谷歌引用10000+ - 获国际最佳论文奖三次、国际学术新星奖两次 - 华人团队首篇多体强化学习Nature子刊 - NeurIPS'22机器人灵巧操作比赛冠军 - ICLR/NeurIPS/AAAI/IJCAI/AAMAS 领域主席 - Neural Network编委 |
- Language Models Resist Alignment: Evidence From Data Compression (ACL 2025 Best Paper) - Efficient and scalable reinforcement learning for large-scale network control (Nature Machine Intelligence) |
| 钟伊凡 | 灵初-北大联合实验室研究员 | 具身智能,灵巧操作,VLA模型,真机强化 | - NeurIPS'22机器人灵巧操作比赛冠军 | - A Survey on Vision-Language-Action Models: An Action Tokenization Perspective - DexGraspVLA: A Vision-Language-Action Framework Towards General Dexterous Grasping |
| 白丰硕 | 灵初-北大联合实验室研究员 | 具身智能,灵巧操作,仿真sim2real,真机强化 | - | - Retrieval Dexterity: Efficient Object Retrieval in Clutters with Dexterous Hand - STAR: Efficient Preference-based Reinforcement Learning via Dual Regularization |
| 陈浩钧 | 灵初-北大联合实验室研究员 | 具身智能,真机强化 | - | - Falcon: Fast Visuomotor Policies via Partial Denoising (ICML 2025) |
| 祝可菲 | 灵初-北大联合实验室研究员 | 具身智能,灵巧操作 | - 提出首个灵巧操作数据飞轮DexFlyWheel,入选NeurIPS'2025 spotlight | - DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation |
| 张柏荣 | 灵初-北大联合实验室研究员 | 具身智能,AI安全,VLA模型 | - 首个VLA具身大模型安全建模框架SafeVLA,入选NeurIPS'2025 spotlight | - SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning,NeurIPS2025,spotlight |
| 李宇 | 灵初-北大联合实验室研究员 | 灵巧操作,真机强化 | - | - ExDex: Dexterous Non-Prehensile Manipulation for Ungraspable Object via Extrinsic Dexterity |
| 梁一韬 | 灵初-北大联合实验室首席科学家 北京大学人工智能研究院助理教授,博导,博雅青年学者,UCLA博士 |
智能体、强化学习、VLA模型 | - 多次获顶会最佳论文提名和奖项 - 担任多个顶会领域主席和编委 |
- Describe, explain, plan and select: interactive planning with LLMs enables open-world multi-task agents(ICLR2024) - Groot: Learning to follow instructions by watching gameplay videos - State of the art control of atari games using shallow reinforcement learning |
| 陈源培 | 灵初联创 & 算法负责人 | 灵巧操作,强化学习,模仿学习,虚实迁移 | - 发表AI期刊顶会论文10余篇,包括TPAMI,AIJ,CoRL,NIPS,ICRA等 - 研究方向聚焦在机器人灵巧操作、强化学习和虚实迁移上,沿着灵巧手强化学习Sim2Real路径发表论文多篇 - 曾搭建了全球首个双手灵巧操作仿真Benchmark,发表全球首个双灵巧手强化学习Sim2Real工作 - 目前国内灵巧手操作领域强化学习Sim2Real方向成果最多的学者,灵巧手Sim2Real复杂应用落地国内第一人 |
- Towards Human-Level Bimanual Dexterous Manipulation with Reinforcement Learning (TPAMI) - Dynamic Handover: Throw and Catch with Bimanual Hands (CoRL 2023) - Sequential Dexterity: Chaining Dexterous Policies for Long-Horizon Manipulation (CoRL 2023) - Object-Centric Dexterous Manipulation from Human Motion Data (CoRL 2024) |
| 温颖 | 灵初RL leader 上海交通大学人工智能学院长聘教轨副教授 上海创智学院全时导师 博导,UCL博士 |
强化学习,多智能体系统,决策大模型 | - 发表ICML, NeurIPS, ICLR, ICRA, CoRL, AAMAS等顶会论文50余篇 - 两次获得顶会最佳论文奖项 - AAAI/ECAI/AAMAS等顶会领域主席 - 22年推出了国内首个多模态决策大模型DB1(VLA前身) |
- Reinforcing Language Agents via Policy Optimization with Action Decomposition (NeurIPS 2024) - Alphazero-like tree-search can guide large language model decoding and training (ICML 2024) - Rema: Learning to meta-think for llms with multi-agent reinforcement learning (NeurIPS 2025) |
Psi-Robot/Docs-Center
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|