Spaces:
Running
Running
| title: README | |
| emoji: 🦀 | |
| colorFrom: purple | |
| colorTo: pink | |
| sdk: static | |
| pinned: false | |
| # Reasoning Kingdom · 推理王国 | |
| **Reasoning Kingdom（推理王国）** is a community research platform affiliated with **Datawhale**, focusing on deep learning, reasoning systems, and the democratization of AI research. | |
| 推理王国是一个挂靠于 **Datawhale** 旗下的社区研究平台，关注深度学习、推理模型、结构化智能与 AI 研究的民主化。 | |
| 我们相信，推理不应只是少数大型机构的封闭能力，而应该成为更多学习者、研究者、开发者能够理解、复现、改造和继续推进的公共技术。 | |
| ## Mission | |
| Our mission is to make reasoning research more accessible, reproducible, and community-driven. | |
| 我们希望让“推理”从少数模型报告里的指标，变成可以被讨论、被实验、被教学、被复现、被质疑、被共同建设的开放研究方向。 | |
| Reasoning Kingdom is built around three core ideas: | |
| 1. **Democratization of Reasoning** | |
| 推理能力不应该只属于大模型公司，也应该属于社区、学生、独立研究者和开源生态。 | |
| 2. **Research as Public Infrastructure** | |
| 研究不是少数人的展示品，而是可以被复用、被继承、被改写的公共基础设施。 | |
| 3. **From Models to Understanding** | |
| 我们不只关心模型是否“答对”，也关心它为什么失败、在哪里坍缩、如何形成结构，以及如何让更多人理解这些机制。 | |
| ## Research Directions | |
| Reasoning Kingdom focuses on topics including but not limited to: | |
| - Deep learning theory and practice | |
| - Reasoning models and reasoning evaluation | |
| - Small model reasoning | |
| - Structural generalization | |
| - Optimization and learning dynamics | |
| - Open-source datasets and benchmarks | |
| - AI for education and research democratization | |
| - Community-driven AI research workflows | |
| 我们关注的问题包括但不限于： | |
| - 深度学习模型的结构与训练机制 | |
| - 推理模型的能力边界与失效模式 | |
| - 小模型推理与低资源推理 | |
| - 推理评测、数据构造与实验复现 | |
| - 学习动力学、优化过程与表示坍缩 | |
| - 面向社区的开源研究工具链 | |
| - 面向学习者的 AI 研究教育 | |
| ## What We Build | |
| This organization may host: | |
| - Research prototypes | |
| - Model checkpoints | |
| - Datasets and benchmark tools | |
| - Educational materials | |
| - Reproducible experiments | |
| - Community research projects | |
| - Open-source implementations related to reasoning and deep learning | |
| 这里会逐步沉淀推理王国相关的模型、数据、实验代码、教学材料与社区研究项目。 | |
| ## Community Spirit | |
| Reasoning Kingdom follows a simple belief: | |
| > Democracy means permission. | |
| > Permission to not have won yet. | |
| > Permission to keep going. | |
| 民主就是允许。 | |
| 允许还没赢，允许继续走。 | |
| We welcome people who care about reasoning, learning, open research, and the future of AI as a public intellectual project. | |
| 这里不要求每个人一开始就拥有完整答案。 | |
| 更重要的是：愿意进入问题，愿意复现，愿意讨论，愿意把一个模糊的问题推进到更清楚一点。 | |
| ## Affiliation | |
| Reasoning Kingdom is a community research platform affiliated with **Datawhale**. | |
| We aim to contribute to the open-source AI community through collaborative research, educational materials, reproducible experiments, and accessible reasoning studies. | |
| ## Contact | |
| For collaboration, research discussion, or community projects, please follow this organization and related repositories. | |
| More projects will be released gradually. |