NLP / LLM
大模型与检索增强，面向中文场景的评测与数据构建。
LLM applications and retrieval augmentation, with evaluation and dataset construction for Chinese scenarios.
CUMT Alethic Insight AI Lab · 真知灼见
Alethic Insight
From noise, structure.
From structure, aletheia.
由繁得构，由构显真。中国矿业大学学生主导的 AI 研究工程共同体。
From noise, structure; from structure, truth. A student-led AI research, engineering, and open-source community at China University of Mining and Technology.
Position
组织定位
Positioning
Alethic Insight，中文名“真知灼见”。AI 研究训练、工程实践、开源协作。
Alethic Insight — “真知灼见” in Chinese. AI research training, engineering practice, and open-source collaboration.
Directions
研究方向
Research Directions
CV
图像与视频理解，从视觉表征到几何与三维重建。
Image and video understanding, from visual representation to geometry and 3D reconstruction.
VLM / MLLM
视觉语言与多模态大模型，跨模态推理的对齐与落地。
Vision-language and multimodal LLMs — alignment and deployment of cross-modal reasoning.
RL
强化学习与对齐训练，在交互环境中打磨策略与反馈回路。
Reinforcement learning and alignment — policies and feedback loops trained in interactive environments.
Agent
工具调用、规划与多智能体编排，面向真实任务的可靠执行。
Tool use, planning, and multi-agent orchestration — reliable execution for real-world tasks.
Memory
长期记忆与检索系统，让智能体沉淀经验、越用越强。
Long-term memory and retrieval systems — agents that accumulate experience and improve with use.
AI Infra
部署、实验追踪与模型服务，GPU 与存储的工程化工作流。
Deployment, experiment tracking, and model serving — engineered workflows for GPUs and storage.
Lab Focus
实验室三大面向
Three Tracks
科研Research
围绕方向议题做公开研究，实验可复现、产物可追溯。
Open research on our directions — reproducible experiments and traceable artifacts.
交付项目Delivered Projects
从招新平台到内部工具，把想法交付成真实可用的系统。
From the recruitment platform to internal tooling — ideas delivered as working systems.
比赛Competitions
以赛促学，组队打榜并沉淀方法与复盘。
Learning by competing — team runs, methods, and post-mortems worth sharing.
Open Work
公开项目
Open Projects
FLUED
Byte-to-latent 语言编码接口研究
Research on a byte-to-latent language encoding interface
website
组织官网源码与部署
Source code and deployment of this site
recruitment
招新平台：报名与机试入口
Recruitment platform: application and coding-test entry
governance-public
公开章程、成员规则和项目治理
Public charter, membership rules, and project governance
awesome-ai
AI 学习与研究工程资源索引
An index of AI learning and research-engineering resources
Hugging Face
模型、数据集、Spaces 和公开实验产物
Models, datasets, Spaces, and public experiment artifacts
Awards
竞赛成果
Competition Results
一等奖
FT2614983D1004
孤独的根号五
第17届中国大学生服务外包创新创业大赛 · 人工智能专项赛道【D01】智能体决策算法（腾讯开悟）
谢圣民（队长）、利楚、程致远、季成洋、罗若涵 · 指导老师：祝汉城
17th China College Student Service Outsourcing Innovation & Entrepreneurship Competition, AI Track D01 "Agent Decision Algorithms" (Tencent Kaiwu)
Xie Shengmin (captain), Li Chu, Cheng Zhiyuan, Ji Chengyang, Luo Ruohan · Advisor: Zhu Hancheng
FT2614983D1004
三等奖
FT2608179D3065
孤独的根号三
第17届中国大学生服务外包创新创业大赛 · 人工智能专项赛道【D01】智能体决策算法（腾讯开悟）
张岩（队长）、韩文艳、李豪、王张嘉、文艺学 · 指导老师：祝汉城
17th China College Student Service Outsourcing Innovation & Entrepreneurship Competition, AI Track D01 "Agent Decision Algorithms" (Tencent Kaiwu)
Zhang Yan (captain), Han Wenyan, Li Hao, Wang Zhangjia, Wen Yixue · Advisor: Zhu Hancheng
FT2608179D3065
三等奖
FT2617974D3101
孤独的根号七
第17届中国大学生服务外包创新创业大赛 · 人工智能专项赛道【D01】智能体决策算法（腾讯开悟）
王天宇（队长）、肖逸涵、刘彬、张雅晗、王紫菲 · 指导老师：姚睿、杜文亮
17th China College Student Service Outsourcing Innovation & Entrepreneurship Competition, AI Track D01 "Agent Decision Algorithms" (Tencent Kaiwu)
Wang Tianyu (captain), Xiao Yihan, Liu Bin, Zhang Yahan, Wang Zifei · Advisors: Yao Rui, Du Wenliang
FT2617974D3101
Operating System
协作公约
Working Agreement
Disclose
公开 AI 辅助的模型、使用范围与验证方式。
We disclose the AI models involved, their scope of use, and how outputs are verified.
Review
一切改动走 Pull Request，讨论全程可追踪。
Every change goes through a pull request, with the full discussion traceable.
Reproduce
实验记录、数据来源与部署过程全部留档。
Experiment logs, data sources, and deployment procedures are all archived.
Handoff
交接文档、权限清单与维护边界预先写明。
Handover docs, permission lists, and maintenance boundaries are written down in advance.
Documents
关键文档跳转
Key Documents
Join
加入方式
How to Join
申请表
Application
机试
Coding Test
面试
Interview
Onboarding