第二届全球技术创新大赛人工智能专题赛暨2026 Hi Tech全球青年创业大赛
赛事指南
一、参赛对象
项目应符合浙江省“315”科技创新体系、“415X”先进制造业集群，以及符合杭州市“296X”先进制造业集群建设发展方向。
二、参赛要求
项目主申请人应同时符合以下条件：
1.年龄一般不超过50周岁；
2.取得硕士及以上学位；
3.为企业主要创始人、第一大股东，股份不低于15%；
4.拥有项目相关的自主知识产权或掌握核心技术。
参赛限制：
1.已入选杭州高新区(滨江)“5050政策”并领取过创业资助的创业者不能报名参赛。
2.已签订过杭州高新区(滨江)产业支持各类协议并享受过相关政策支持的不能报名参赛。
3.往届总决赛获奖者(项目)不得再次报名，在系统中报名两次及以上的参赛人(项目)不得再次报名，同一项目不得以换参赛人或项目名称的形式多次报名。
三、参赛赛题
赛题一：多模态视觉智能与复杂场景感知技术创新
1.工业复杂场景高精度视觉质检：面向高端制造、电子元器件等工业场景，研发小样本、零缺陷样本条件下的异常检测算法，突破复杂纹理、微小缺陷等检测难点，构建多光谱、多角度融合的视觉质检方案。
2.具身智能三维环境语义理解：面向人形机器人、四足机器人等具身智能体，研发单目/多目视觉与深度信息融合的实时三维建图与语义分割技术，构建动态障碍物检测、物体位姿估计一体化感知系统。
3.端侧设备轻量化视觉推理：面向AI摄像头、智能终端等算力受限设备，研发模型压缩、知识蒸馏、量化加速等轻量化技术，构建面向端侧部署的高效视觉检测与识别算法，支撑低功耗、实时性要求高的视觉应用场景。
4.多模态跨域视觉语义推理：突破视觉与文本、语音、红外等多模态数据的语义对齐与联合推理技术，实现空间关系推理、行为意图识别等高级认知能力，提升视觉系统对真实世界复杂情境的理解与决策支撑水平。
赛题二：行业大模型与智能体关键技术创新
1.垂直领域行业大模型构建与优化：面向金融、制造、医疗等特定行业，研发领域知识注入、持续学习等关键技术，构建具备行业深度认知能力的垂直大模型，实现行业复杂问题的精准解答与专业任务处理。
2.多智能体协同与任务编排技术：研发多智能体分工协作、任务拆解、动态调度核心技术，构建具备角色定义、工具调用能力的智能体开发框架，实现复杂业务场景下多智能体的自主协同与闭环执行。
3.知识增强与可信推理技术：研发RAG检索增强、领域知识库构建、事实一致性校验等技术，构建大模型与结构化/非结构化业务数据的融合推理系统，提升大模型回答的准确性与可溯源性。
4.大模型轻量化与私有化部署：面向企业私有化部署需求，研发大模型参数高效微调、量化压缩、推理加速等技术，构建支持国产算力芯片适配的轻量化大模型部署方案，实现中等规模大模型在单卡/单机环境下的高效推理，降低行业大模型落地的算力门槛与成本。
赛题三：AI+产业融合创新应用方案
1.AI+智能制造升级应用：面向离散制造、流程工业等生产场景，研发基于AI的生产质量检测、设备预测性维护、工艺参数优化等应用方案，推动AI技术深度融入研发设计、生产制造全流程，助力制造业数字化智能化转型。
2.AI+金融科技创新应用：面向银行、证券、保险等金融场景，研发智能风控、智能投研等AI应用方案，利用大模型、知识图谱等技术提升金融机构风险识别能力、客户服务水平与运营管理效率。
3.AI+医疗健康惠民应用：面向临床辅助诊断、医学影像分析、健康管理、药物研发等医疗健康场景，研发安全可靠的AI辅助诊疗系统、智能健康管理工具、医疗数据智能分析平台等应用方案。
4.AI+城市治理智慧应用：面向城市交通、安防、环保、政务服务等治理场景，研发AI赋能的城市运行监测、事件智能发现、应急指挥调度、政务智能问答等应用方案，构建“感知-分析-决策-执行”闭环的城市智慧治理体系，提升城市治理精细化、智能化水平。
四、报名方式
可选择以下任一方式完成报名。
1.大赛官网在线报名：https://reg.gticcn.com/business/
2.专题赛网站在线报名：https://hitech.5050plan.com/
五、赛事奖励
参赛奖金：专题赛设置若干奖项，一等奖、二等奖、三等奖和优胜奖分别最高给予20万元、10万元、5万元和1万元项目奖金。
The AI Special Track of the 2nd Global Technology Innovation Competition & 2026 Hi Tech Global Youth Entrepreneurship Competition
EVENT GUIDE
Ⅰ.ELIGIBILITY
Submissions should align with Zhejiang Province’s “315” Sci-Tech Innovation Program and its “415X” Advanced Manufacturing Action Plan, as well as Hangzhou’s “296X” Advanced Manufacturing Action Plan.
Ⅱ.REQUIREMENTS
Lead founders must meet all of the following requirements:
1.Under the age of 50;
2.Graduating with a Master’s degree or above;
3.Be the primary founder and largest shareholder of the company, with at least 15% equity ownership;
4.In possession of independent intellectual property rights or core technologies related to the project.
Notes:
1.Founders who have previously been selected into Hangzhou Hi-tech Zone (Binjiang)'s “5050 Talent Program” and received startup funding are ineligible.
2.Founders who have already signed industry priority agreements with Hangzhou Hi-tech Zone (Binjiang) and received its supports and benefits are ineligible.
3.Previous grand final winners (or projects) are not eligible to re-enter.Participants (or projects) who have registered for two or more previous editions are not eligible to re-enter. The same project may not be entered multiple times under a different participant name or project title.
Ⅲ.COMPETITION TRACKS
Competition Topic 1:Innovations in Multimodal Visual Intelligence and Complex Scene Perception
1.High-Precision Visual Quality Inspection in Complex Industrial Scenes：For high-end manufacturing and electronic component production, develop anomaly detection algorithms under few-shot and zero-defect sample conditions, overcoming challenges such as complex textures and micro-defects, and construct a multi-spectral, multi-angle fused visual inspection solution.
2.Embodied 3D Semantic Understanding for Intelligent Agents：For embodied agents including humanoid and quadruped robots, develop real-time 3D mapping and semantic segmentation technologies that fuse monocular/stereo vision with depth information, and build an integrated perception system encompassing dynamic obstacle detection and object pose estimation.
3.Lightweight Visual Inference for Edge Devices：For compute-constrained devices such as AI cameras and intelligent terminals, develop lightweight techniques including model compression, knowledge distillation, and quantization acceleration, and implement efficient vision detection and recognition algorithms optimized for edge deployment, supporting low-power, high-real-time visual application scenarios.
4.Multimodal Cross-Domain Visual Semantic Reasoning：Achieve semantic alignment and joint reasoning across multimodal data including vision, text, speech, and infrared, enabling advanced cognitive capabilities such as spatial relation reasoning and behavioral intent recognition, thereby enhancing the visual system's ability to understand and support decision-making in complex real-world contexts.
Competition Topic 2:Key Innovations in Industry-Specific Large Models and Intelligent Agents
1.Construction and Optimization of Vertical-Domain Large Models：For specific industries such as finance, manufacturing, and healthcare, develop key technologies including domain-knowledge infusion and continual learning, build vertical large models with deep domain cognitive capabilities, and enable precise answers to complex industry-specific problems and professional task processing.
2.Multi-Agent Collaboration and Task Orchestration：Develop core technologies for multi-agent division of labor, task decomposition, and dynamic scheduling, build an agent development framework with role definition and tool-calling capabilities, and achieve autonomous collaboration and closed-loop execution among multiple agents in complex business scenarios.
3.Knowledge-Enhanced and Trustworthy Reasoning：Develop technologies such as retrieval-augmented generation (RAG), domain knowledge-base construction, and fact-consistency verification, build a fused reasoning system that integrates large models with structured and unstructured business data, and enhance the accuracy and traceability of model-generated responses.
4.Lightweight and Private Deployment of Large Models：For enterprise private deployment needs, develop parameter-efficient fine-tuning, quantization compression, and inference acceleration technologies, build a lightweight large-model deployment solution compatible with domestic AI accelerator chips, and enable efficient inference for medium-scale large models on single-card or stand-alone environments, thereby lowering the computational cost and barriers for industrial adoption.
Competition Topic 3:AI + Industry Integration Innovative Application Solutions
1.AI + Smart Manufacturing Upgrade Applications：For production scenarios in discrete manufacturing and process industries, develop AI-based application solutions for production quality inspection, predictive equipment maintenance, and process-parameter optimization, driving deep integration of AI throughout the entire R&D and production workflow, and facilitating the digital and intelligent transformation of manufacturing.
2.AI + FinTech Innovative Applications：For financial scenarios in banking, securities, and insurance, develop AI application solutions for intelligent risk control and smart investment research, leveraging large models and knowledge graphs to enhance risk identification, customer service, and operational management efficiency for financial institutions.
3.AI + Healthcare Beneficial Applications：For healthcare scenarios including clinical assisted diagnosis, medical image analysis, health management, and drug discovery, develop safe and reliable AI-assisted diagnostic systems, intelligent health management tools, and smart medical data analytics platforms.
4.AI + Smart City Governance Applications：For urban governance scenarios such as transportation, public safety, environmental protection, and government services, develop AI-enabled solutions for urban operation monitoring, intelligent event detection, emergency command and dispatch, and government intelligent Q&A, building a closed-loop "perception–analysis–decision–execution" smart urban governance system to enhance the refinement and intelligence level of city management.
Ⅳ.REGISTRATION METHOD
Registration may be completed through either of the following methods:
1.Online registration on the official competition website: https://reg.gticcn.com/business/
2.Online registration on the special-track competition website: https://hitech.5050plan.com/
Ⅴ.Competition Prizes
Award and Prizes:The Special Topic Competition features several awards; the first prize, second prize, third prize, and Excellence Award will each be awarded a project prize of up to RMB 200,000, RMB 100,000, RMB 50,000, and RMB 10,000, respectively.
