Skip to content
View wxhfy's full-sized avatar

Highlights

  • Pro

Block or report wxhfy

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
wxhfy/README.md

👋 Hi, I'm YueYi (wxhfy)

Profile Views

AI Researcher & Developer | 多智能体系统 | 深度学习


🎯 About Me

  • 🔬 AI Researcher — 专注于多智能体系统、肽段生成、跨模态融合
  • 🎮 AI Werewolf Developer — 构建了完整的 AI 狼人杀多智能体对战系统
  • 🤖 OpenClaw Contributor — 参与 AI Agent 自动化平台开发
  • 🧬 Research Projects: TopoBasin, SUCF
  • 🎓 燕山大学 — 热爱技术,追求卓越

🛠️ Tech Stack

Languages

Python TypeScript JavaScript SQL LaTeX

AI/ML Frameworks

PyTorch TensorFlow scikit-learn NumPy Pandas Hugging Face

Web Frameworks

FastAPI Next.js React Flask

Databases & Cache

PostgreSQL MySQL SQLite Redis MongoDB

DevOps & Tools

Docker Kubernetes Git GitHub Actions Linux VS Code Vim

Cloud & Storage

AWS Google Cloud MinIO

Data Science & Visualization

Jupyter Matplotlib Plotly

Bioinformatics

AlphaFold ESM PyMOL

LLM & AI APIs

OpenAI LangChain LlamaIndex


🎮 Featured Projects

面向狼人杀博弈的多智能体认知决策与自进化系统

核心架构: Play → Evaluate → Evolve

模块 能力
对局引擎 7-12人板子、昼夜流程、警徽/PK/遗言/猎人开枪/白狼王自爆
Agent 系统 三层 Prompt (Persona+Role+Strategy)、Memory、SocialModel、工具调用
信息隔离 PlayerView 投影、狼队合法视图、严格公私分离
Track B 评测 三级 LLM Judge (规则→轻量LLM→3人评委团)、多维评分、Leaderboard
Track C 进化 策略知识库、检索增强、A/B 因果验证、自动生命周期管理
前端 Next.js 暗色主题、实时观战、发言气泡、投票可视化、复盘报告

🔬 实验数据 (10 模型 × 火山引擎 Ark):

模型 胜率 平均决策分
Doubao-Seed-2.0-Pro 61.5% 53.0
DeepSeek V4 Flash 50.0% 51.8
Doubao-Seed-2.0-Lite 46.2% 48.8
Kimi K2.5 30.8% 63.6
GLM 4.7 30.8% 46.4
DeepSeek V3.2 30.8% 36.1
DeepSeek V4 Pro 14.3% 43.7

基于 99 局真实 LLM 对局 + 60 局 Track C A/B 对照实验

Tech Stack: Python, FastAPI, Next.js, PostgreSQL, WebSocket, GSAP, ECharts


由自动进化系统生成的技能库

  • ✅ 自动识别重复操作 (17次触发,100%成功率)
  • ✅ 自动提取最佳实践并封装为可复用技能
  • ✅ 持续优化和进化

Tech Stack: Python, OpenClaw


🔬 Research Areas

🤖 Multi-Agent Systems    🧬 Peptide Generation    🔗 Cross-Modal Fusion
    ↓                           ↓                           ↓
AI Werewolf              TopoBasin                  SUCF
(Play→Eval→Evolve)       (拓扑持久性)            (RGAT+Mamba)

🌐 Connect With Me

GitHub


GitHub followers GitHub stars

Pinned Loading

  1. SUCF_AMP SUCF_AMP Public

    🧬 SUCF: 结构门控的跨模态融合网络 — 通过融合蛋白质序列和结构信息预测生物活性,使用 pLDDT 门控 RGAT + Mamba 融合,Baseline 93.89% Accuracy

    Python 1

  2. AIwerewolf AIwerewolf Public

    🎮 AI 狼人杀多智能体对战系统 — 支持 7-12 人板子、LLM/Human Agent、多维评分、自进化策略系统

    Python 4 1