Hi there 👋 I'am Ming Yang. Welecome to my site~

I am a second-year Master at School of Data Science, Fudan University, advised by Prof. Weiguo Zheng. My research focuses on LLM related technologies, including

• Retrieval-Augmented Generation (RAG) and Approximate Nearest Neighbor Search (ANNS);
• LLM Application Algorithm, including RLVR, SFT, and multi-turn tool-integrated reasoning;
• Agentic System, with an interest in turning foundation models into usable products.

I am a recipient of both the National Scholarship for Undergraduate and Graduate Students. With an expected graduation in July 2027, I am actively seeking opportunities to contribute to LLM systems, AI reasoning/search, and AI agent products.

• [2026/02] One paper accepted by ACL 2026 arXiv
• [2026/03] Our open-source multi-agent systems for presentation delivery is coming DeepSlide
• [2025/09] New work about LLM RLVR arXiv
• [2025/05] One paper accpeted by KDD 2025
• [2024/09] One paper accepted by NeurIPS 2024

• CSPG: Crossing Sparse Proximity Graphs for Approximate Nearest Neighbor Search.
  Ming Yang, Yuzheng Cai, and Weiguo Zheng.
  Thirty-eighth Annual Conference on Neural Information Processing Systems. NeurIPS 2024.
  [link] [pdf] [code] [slides] [poster]

• Hi-PNG: Efficient Interval-Filtering ANNS via Hierarchical Interval Partition Navigating Graph.
  Ming Yang, Yuzheng Cai, and Weiguo Zheng.
  Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining. KDD 2025.
  [link] [code]

• Ant Group — LLM Application Algorithm Intern (AI4Data, Tiansuan Lab · Jun 2025 – Jan 2026)

  • Solved a critical infrastructure issue in the VeRL + vLLM training stack under none-Nvidia GPU + FSDP settings by fixing tensor-view / protobuf synchronization problems, improving stability on internal hardware.
  • Led benchmark construction for related bussiness with AI search and proposed a business-oriented reinforcement strategy for AI search.
  • Produced multiple research outputs, including HAMMER, DiPO, and TSPO.

• Lingjun Investment — LLM Quant Application Algorithm Intern (Dec 2024 – Apr 2025)

  • Built an end-to-end financial report analysis framework for long-document LLM applications
  • Designed multiple information factors for financial text understanding, including confidence expression, positional signals, information density, and structure-aware indicators
  • Closed the loop between data construction and model optimization; improved extraction success rate from 57.3% to 82.7% using rule-based data generation and LlamaFactory fine-tuning

• CSDN — NLP Algorithm Intern (Mar 2024 – Jun 2024)

  • Contributed to (CSDN Tech AI Chatbot), an AI QA product for the developer ecosystem
  • Built data filtering and fine-tuning pipelines for technical QA using Qwen1.5-32B LoRA

• HAMMER: Hamiltonian Curiosity Augmented Large Language Model Reinforcement (first author, under review)
• DiPO: Disentangled Perplexity Policy Optimization for Fine-grained Exploration-Exploitation Trade-Off (co-author, under review)
• TSPO: Breaking the Double Homogenization Dilemma in Multi-turn Search Policy Optimization (co-author, under review)
• DeepSlide: From Artifacts to Presentation Delivery (first author, project lead, under review)