Zeqi Lin
Principal Researcher
About
Dr. Zeqi Lin is currently a Principal Researcher at Microsoft CoreAI Post-Training Team. He obtained his Ph.D from Software Engineering Institute, Peking University in 2019 (supervised by Prof. Bing Xie and Prof. Lu Zhang). He joined Microsoft Research Asia (Beijing) in 2019 and Microsoft GenAI (Redmond) in 2024, focusing on LLM-Powered Coding and Reasoning.
Projects
- Reinforcement Learning of GPT-4o Copilot Completion Model (opens in new tab)
- Phi-3: A Highly Capable Language Model Locally on Your Phone (opens in new tab)
- Azure OpenAI on Your Data (opens in new tab)
- Project Sophia, A New Generation AI-First Business Application (opens in new tab)
- Power BI Quick Measures Using Natural Language (opens in new tab)
- Power Automate Flow with Natural Language (opens in new tab)
- Write Power Fx Formulas with Natural Language (opens in new tab)
- Natural Language Queries in Excel Ideas (opens in new tab)
- NuGetSolver: A Powerful Tool for Resolving NuGet Dependency Conflicts in Visual Studio (opens in new tab)
Publications
- Self-Evolved Reward Learning for LLMs (opens in new tab) (ICLR 2025)
- Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs (opens in new tab)
- STAND-Guard: A Small Task-Adaptive Content Moderation Model (opens in new tab) (COLING 2025 Industry Track)
- Make Your LLM Fully Utilize the Context (opens in new tab) (NeurIPS 2024)
- Can LLMs Learn From Mistakes? An Empirical Study on Reasoning Tasks (opens in new tab) (EMNLP 2024 Findings)
- Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone (opens in new tab)
- Compositional API Recommendation for Library-Oriented Code Generation (opens in new tab) (ICPC 2024)
- CodeT: Code Generation with Generated Tests (opens in new tab) (ICLR 2023)
- Making Language Models Better Reasoners with Step-Aware Verifier (opens in new tab) (ACL 2023)
- How Do In-Context Examples Affect Compositional Generalization? (opens in new tab) (ACL 2023)
- Does Deep Learning Learn to Abstract? A Systematic Probing Framework (opens in new tab) (ICLR 2023)
- Skill-Based Few-Shot Selection for In-Context Learning (opens in new tab) (EMNLP 2023)
- TAPEX: Table Pre-Training via Learning a Neural SQL Executor (opens in new tab) (ICLR 2022)
- Nufix: Escape From NuGet Dependency Maze (opens in new tab) (ICSE 2022)
- Reasoning Like Program Executors (opens in new tab) (EMNLP 2022)
- CERT: Continual Pre-Training on Sketches for Library-Oriented Code Generation (opens in new tab) (IJCAI 2022)
- When Language Model Meets Private Library (opens in new tab) (EMNLP 2022 Findings)
- Can Neural Clone Detection Generalize to Unseen Functionalities? (ASE 2021)
- Learning Algebraic Recombination for Compositional Generalization (opens in new tab) (ACL 2021 Findings)
- Revisiting Iterative Back-Translation from the Perspective of Compositional Generalization (opens in new tab) (AAAI 2021)
- Iterative Utterance Segmentation for Neural Semantic Parsing (opens in new tab) (AAAI 2021)
- Compositional Generalization by Learning Analytical Expressions (opens in new tab) (NeurIPS 2020)
- Hierarchical Poset Decoding for Compositional Generalization in Language (opens in new tab) (NeurIPS 2020)
- Adaptive Deep Code Search (opens in new tab) (ICPC 2020)
- CoRA: Decomposing and Describing Tangled Code Changes for Reviewer (opens in new tab) (ASE 2019)
- Improving Software Text Retrieval Using Conceptual Knowledge in Source Code (opens in new tab) (ASE 2017)
- Intelligent Development Environment and Software Knowledge Graph (opens in new tab) (JCST 2017)
- Mining API Usage Examples from Test Code (opens in new tab) (ICSME 2014)