Vũ Lê
About
I am a Research Scientist in Microsoft's PROSE team,
where I study how AI can understand, reason about, and automate spreadsheets. My research centers on spreadsheet intelligence, from learning rich spreadsheet representations and adapting foundation models for tasks such as formula and action prediction, formula repair, and text-to-formula, to building, evaluating, and training autonomous Excel agents.
My recent work explores reinforcement learning for spreadsheet agents capable of solving complex, multi-step tasks in realistic spreadsheet environments.
My research has contributed to AI experiences across Microsoft products including Excel, Visual Studio, Power BI, SQL Server, Power Apps, Power Automate, and PowerShell. Prior to my current work, I developed novel test generation techniques that uncovered more than 600 previously unknown bugs in the GCC and LLVM compilers.
- Email:
- Phone: 425-421-4582
- Address: 1 Microsoft Way Redmond, WA 98052
- Curriculum Vitae: Download
Awards
- Most Influential OOPSLA Paper Award, OOPSLA '25, Singapore, 2025
- PLDI Distinguished Paper Award, PLDI '14, Edinburgh, 2014
- Best Graduate Researcher Award, Computer Science Department, UC Davis, 2014
- Gold Medal at ACM Student Research Competition, PLDI '13, Seattle, 2013
- PhD Fellowship, Vietnam Education Foundation, 2009
Publications
- A Benchmark and Framework for Evaluating Next Action Predictions in Spreadsheets. ICML '26
Tejas Agrawal, Vu Le, Sumit Gulwani, Gust Verbruggen
- jqBench: a benchmark for reading and editing JSON from natural language and/or examples. ICLR '26
Gust Verbruggen, Chris Parnin, Vu Le, Sumit Gulwani
- The Invisible Mentor: Inferring User Actions from Screen Recordings to Recommend Better Workflows. CHI '26
Litao Yan, Andrew Head, Ken Milne, Vu Le, Sumit Gulwani, Chris Parnin, Emerson Murphy-Hill
- TableTalk: Scaffolding Spreadsheet Development with a Language Agent. TOCHI '25
Jenny T. Liang, Aayush Kumar, Yasharth Bajpai, Sumit Gulwani, Vu Le, Chris Parnin, Arjun Radhakrishna, Ashish Tiwari, Emerson Murphy-Hill, Gustavo Soares
- Evaluating the evaluator: Measuring LLMs’ adherence to task evaluation instructions. AAAI '25
Bhuvanashree Murugadoss, Christian Poelitz, Ian Drosos, Vu Le, Nick McKenna, Carina Suzana Negreanu, Chris Parnin, Advait Sarkar
- Execution-guided within-prompt search for programming-by-example. ICLR '25
Gust Verbruggen, Mukul Singh, Ashish Tiwari, Vu Le, Sumit Gulwani
- TeCoFeS: Text Column Featurization using Semantic Analysis. Findings of NAACL '25
Ananya Singha, Mukul Singh, Ashish Tiwari, Sumit Gulwani, Vu Le, Chris Parnin
- An empirical study of validating synthetic data for formula generation. Findings of NAACL '25
Usneek Singh, José Cambronero, Sumit Gulwani, Aditya Kanade, Anirudh Khatry, Vu Le, Gust Verbruggen
- DataVinci: Learning Syntactic and Semantic String Repairs. SIGMOD '25
Mukul Singh, José Pablo Cambronero, Sumit Gulwani, Vu Le, Carina Negreanu, Arjun Radharishna, Gust Verbruggen
- One-to-many testing for code generation from (just) natural language. Findings of EMNLP '24
Mansi Uniyal, Mukul Singh, Gust Verbruggen, Sumit Gulwani, Vu Le
- RAR: Retrieval Augmented Retrieval for Code Generation in Low Resource Languages. EMNLP '24
Avik Dutta, Mukul Singh, Sumit Gulwani, Vu Le, Gust Verbruggen
- Tabularis Revilio: Converting Text to Tables. CIKM '24
Mukul Singh, Sumit Gulwani, Vu Le, Gust Verbruggen
- Solving Data-centric Tasks using Large Language Models. NAACL '24
Shraddha Barke, Christian Poelitz, Carina Negreanu, Ben Zorn, José Cambronero, Andy Gordon, Vu Le, Elnaz Nouri, Nadia Polikarpova, Advait Sarkar, Brian Slininger, Neil Toronto, Jack Williams
- Semantically Aligned Question and Code Generation for Automated Insight Generation. LLM4Code Workshop, ICSE '24
Ananya Singha, Bhavya Chopra, Anirudh Khatry, Sumit Gulwani, Austin Henley, Vu Le, Chris Parnin, Mukul Singh, Gust Verbruggen
- PyDex: Repairing Bugs in Introductory Python Assignments using LLMs. OOPSLA '24
Jialu Zhang, José Pablo Cambronero, Sumit Gulwani, Vu Le, Ruzica Piskac, Gustavo Soares, and Gust Verbruggen
- FLAME: A Small Language Model for Spreadsheet Formulas. AAAI '24
Harshit Joshi, Abishai Ebenezer, José Cambronero, Sumit Gulwani, Aditya Kanade, Vu Le, Ivan Radicek, Gust Verbruggen
- FormaT5: Abstention and Examples for Conditional Table Formatting with Natural Language. VLDB '24
Mukul Singh, José Cambronero, Sumit Gulwani, Vu Le, Gust Verbruggen
- CODEFUSION: A Pre-trained Diffusion Model for Code Generation. EMNLP '23
Mukul Singh, José Cambronero, Sumit Gulwani, Vu Le, Gust Verbruggen, Carina Negreanu
- TSTR: Target Similarity Tuning Meets the Real World. Findings of EMNLP '23
Anirudh Khatry, Sumit Gulwani, Priyanshu Gupta, Vu Le, Ananya Singha, Mukul Singh, Gust Verbruggen
- EmFore: Online Learning of Email Folder Classification Rules. CIKM '23
Harshit Joshi, José Cambronero, Sumit Gulwani, Vu Le, Ivan Radicek, Gust Verbruggen
- Cornet: Learning Table Formatting Rules By Example. VLDBI '23
Mukul Singh, José Cambronero, Sumit Gulwani, Vu Le, Carina Negreanu, Mohammad Raza, Gust Verbruggen
- Repair Is Nearly Generation: Multilingual Program Repair with LLMs. AAAI '23
Harshit Joshi, José Cambronero, Sumit Gulwani, Vu Le, Ivan Radicek, Gust Verbruggen
- FlashFill++: Scaling Programming by Example by Cutting to the Chase. POPL '23
José Cambronero, Sumit Gulwani, Vu Le, Daniel Perelman, Arjun Radhakrishna, Clint Simon, Ashish Tiwari
- Neurosymbolic Repair for Low-Code Formula Languages. OOPSLA '22
Rohan Bavishi, Harshit Joshi, José Cambronero, Anna Fariha, Sumit Gulwani, Vu Le, Ivan Radicek, Ashish Tiwari
- Personalized Action Suggestions in Low-Code Automation Platforms. ICSE Industry Track '22
Saksham Gupta, Gust Verbruggen, Mukul Singh, Sumit Gulwani, Vu Le
- NL2Viz: Natural Language to Visualization via Constrained Syntax-Guided Synthesis. FSE '22
Zhengkai Wu, Vu Le, Ashish Tiwari, Sumit Gulwani, Arjun Radhakrishna, Ivan Radiček, Gustavo Soares, Xinyu Wang, Zhenwen Li, Tao Xie
- Synchromesh: Reliable Code Generation From Pre-trained Language Models. ICLR '22
Gabriel Poesia, Alex Polozov, Vu Le, Ashish Tiwari, Gustavo Soares, Chris Meek, Sumit Gulwani
- Semantic Programming by Example with Pre-trained Models. OOPSLA '21
Gust Verbruggen, Vu Le, Sumit Gulwani
- Multi-modal Program Inference: a Marriage of Pre-trained Language Models and Component-based Synthesis. OOPSLA '21
Kia Rahmani, Mohammad Raza, Sumit Gulwani, Vu Le, Dan Morris, Arjun Radhakrishna, Gustavo Soares, Ashish Tiwari
- reCode: A Lightweight Find-and-Replace Interaction in the IDE for Transforming Code by Example. UIST '21 [UIST21]
Wode Ni, Joshua Sunshine, Vu Le, Sumit Gulwani, Titus Barik
- Can Program Synthesis be Used to Learn Merge Conflict Resolutions? An Empirical Analysis. ICSE '21 [ICSE21]
Rangeet Pan, Vu Le, Nachi Nagappan, Sumit Gulwani, Shuvendu Lahiri, Mike Kaufman
- An Integrated Approach of Deep Learning and Symbolic Analysis for Digital PDF Table Extraction. ICPR '20 [ICPR20]
Mengshi Zhang, Daniel Perelman, Vu Le, Sumit Gulwani
- Structure Interpretation of Text Formats. OOPSLA '20 [OOPSLA20]
Sumit Gulwani, Vu Le, Arjun Radhakrishna, Ivan Radiček, Mohammad Raza
- On the Fly Synthesis of Edit Suggestions. OOPSLA '19 [OOPSLA19]
Anders Miltner, Sumit Gulwani, Vu Le, Alan Leung, Arjun Radhakrishna, Gustavo Soares, Ashish Tiwari, Abhishek Udupa
Blue-Pencil has been shipped with Visual Studio. Blog post.. - Interactive Program Synthesis. ArXiv preprint '17
Vu Le, Daniel Perelman, Alex Polozov, Mohammad Raza, Abhishek Udupa, Sumit Gulwani - Finding Compiler Bugs via Live Code Mutation. OOPSLA '16 [OOPSLA16]
Chengnian Sun, Vu Le, and Zhendong Su - Toward Understanding Compiler Bugs in GCC and LLVM. ISSTA '16 [ISSTA16]
Chengnian Sun, Vu Le, Qirun Zhang, and Zhendong Su - Finding and Analyzing Compiler Warning Defects. ICSE '16 [ICSE16]
Chengnian Sun, Vu Le, and Zhendong Su - User Interaction Models for Disambiguation in Programming by Example. UIST '15 [UIST15]
Mikaël Mayer, Gustavo Soares, Maxim Grechkin, Vu Le, Mark Marron, Alex Polozov, Rishabh Singh, Ben Zorn, and Sumit Gulwani - Finding Deep Compiler Bugs via Guided Stochastic Program Mutation. [slides] OOPSLA '15 [OOPSLA15]
Vu Le, Chengnian Sun, and Zhendong Su - Randomized Stress-Testing of Link-Time Optimizers. ISSTA '15 [ISSTA15, 27.7%]
Vu Le, Chengnian Sun, and Zhendong Su - Compiler Validation via Equivalence Modulo Inputs. PLDI '14 [PLDI14a, 18%]
Vu Le, Mehrdad Afshari, and Zhendong Su
Distinguished Paper Award
Our technique has found more than 300 bugs in GCC and LLVM, most of which have already been fixed. - FlashExtract: A Framework for Data Extraction by Examples. PLDI '14 [PLDI14b, 18%]
Vu Le and Sumit Gulwani
FlashExtract (demo video) has been shipped as the cmdlet ConvertFrom-String in PowerShell, and as Custom Fields feature in Microsoft Operations Management Suites.
Read about ConvertFrom-String in this PowerShell's blog post.
Other external blog posts about ConvertFrom-String: PowerShell Magazine, Development in a Blink, LazyWinAdmin, Happy SysAdm. - SmartSynth: Synthesizing Smartphone Automation Scripts from Natural Language. MobiSys '13 [MobiSys13a, 16%]
Vu Le, Sumit Gulwani, and Zhendong Su
SmartSynth has been integrated into TouchDevelop (demo video). - Automatic Detection of Floating-Point Exceptions. POPL '13 [POPL13, 18%]
Earl Barr, Thanh Vo, Vu Le, and Zhendong Su
Ariadne has found 2,091 floating-point exceptions in the GNU Scientific Library (GSL).
Education
University of California, Davis, USA
Ph.D. (2009 - 2015)
Advisors: Zhendong Su and Sumit Gulwani Thesis: Program Synthesis for Empowering End Users and Stress-Testing Compilers