Jingxuan He

I am a PostDoc at UC Berkeley, working with Dawn Song. I received my PhD from the SRI Lab of ETH Zurich, where I was advised by Martin Vechev. I did my undergraduate at Zhejiang University.

何静轩 / jingxuan.he [at] berkeley.edu / Scholar / Twitter / LinkedIn

Honors and Awards

Two Spotlight Papers at ICML 2025
ETH Medal for Oustanding Doctoral Thesis, 2024
ACM CCS Distinguished Paper Award, 2023
NeurIPS Top Reviewer, 2023

Research

I am broadly interested in security, machine learning, and programming languages. My recent research focuses on enhancing the security of AI-generated code, utilizing learning-based techniques that improve AI's security reasoning and formal methods that provide rigorous guarantees (see highlighted papers). I also work on related topics, including AI for programming and AI safety.

You can selectively view my publications by the focused area of the venue:

Type-Constrained Code Generation with Language Models

Niels Mündler*, Jingxuan He*, Hao Wang, Koushik Sen, Dawn Song, Martin Vechev
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), 2025
paper / code /

BaxBench: Can LLMs Generate Secure and Correct Backends?

Mark Vero, Niels Mündler, Victor Chibotaru, Veselin Raychev, Maximilian Baader, Nikola Jovanović, Jingxuan He, Martin Vechev
International Conference on Machine Learning (ICML), 2025

(Spotlight)

paper / code / website /

Formal Mathematical Reasoning: A New Frontier in AI

Kaiyu Yang, Gabriel Poesia, Jingxuan He, Wenda Li, Kristin Lauter, Swarat Chaudhuri, Dawn Song
International Conference on Machine Learning (ICML), 2025

(Spotlight)

paper /

Mind the Gap: A Practical Attack on GGUF Quantization

Kazuki Egashira, Robin Staab, Mark Vero, Jingxuan He, Martin Vechev
International Conference on Machine Learning (ICML), 2025
ICLR 2025 Workshop on Building Trust in Language Models and Applications, 2025

(Oral)

                Black-Box Adversarial Attacks on LLM-Based Code Completion
                
                Slobodan Jenko*, Niels Mündler*, Jingxuan He, Mark Vero, Martin Vechev

                International Conference on Machine Learning (ICML), 2025

Exploiting LLM Quantization

Kazuki Egashira, Mark Vero, Robin Staab, Jingxuan He, Martin Vechev
Neural Information Processing Systems (NeurIPS), 2024
ICML 2024 Workshop on the Next Generation of AI Safety, 2024

(Oral)

paper / code / website /

SWT-Bench: Testing and Validating Real-World Bug-Fixes with Code Agents

Niels Mündler, Mark Niklas Müller, Jingxuan He, Martin Vechev
Neural Information Processing Systems (NeurIPS), 2024
paper / code / poster /

Instruction Tuning for Secure Code Generation

Jingxuan He*, Mark Vero*, Gabriela Krasnopolska, Martin Vechev
International Conference on Machine Learning (ICML), 2024
paper / code /

Self-contradictory Hallucinations of Large Language Models: Evaluation, Detection and Mitigation

Niels Mündler, Jingxuan He, Slobodan Jenko, Martin Vechev
International Conference on Learning Representations (ICLR), 2024
paper / code / website /

Large Language Models for Code: Security Hardening and Adversarial Testing

Jingxuan He, Martin Vechev
ACM Conference on Computer and Communications Security (CCS), 2023

(Distinguished Paper)

paper / code / slides /

On Distribution Shift in Learning-based Bug Detectors

Jingxuan He, Luca Beurer-Kellner, Martin Vechev
International Conference on Machine Learning (ICML), 2022
paper / code /

Learning to Explore Paths for Symbolic Execution

Jingxuan He, Gishor Sivanrupan, Petar Tsankov, Martin Vechev
ACM Conference on Computer and Communications Security (CCS), 2021
paper / code / slides /

TFix: Learning to Fix Coding Errors with a Text-to-Text Transformer

Berkay Berabi, Jingxuan He, Veselin Raychev, Martin Vechev
International Conference on Machine Learning (ICML), 2021
paper / code / talk / slides /

Learning to Find Naming Issues with Big Code and Small Supervision

Jingxuan He, Cheng-Chun Lee, Veselin Raychev, Martin Vechev
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), 2021
paper / talk / slides /

Learning Fast and Precise Numerical Analysis

Jingxuan He, Gagandeep Singh, Markus Püschel, Martin Vechev
ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI), 2020
paper / code / talk / slides /

Learning to Fuzz from Symbolic Execution with Application to Smart Contracts

Jingxuan He, Mislav Balunović, Nodar Ambroladze, Petar Tsankov, Martin Vechev
ACM Conference on Computer and Communications Security (CCS), 2019
paper / code / talk / slides /

DeBin: Predicting Debug Information in Stripped Binaries

Jingxuan He, Pesho Ivanov, Petar Tsankov, Veselin Raychev, Martin Vechev
ACM Conference on Computer and Communications Security (CCS), 2018
paper / code / talk / slides /

Design and source code from Leonid Keselman's and Jon Barron's websites.