01Verification & safety
Can we check what AI produces?
Formal specifications, verifiable transformations, and symbolic guidance for more reliable generation and repair.
Postdoctoral ResearcherLinköping University
Joonghyuk Hahn
한중혁Trustworthy AI,
grounded in language and structure.
I am a postdoctoral researcher in the Trustworthy Systems Group (TSG) at Linköping University. Working with Simin Nadjm-Tehrani, I contribute to RESIST, a research center for resilient and secure AI systems.
Currently based inLinköping, Sweden
Research center
RESIST Resilience and Security for Trustworthy AI SystemsNow at Linköping
Today's language models read untrusted content and act through tools, so trusting them takes more than a better filter. At Linköping, as part of RESIST, I work where LLM verification meets LLM security: checking what these systems produce, understanding how they fail under attack, and recognizing when they should refuse.
Research directions
From structure
to trust.
My work connects formal language theory with modern AI: evaluating what language models can reliably do, checking the correctness of their outputs, and improving how they reason about programs.
02Code intelligence
Can AI reason about programs?
Structure-aware methods for code analysis, complexity prediction, and evaluation beyond functional correctness.
03Language & foundations
What does structure make possible?
Formal grammars, automata, and symbolic learning as foundations for understanding language and computation.
Selected work
A closer look at the research.
Query4Regex
Query4Regex: Verifiable Regex Transformation through Formal Operations from NL and DSL Queries
Evaluating regex transformations with formal operations and deterministic finite automata equivalence checks.
Regex repair
Repairing Regex Vulnerabilities via Localization-Guided Instructions
Localizing vulnerable subpatterns symbolically, then guiding language models to repair regular expressions.
ContractEval
ContractEval: A Benchmark for Evaluating Contract-Satisfying Assertions in Code Generation
Evaluating whether generated code enforces input preconditions, beyond correctness on valid inputs.
* Equal contribution. Illustrations describe research concepts; see each paper for methods and results.
Updates
Recently.
Sep 2026
I joined Linköping University as a postdoctoral researcher in the Trustworthy Systems Group, within the Cybersecurity division, working with Simin Nadjm-Tehrani as part of RESIST. About RESIST
Jul 2026
ContractEval, a benchmark for checking whether generated code enforces its input contracts, appears in Findings of ACL 2026. Paper
Mar 2026
Three papers at EACL 2026: Query4Regex in Findings, and work on regex vulnerability repair and regex minimization in the main conference. Read the papers
Earlier updates
Feb 2026
I received my Ph.D. in Computer Science from Yonsei University, advised by Yo-Sub Han. Background
2026
EnCur, our work on curriculum-based in-context learning for code time complexity prediction, appears in Expert Systems with Applications. Paper
Aug 2025
I presented work on the sparsity of M-unambiguous languages at DLT 2025. Conference program
2025
TCProF, our semi-supervised framework for time-complexity prediction, was presented at NAACL 2025. Paper
Background
Along the way.
- Sep 2026–present
Postdoctoral Researcher
Linköping University · TSG / RESIST
- 2019–Feb 2026
Ph.D. in Computer Science
Yonsei University
Advisor: Yo-Sub Han - 2015–2019
Computer Science
Bachelor’s degree · Yonsei University