IBM open sources CodeAlchemy, a massive synthetic dataset of high-quality code
Release
Kim Martineau
AI for Code
Most enterprises rely on large volumes of aging code that are a challenge to debug, maintain, and update. At IBM Research, we’re using AI to help modernize software stacks, designing automated programming and refactoring systems to help businesses keep up with the speed of modern life.
Our work
IBM’s software engineering agent tops the Multi-SWE-bench leaderboard for Java
NewsPeter HessWatsonx Code Assistant for Z is the Rosetta Stone for mainframes
NewsPeter HessThese new IBM agents will give developers new ways to solve problems — and clear their backlog
NewsMike MurphyWhat’s an LLM context window and why is it getting larger?
NewsKim MartineauSoftware has eaten the world. What now?
Q & AKim Martineau- See more of our work on AI for Code
Publications
Understanding Automated Program Repair Agents Through the Lens of Traceability: An Empirical Study
- Ira Ceka
- Hailie Mitchell
- et al.
- 2026
- ISSTA 2026
Conference paperText-to-SQL Evaluation Toolkit
- 2026
- VLDB 2026
Demo paperPPDL: LLM-Based Flows as Probabilistic Programs
- Louis Mandel
- Guillaume Baudart
- et al.
- 2026
- ICML 2026
Conference paperCharacterizing Agents in Production
- Melissa Pan
- Negar Arabzadeh
- et al.
- 2026
- ICML 2026
Conference paperInvestigating Test Overfitting on SWE-bench
- Toufique Ahmed
- Jatin Ganhotra
- et al.
- 2026
- FSE 2026
Short paperCan Old Tests do New Tricks for Resolving SWE Issues?
- Yang Chen
- Toufique Ahmed
- et al.
- 2026
- FSE 2026
Conference paper
