Hype Meets Reality: Large Language Models as Mutators in Search-based Automated Program Repair of Simulink-Stateflow Models
- Source
- arxiv.org
- Author
- Ayesha Irshad, Pablo Valle, Jon Ayerdi, Aitor Arrieta
- Date

Context
Replacing carefully designed repair operations with language-model suggestions made one automated repair system less effective in a controlled comparison. The study abstract describes modifying FlowRepair to generate some candidate changes with language models, then testing it on 19 faulty Stateflow models under the same elapsed-time budget as the original system.
The modified versions produced patches classified as valid for four models, compared with 16 for the original approach. The authors attribute the difficulty to imprecise symbolic changes, missing behavioral feedback and too many unhelpful candidates. The result concerns this particular integration: flexible suggestions were less useful than structured repair operations within the tested search process.
- LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
articleRethinking Automated Program Repair: The Impact of Bug Complexity, Fault Localization, and LLM Cost-efficiencyJunchi Liu, Ali Bigdeli, Roya Daneshi, Atu Ambala, Sudipto Ghosh, Fabio Santos
articleHow effective are traditional test criteria at detecting bugs in large language models generated code?Asma Hamidi, Michael Konstantinou, Renzo Degiovanni, Mike Papadakis
articleBreaking Models to Test the Judge: A Mutation Testing Approach for Semantic Evaluators of Domain Class DiagramsKevin Delcourt, Meriem Ben Chaaben, Abdelhamid Rouatbi, Luciano Marchezan, Houari Sahraoui
Checking sign-in…
Loading comments…