ScholarlyArticle: "Mutation-Guided LLM-based Test Generation at Meta" (2025) https://arxiv.org/abs/2501.12862v1

Good call, thanks for linking the research paper directly here.

Can this unit test generation capability be connected to the models listed on the SWE-bench [Multimodal] leaderboard?

I'm currently working on running an agent through SWE-Bench (RA.Aid).

What do you mean by connecting the test generation capability to it?

Do you mean generating new eval test cases? I think it could potentially have a use there.

To resolve an [SWE bench] Issue, write tests for the issue and relevant or adjacent code in different commits, w/ FB ACH