Multi-SWE-bench
A benchmark that extends the SWE-bench paradigm to multiple programming languages, covering Java, Go, Rust and others, and is used to examine cross-language generalization of models.
Timeline (last 90 days)
Multi-SWE-bench extends the SWE-bench paradigm to multiple programming languages, covering Java, Go, Rust, and others, to test models' cross-language generalization ability.
SWE-TQ is a new benchmark covering eight programming languages, obtained by cleaning human annotations from SWE-bench and Multi-SWE-bench, for evaluating task clarity and test coverage.
All Facts (2)
Multi-SWE-bench extends the SWE-bench paradigm to multiple programming languages, covering Java, Go, Rust, and others, to test models' cross-language generalization ability.
50%UnverifiedSWE-TQ is a new benchmark covering eight programming languages, obtained by cleaning human annotations from SWE-bench and Multi-SWE-bench, for evaluating task clarity and test coverage.
50%