13 Models and 4 Agents on SWE Tasks: Go, Java, Python, Rust, TS
SWE-rebench: A Continuously Evolving and Decontaminated Benchmark for Software Engineering LLMs.
Summary
The article discusses SWE-rebench, a new benchmark designed for evaluating software engineering language models across multiple programming languages. It highlights how this tool aims to improve the performance and reliability of these models in software development tasks.
Comments (11 of ) ⌛