SWE-BENCH VERIFIED

Summary from the TLDR AI newsletter

OpenAI has introduced a subset of SWE-bench that is easier and more in line with what humans and AI can solve today. It is a good benchmark for validating and working towards before running the entire original benchmark.