1 articles with this tag
OpenAI's audit reveals approximately 30% of SWE-Bench Pro's coding tasks are flawed, prompting the company to retract its recommendation for the benchmark.
