BrowseComp benchmark for evaluating various AI systems, including language models, Python-based agents, and executable agents in binary format. - View it on GitHub