Evaluating LLMs for Scientific Discovery: Insights from ScienceAgentBench
How well can AI tackle real-world science? ScienceAgentBench puts 102 tasks to the test, exposing both its breakthroughs and struggles.
Follow the subjects shaping intelligence, science, culture, and the future of technology.
Find a subject, then follow its latest reporting and ideas.
Stories, interviews, and analysis in this subject.