Evaluating AI research agents through search strategy, not just results

Original: (1/N) How should we evaluate AI agents that conduct ML research?

Source: arxiv.org ↗