Questions by the AI Safety Student Team at Harvard, inspired by Eli Lifland’s work, on evaluations of AI models against technical benchmarks such as GPQA, InterCode and GAIA.