ai-bench Repository for benchmarking AI models on code-understanding problems. Some example problems are in folder tasks/assert-tasks To set up repository execute following commands: source ./scripts/bootstrap python3 eval.py -h