HCAST: Human-Calibrated Autonomy Software Tasks
This provides a practical benchmark for assessing AI autonomy in real-world tasks, though it is incremental in refining evaluation methods.
The paper tackles the problem of evaluating AI systems' societal impacts by introducing HCAST, a benchmark of 189 tasks with human-calibrated metrics, and finds that current AI agents succeed 70-80% on tasks taking humans under an hour but less than 20% on tasks taking over 4 hours.
To understand and predict the societal impacts of highly autonomous AI systems, we need benchmarks with grounding, i.e., metrics that directly connect AI performance to real-world effects we care about. We present HCAST (Human-Calibrated Autonomy Software Tasks), a benchmark of 189 machine learning engineering, cybersecurity, software engineering, and general reasoning tasks. We collect 563 human baselines (totaling over 1500 hours) from people skilled in these domains, working under identical conditions as AI agents, which lets us estimate that HCAST tasks take humans between one minute and 8+ hours. Measuring the time tasks take for humans provides an intuitive metric for evaluating AI capabilities, helping answer the question "can an agent be trusted to complete a task that would take a human X hours?" We evaluate the success rates of AI agents built on frontier foundation models, and we find that current agents succeed 70-80% of the time on tasks that take humans less than one hour, and less than 20% of the time on tasks that take humans more than 4 hours.