SciPredict: Can LLMs Predict the Outcomes of Scientific Experiments in Natural Sciences?

Accelerating scientific discovery requires the identification of which experiments would yield the best outcomes before committing resources to costly physical validation. While existing benchmarks evaluate LLMs on scientific knowledge and reasoning, their ability to predict experimental outcomes---a task where AI could significantly exceed human capabilities---remains largely underexplored. We introduce SciPredict, a benchmark comprising 405 tasks derived from recent empirical studies in 33 specialized sub-fields of physics, biology, and chemistry. SciPredict addresses two critical questions: (a) *can LLMs predict the outcome of scientific experiments with sufficient accuracy?* and (b) *can such predictions be reliably used in the scientific research process?* Evaluations reveal fundamental limitations on both fronts. Model accuracies are 14-26\% and human expert performance is $\approx$20\%. Although some frontier models exceed human performance model accuracy is still far below what would enable reliable experimental guidance. Even within the limited performance, models fail to distinguish reliable predictions from unreliable ones, achieving only $\approx$20\% accuracy regardless of their confidence or whether they judge outcomes as predictable without physical experimentation. Human experts, in contrast, demonstrate strong calibration: their accuracy increases from $\approx$5\% to $\approx$80\% as they deem outcomes more predictable without conducting the experiment. SciPredict establishes a rigorous framework demonstrating that superhuman performance in experimental science requires not just better predictions, but better awareness of prediction reliability. For reproducibility all our data and code are provided at https://anonymous.4open.science/r/SciPredict-AI01.

Can AI predict what will happen in a science experiment before it's actually run? This paper introduces SciPredict, a test of 405 questions drawn from real, recently published experiments in physics, biology, and chemistry to find out. The results are humbling: the best AI models only get about 14–26% of predictions right, roughly on par with human experts (\~20%). But the bigger problem isn't just accuracy — it's that AI models can't tell when they're guessing well versus guessing badly. They report similar confidence whether they're right or wrong. Human experts, by contrast, have good instincts about which questions they can answer without running the experiment (reaching \~80% accuracy on those) and which ones they can't (\~5%). The study also found that giving AI relevant background information helps a little, but when models try to come up with their own background knowledge, it actually makes their predictions worse. The takeaway: AI isn't yet reliable enough to guide scientists on which experiments are worth pursuing, and the missing piece isn't just smarter predictions — it's knowing when to trust those predictions.