AI's Exam

Latest news

Topics

Sea surface temperature anomalies across the tropical Pacific in August 2026, with the warm El Niño tongue along the equator
Tropical Pacific, August 2026 · NOAA ERSSTv5
Climate science

SciExam for ENSO

Can AI agents build climate models? Agents turn raw ocean data into a model of El Niño and La Niña.

Open12 agent systemsCode and data
Four views of the same bedroom. In one of them the scene has been edited, and the model must find which.
Four views, one contradicts the scene
Vision

Same Space or Not (SSoN)

Can AI models keep one scene in mind? Models find the one view of a room that contradicts the other three.

New1,000 questions6 models and humans

Results

Composite score of each system's final modelEach bar is the composite score, 0.3 × statistics + 0.3 × dynamical consistency + 0.4 × prediction, split into what each part contributes. The published reference model is the baseline, marked by the dashed line.
StatisticsDynamical consistencyPrediction

Principles

  1. 01

    Real scientific questions

    Each exam is built on a question that scientists are still working to answer.

  2. 02

    Scientific validation

    What an agent produces is checked the way a field checks new results, against data and its established standards.

  3. 03

    New insights from agents

    Each exam starts from what a field already knows and asks what an agent’s approach can add to it.

About

AI's Exam is a collection of exams for AI, starting with climate science and vision. Each topic poses its own problem with its own data, tasks and evaluation, and each has its own page with the full results.