Luo et al. (2025)
View original ↗At a glance
- Tests
- Inhibition (Executive functions): holding back the automatic response, reading the word, to give the one asked for: its ink colour.
- How
- Colour words printed in matching or clashing ink; the task is to name the ink. The models see the words as images, as in the original test on paper.
- Best AI
- 93.1%
- Humans
- —
- Our items
- 3
- Executive functions gap, overall
- +17.5
Example item
A specimen written for this page. Live items are similar but never shown, so posted answers can't spoil them.
Name the colour of the ink, not the word.
(a) One word at a time: name the ink
Human vs AI
What the benchmark reports, and where the faculty stands overall
Humans
—
This benchmark doesn't report a human figure.
Executive functions overall · faculty reference, 0–100
Measured on other tests, not on Luo et al. (2025).
Human: GAIA, university-educated testers. AI: GAIA, Claude Sonnet 4.5 (HAL).
How we mirror it
Sixteen colour words per item, shown one at a time, half of them printed in a clashing ink. Tap the ink colour, or press R, G, B or Y.
Every answer and its response time is stored. How much longer clashing words take than matching ones is the classic Stroop effect.
Items are original, versioned, and assigned at random, so a posted answer spoils only one variant.
What we record