Spatial and physical reasoning challenges verified by human intuition. Builds a dataset of causal reasoning pairs — one of the hardest capabilities to train in AI.
A bowling ball and a tennis ball are dropped from the same height in a vacuum. Which hits the ground first?
A ball rolls off a table moving horizontally at 5 m/s. What happens immediately after it leaves the edge?
An ice cube melts completely in a full glass of water. What happens to the water level?
Humans develop physical intuition from birth. AI must be explicitly taught what happens when objects collide, fall, or interact.
Each labeled response teaches Lyra a (situation → outcome) mapping — building a model of causality from human common sense.
Unlike abstract language tasks, physics reasoning is grounded in how the real, physical world actually behaves.