Apoorva Sharma

Litmus redefines AI evaluation by showing that zero-label metric specification can outperform traditional methods, achieving higher validity and broader concern coverage.

Steering imaginations in video world models can reveal critical failure points in robotic actions that traditional methods might overlook.