ResearchBenchmarksPredictive Video Reasoning on CLEVRER (val)Follow87.5AccuracyALOE67.94873.02478.183.176May 15, 2026Evaluation ResultsMethodMethodLinksAccuracyALOEevaluation_format=per-...evaluation_format=per-question, n=3,5572026.0587.5Substrateevaluation_format=per-...evaluation_format=per-option, n=3,5572026.0584.07NS-DRevaluation_format=per-...evaluation_format=per-option, n=3,5572026.0582.9Substrateevaluation_format=per-...evaluation_format=per-question, n=3,5572026.0569.5NS-DRevaluation_format=per-...evaluation_format=per-question, n=3,5572026.0568.7