model	object_permanence_occlusion	dynamic_spatial	action_intent	predictive_counterfactual	object_deformation_state	egocentric_allocentric
Qwen3-VL-8B-Instruct	49.82	44.51	57.40	44.81	60.66	43.08
LLaVA-OneVision-1.5-8B-Instruct	39.59	44.15	50.60	42.42	51.18	45.18
ERNIE-4.5-VL-28B-A3B-Thinking	41.34	40.17	51.66	40.28	57.35	41.22
ISI-8B-Instruct	46.59	43.01	56.50	43.46	50.24	43.64
Escher-8B-Instruct	46.96	49.80	63.60	48.31	56.87	47.50
Qwen-3-VL-Plus	44.38	43.21	55.52	44.61	57.82	42.38
GPT5	52.24	53.28	64.66	50.41	67.99	50.58
Gemini-2.5-pro	56.17	57.29	69.53	54.20	73.11	54.39
