Share your thoughts, 1 month free Claude Pro on usSee more

Goal-conditioned visual planning on CrossTask T=3 88 (test)

51.71Success Rate (SR)

GeoWorld ViT-g384

Updated 15d ago

Evaluation Results

Method	Links
GeoWorld ViT-g384 2026.02		51.71	77.3	92.95
V-JEPA 2 ViT-g384 2026.02		50.16	74.86	91.73
GPT-5 2026.02		50.03	72.38	91.18
GeoWorld ViT-g 2026.02		49.23	76.64	90.61
Gemini 2.5 Pro 2026.02		48.91	73.82	90.3
V-JEPA 2 ViT-g 2026.02		48.13	73.42	89.62
GeoWorld ViT-H 2026.02		47.79	74.42	88.84
V-JEPA 2 ViT-H 2026.02		46.02	71.98	87.29
Qwen3-VL-Max 2026.02		45.47	70.93	86.18
GeoWorld ViT-L 2026.02		44.8	70.54	86.3
InternVL3.5-241B 2026.02		44.03	70.01	84.41
V-JEPA 2 ViT-L 2026.02		43.36	69.55	84.75
VideoWorld 2026.02		41.59	66.11	82.64