List your company

Multimodal Physical Reasoning on WMW-TRACEBANK synthetic controlled split external-transfer pool

76Answer AccuracyClaude Opus 4.7
40.6449.825968.18May 28, 2026

Evaluation results

MethodLinks
Claude Opus 4.72026.05768168911835
GPT-5.52026.05727761882446
GPT-4o2026.05636849833147
Qwen2.5-VL-72B2026.05606547823336
InternVL3-78B2026.05586344803435
GPT-4o-mini2026.05525538783524
Qwen2.5-VL-7B2026.05424630724223