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

Perception on TreeBench

55.3Overall Accuracy

PFlowNet

Updated 29d ago

Evaluation Results

Method	Links
PFlowNet 2026.05		55.3	65.5	69.2	80.2	75	77.9
TreeVGR 2026.05		50.4	65.5	53.8	82.6	68.8	63.3
InternVL3-78B 2026.05		46.4	62.1	61.5	52.2	68.8	52.9
Qwen3-VL-32B 2026.05		45.2	60.3	63.4	58.1	83.6	30.3
Qwen3-VL-8B 2026.05		44.9	65.5	53.9	65.2	75	64.7
Qwen2.5-VL-32B 2026.05		42.5	51.7	53.8	69.6	62.5	54.4
Qwen2.5-VL-72B 2026.05		42.2	65.5	69.2	56.5	56.3	48.5
Qwen3-VL-4B 2026.05		42.2	48.3	61.5	65.2	81.3	35.3
InternVL3-38B 2026.05		42	51.7	61.5	52.2	68.8	51.5
DeepEyesV2 2026.05		40.7	65.5	69.2	56.5	62.5	55.9
LLaVA-OV-72B 2026.05		40.5	62.1	53.8	65.2	62.3	36.8
Pixel-Reasoner 2026.05		39	58.6	61.5	65.2	50	48.5
InternVL3-8B 2026.05		38.8	51.7	69.2	56.5	56.3	33.7
Thyme 2026.05		38.2	48.2	46.1	69.5	50	51.4
DeepEyes 2026.05		37.5	62.1	53.8	65.2	68.8	51.5
LLaVA-OV-7B 2026.05		37.3	55.2	53.8	56.5	50	32.4
Qwen2.5-VL-7B 2026.05		37	55.2	53.8	56.5	62.5	27.9