List your company

Visual Patch Grounding on MedSG-Bench n=1000 (VPG)

30.81IoUQwen2.5-VL-7B
13.733218.166622.627.0334May 28, 2026

Evaluation results

MethodLinks
Qwen2.5-VL-7BPipeline=PDF→code (direct), max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0530.819.8
Qwen2.5-VL-7BPipeline=PDF→CT→code, max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0530.179.7
Qwen2.5-VL-7BPipeline=Reported2026.0528.875.7
Qwen2.5-VL-72BPipeline=PDF→code (direct), max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0528.087
Qwen2.5-VL-3BPipeline=Reported2026.0527.363.4
Qwen2.5-VL-3BPipeline=PDF→code (direct), max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0527.223.2
Qwen2.5-VL-32BPipeline=PDF→CT→code, max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0527.27.8
Qwen2.5-VL-32BPipeline=PDF→code (direct), max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0527.27.8
Qwen2.5-VL-32BPipeline=Reported2026.0526.927.1
Qwen2.5-VL-3BPipeline=PDF→CT→code, max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0526.833.1
Qwen2.5-VL-72BPipeline=Reported2026.0526.456.3
Qwen2.5-VL-72BPipeline=PDF→CT→code, max_new_tokens=128, zero-shot=true, precision=bf16, batch size=1, do_sample=False2026.0514.392