Scene Text Understanding on OCRBench

804Score

InternVL2.5

Updated 19d ago

Evaluation Results

Method	Links
InternVL2.5 2025.12		804
Qwen2.5-VL 3B (reproduced) 2025.12		804
Qwen2.5-VL 3B (reported) 2025.12		797
CASA⊕ Qwen2.5-VL 2025.12		790
VideoLLaMA3 2025.12		779
Qwen2-VL 2025.12		767
SmolVLM 2025.12		729
InsertionHe-2B 2025.12		728
CASA⊕ 2025.12		728
CASA→ 2025.12		723
CASA∨ 2025.12		694
mPLUG-Owl3 8B 2025.12		527
mPLUG-Owl3 2B 2025.12		450
InternVL3.5-38B 2026.07		0.866
Qwen2.5-VL-72B-Instruct 2026.07		0.84
Qwen2.5-VL-7B-Instruct 2026.07		0.837
InternVL3.5-8B 2026.07		0.827
InternVL3.5-4B 2026.07		0.813
Optimization-based Token Selection 2026.07		0.812
Qwen2.5-VL-3B-Instruct 2026.07		0.789
Permutation Attack 2026.07		0.776
Masking Attack 2026.07		0.735
Gaussian Perturbation Attack 2026.07		0.668
Optimization-based Token Selection 2026.07		0.655
VTM-Attack (Naïve Average) 2026.07		0.634
Sign Flip Attack 2026.07		0.357
Optimization-based Token Selection 2026.07		0.305
Optimization-based Token Selection 2026.07		0.157
Optimization-based Token Selection 2026.07		0.15
Optimization-based Token Selection 2026.07		0.02