Optical Character Recognition On Ocrbench V2 Chinese
Metriken
Accuracy
Ergebnisse
Leistungsergebnisse verschiedener Modelle zu diesem Benchmark
Modellname | Accuracy | Paper Title | Repository |
|---|---|---|---|
| Ovis2.5-8B | 55.4 | Ovis2.5 Technical Report | - |
| Seed 1.6-vision | 60.5 | - | - |
| Kimi-VL-A3B-16B | 54.1 | Kimi-VL Technical Report | |
| MiniCPM-V-4.5-8B | 58.8 | MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe | - |
| Gemini-2.5-Pro | 62.2 | - | - |
| Ovis2-8B | 56.0 | - | - |
| Qwen3-Omni-30B-A3B-Instruct | 60.0 | Qwen3-Omni Technical Report | - |
| WeThink-Qwen2.5VL-7B | 55.8 | WeThink: Toward General-purpose Vision-Language Reasoning via Reinforcement Learning | |
| SAIL-VL2-8B | 57.6 | SAIL-VL2 Technical Report | - |
| Gemini1.5-Pro | 55.5 | - | - |
0 of 10 row(s) selected.