Document Visual Question Answering on CORD
88.8ANLSDocVAL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DocVALModel Size=12B, Base Model=Gemma3-12B2025.11 | 88.8 | 78.1 | |
| DocVALModel Size=4B, Base Model=Gemma3-4B2025.11 | 85.5 | 70.2 | |
| Gemma3-12BModel Size=12B2025.11 | 85.3 | — | |
| DLaVAModel Size=Pixtral-12B2025.11 | 84.4 | 57.9 | |
| InternVL3.5-14BModel Size=14B2025.11 | 84.1 | — | |
| InternVL3.5-8BModel Size=8B2025.11 | 82.9 | — | |
| LayTextLLMModel Size=Llama2-7B2025.11 | 82.5 | — | |
| Llama-3.2-11B-VisionModel Size=11B2025.11 | 81.6 | — | |
| Qwen3-VL-8B-ThinkingModel Size=8B2025.11 | 80.7 | — | |
| Gemma3-4BModel Size=4B2025.11 | 78.3 | — | |
| DocLayLLMModel Size=Llama3-7B2025.11 | 71.3 | — | |
| LayoutLLMModel Size=Vicuna1.5-7B2025.11 | 63.1 | — |