Long multimodal document understanding on MM-NIAH
46.1Overall ScoreInternVL1.5-RAG
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| InternVL1.5-RAG#param=45B, zero-shot=true2024.10 | 46.1 | 59.5 | 50.1 | 44.9 | 39.3 | |
| PDF-WuKong#param=8.5B, zero-shot=true2024.10 | 43.3 | 53 | 43.9 | 43 | 42.1 | |
| InternVL1.5#param=26B, zero-shot=true2024.10 | 41.1 | 59.5 | 50.1 | 41.9 | 16.6 | |
| llava-v1.6-34b#param=34B, zero-shot=true2024.10 | 20.6 | 57.4 | 45.1 | 8.2 | 0 | |
| llava-v1.6-13b#param=13B, zero-shot=true2024.10 | 16.9 | 43.7 | 34.9 | 13.6 | 0 | |
| VILA1.0-13b#param=13B, zero-shot=true2024.10 | 15.7 | 41.9 | 33.2 | 8.6 | 0.1 | |
| Emu2-Chat#param=37B, zero-shot=true2024.10 | 8.8 | 38.9 | 18.2 | 0 | 0 |