Commit
•
5e68c9e
1
Parent(s):
aaae7db
Upload folder using huggingface_hub
Browse files
README.md
CHANGED
|
@@ -19,26 +19,26 @@ InternVL 2.0 is trained with an 8k context window and utilizes training data con
|
|
| 19 |
|
| 20 |
## Performance
|
| 21 |
|
| 22 |
-
|
|
| 23 |
-
|
|
| 24 |
-
|
|
| 25 |
-
|
|
| 26 |
-
|
|
| 27 |
-
|
|
| 28 |
-
|
|
| 29 |
-
|
|
| 30 |
-
|
|
| 31 |
-
|
|
| 32 |
-
|
|
| 33 |
-
|
|
| 34 |
-
|
|
| 35 |
-
|
|
| 36 |
-
|
|
| 37 |
-
|
|
| 38 |
-
|
|
| 39 |
-
|
|
| 40 |
-
|
|
| 41 |
-
| MathVista<sub>
|
| 42 |
|
| 43 |
- We simultaneously use InternVL and VLMEvalKit repositories for model evaluation. Specifically, the results reported for DocVQA, ChartQA, InfoVQA, TextVQA, MME, AI2D, MMBench, CCBench, MMVet, and SEED-Image were tested using the InternVL repository. MMMU, OCRBench, RealWorldQA, HallBench, and MathVista were evaluated using the VLMEvalKit.
|
| 44 |
|
|
|
|
| 19 |
|
| 20 |
## Performance
|
| 21 |
|
| 22 |
+
| Benchmark | PaliGemma-3B | Phi-3-Vision | Mini-InternVL-4B-1.5 | InternVL2-4B |
|
| 23 |
+
| :--------------------------: | :----------: | :----------: | :------------------: | :----------: |
|
| 24 |
+
| Model Size | 2.9B | 4.2B | 4.2B | 4.2B |
|
| 25 |
+
| | | | | |
|
| 26 |
+
| DocVQA<sub>test</sub> | - | - | 87.7 | 89.2 |
|
| 27 |
+
| ChartQA<sub>test</sub> | - | 81.4 | 81.0 | 81.5 |
|
| 28 |
+
| InfoVQA<sub>test</sub> | - | - | 64.6 | 67.0 |
|
| 29 |
+
| TextVQA<sub>val</sub> | 68.1 | 70.9 | 72.5 | 74.4 |
|
| 30 |
+
| OCRBench | 614 | 639 | 638 | 788 |
|
| 31 |
+
| MME<sub>sum</sub> | 1686.1 | 1508.0 | 2053.6 | 2064.1 |
|
| 32 |
+
| RealWorldQA | 55.2 | 58.8 | 60.1 | 60.7 |
|
| 33 |
+
| AI2D<sub>test</sub> | 68.3 | 76.7 | 76.9 | 78.9 |
|
| 34 |
+
| MMMU<sub>val</sub> | 34.9 | 40.4 | 43.3 | 47.0 |
|
| 35 |
+
| MMBench-EN<sub>test</sub> | 71.0 | 73.6 | 76.2 | 78.6 |
|
| 36 |
+
| MMBench-CN<sub>test</sub> | 63.6 | - | 70.3 | 73.9 |
|
| 37 |
+
| CCBench<sub>dev</sub> | 29.6 | 24.1 | 58.8 | 66.5 |
|
| 38 |
+
| MMVet<sub>GPT-4-0613</sub> | 33.1 | - | 46.7 | 55.7 |
|
| 39 |
+
| SEED-Image | 69.6 | 70.9 | 72.5 | 73.7 |
|
| 40 |
+
| HallBench<sub>avg</sub> | 32.2 | 39.0 | 42.8 | 41.9 |
|
| 41 |
+
| MathVista<sub>testmini</sub> | 28.7 | 44.5 | 53.7 | 58.6 |
|
| 42 |
|
| 43 |
- We simultaneously use InternVL and VLMEvalKit repositories for model evaluation. Specifically, the results reported for DocVQA, ChartQA, InfoVQA, TextVQA, MME, AI2D, MMBench, CCBench, MMVet, and SEED-Image were tested using the InternVL repository. MMMU, OCRBench, RealWorldQA, HallBench, and MathVista were evaluated using the VLMEvalKit.
|
| 44 |
|