Datasets:
docs: emoji nav links; point Code/eval/load_dataset to AIGrounding
Browse files
README.md
CHANGED
|
@@ -138,7 +138,7 @@ dataset_info:
|
|
| 138 |
|
| 139 |
**ECCV 2026**
|
| 140 |
|
| 141 |
-
🏠 Homepage *(coming soon)* · 💻 [Code](https://github.com/
|
| 142 |
|
| 143 |
**Diagram-MMU** is a benchmark for evaluating Multimodal Large Language Models (MLLMs) on understanding, parsing, and editing **scientific diagrams**. It contains **3,744** curated diagrams (each with compilable source code) and **18,305** human-validated evaluation instances across **six domains** (`charts`, `planar_geometry`, `3d_shapes`, `graph_structures`, `chemistry`, `circuit_diagrams`), over three tasks: diagram-to-code parsing (**D2C-P**), diagram-to-code editing (**D2C-E**), and diagram question answering (**DQA**).
|
| 144 |
|
|
@@ -168,9 +168,9 @@ Each config provides two splits: **`test`** (full benchmark) and **`testmini`**
|
|
| 168 |
```python
|
| 169 |
from datasets import load_dataset
|
| 170 |
|
| 171 |
-
d2cp = load_dataset("
|
| 172 |
-
dqa = load_dataset("
|
| 173 |
-
diagrams = load_dataset("
|
| 174 |
|
| 175 |
ex = d2cp[0]
|
| 176 |
ex["image"] # PIL.Image (decoded automatically)
|
|
@@ -181,7 +181,7 @@ ex["code"] # ground-truth source
|
|
| 181 |
|
| 182 |
All ground truth is public, so evaluation runs locally — no submission. The official evaluation code (object / code / image metrics for **D2C-P** & **D2C-E**, and the rule-based + LLM-as-judge pipeline for **DQA**) will be released separately:
|
| 183 |
|
| 184 |
-
- **Evaluation repository:** [github.com/
|
| 185 |
|
| 186 |
## License & Citation
|
| 187 |
|
|
|
|
| 138 |
|
| 139 |
**ECCV 2026**
|
| 140 |
|
| 141 |
+
🏠 Homepage *(coming soon)* · 💻 [Code](https://github.com/AIGrounding/Diagram-MMU) · 📄 Paper *(coming soon)*
|
| 142 |
|
| 143 |
**Diagram-MMU** is a benchmark for evaluating Multimodal Large Language Models (MLLMs) on understanding, parsing, and editing **scientific diagrams**. It contains **3,744** curated diagrams (each with compilable source code) and **18,305** human-validated evaluation instances across **six domains** (`charts`, `planar_geometry`, `3d_shapes`, `graph_structures`, `chemistry`, `circuit_diagrams`), over three tasks: diagram-to-code parsing (**D2C-P**), diagram-to-code editing (**D2C-E**), and diagram question answering (**DQA**).
|
| 144 |
|
|
|
|
| 168 |
```python
|
| 169 |
from datasets import load_dataset
|
| 170 |
|
| 171 |
+
d2cp = load_dataset("AIGrounding/Diagram-MMU", "d2c-p", split="test") # parsing
|
| 172 |
+
dqa = load_dataset("AIGrounding/Diagram-MMU", "dqa", split="testmini") # QA, dev subset
|
| 173 |
+
diagrams = load_dataset("AIGrounding/Diagram-MMU", "diagrams", split="test")
|
| 174 |
|
| 175 |
ex = d2cp[0]
|
| 176 |
ex["image"] # PIL.Image (decoded automatically)
|
|
|
|
| 181 |
|
| 182 |
All ground truth is public, so evaluation runs locally — no submission. The official evaluation code (object / code / image metrics for **D2C-P** & **D2C-E**, and the rule-based + LLM-as-judge pipeline for **DQA**) will be released separately:
|
| 183 |
|
| 184 |
+
- **Evaluation repository:** [github.com/AIGrounding/Diagram-MMU](https://github.com/AIGrounding/Diagram-MMU)
|
| 185 |
|
| 186 |
## License & Citation
|
| 187 |
|