AIGrounding commited on
Commit
50f00b9
·
verified ·
1 Parent(s): 0357782

docs: emoji nav links; point Code/eval/load_dataset to AIGrounding

Browse files
Files changed (1) hide show
  1. README.md +5 -5
README.md CHANGED
@@ -138,7 +138,7 @@ dataset_info:
138
 
139
  **ECCV 2026**
140
 
141
- 🏠 Homepage *(coming soon)* · 💻 [Code](https://github.com/weihao-bo/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
 
@@ -168,9 +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("weihao-bo/Diagram-MMU", "d2c-p", split="test") # parsing
172
- dqa = load_dataset("weihao-bo/Diagram-MMU", "dqa", split="testmini") # QA, dev subset
173
- diagrams = load_dataset("weihao-bo/Diagram-MMU", "diagrams", split="test")
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/weihao-bo/Diagram-MMU](https://github.com/weihao-bo/Diagram-MMU)
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