Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,7 @@
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
|
|
|
|
3 |
---
|
4 |
|
5 |
# 🧠 CC-HARD: A Challenging Dataset for Design-to-Code Generation
|
@@ -7,7 +9,7 @@ license: cc-by-4.0
|
|
7 |
[📄 Paper on arXiv](https://arxiv.org/pdf/2508.03560)
|
8 |
[📄 Paper on ACM](https://arxiv.org/pdf/2508.03560)
|
9 |
|
10 |
-
**CC-HARD**
|
11 |
It was specifically designed to evaluate layout fidelity in **webpage design-to-code generation**.
|
12 |
|
13 |
The dataset consists of **128 webpage screenshots** and their corresponding **HTML/CSS code**, manually curated from the Common Crawl corpus. Unlike prior datasets, CC-HARD emphasizes:
|
@@ -27,10 +29,8 @@ Each example in **CC-HARD** consists of:
|
|
27 |
- `image`: A high-resolution PNG screenshot of a real-world webpage design
|
28 |
- `text`: The corresponding HTML/CSS code used to render that design
|
29 |
|
30 |
-
---
|
31 |
-
|
32 |
|
33 |
-
##
|
34 |
|
35 |
We compare **CC-HARD** with **Design2Code-HARD** dataset across multiple structural and content dimensions.
|
36 |
The following table summarizes the key statistics, as reported in the *LaTCoder* paper (Table 2):
|
@@ -45,7 +45,7 @@ The following table summarizes the key statistics, as reported in the *LaTCoder*
|
|
45 |
| Avg. Unique Tags | 23 ± 5 | 27 ± 5 |
|
46 |
|
47 |
|
48 |
-
###
|
49 |
|
50 |
As discussed in the paper:
|
51 |
|
@@ -57,6 +57,7 @@ As discussed in the paper:
|
|
57 |
As a result, models that perform well on Design2Code-HARD often struggle on CC-HARD — a trend clearly shown in the benchmark results.
|
58 |
This highlights the **increased layout sensitivity and real-world difficulty** embedded in CC-HARD, making it a more suitable testbed for evaluating layout-aware design-to-code systems.
|
59 |
|
|
|
60 |
|
61 |
## 🧾 Citation
|
62 |
|
@@ -76,6 +77,4 @@ This highlights the **increased layout sensitivity and real-world difficulty** e
|
|
76 |
keywords = {code generation, design to code, ui automation},
|
77 |
location = {Toronto ON, Canada},
|
78 |
series = {KDD '25}
|
79 |
-
}
|
80 |
-
|
81 |
-
|
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
3 |
+
task_categories:
|
4 |
+
- text-generation
|
5 |
---
|
6 |
|
7 |
# 🧠 CC-HARD: A Challenging Dataset for Design-to-Code Generation
|
|
|
9 |
[📄 Paper on arXiv](https://arxiv.org/pdf/2508.03560)
|
10 |
[📄 Paper on ACM](https://arxiv.org/pdf/2508.03560)
|
11 |
|
12 |
+
**CC-HARD** is a challenging benchmark dataset introduced in the KDD 2025 paper *LaTCoder: Converting Webpage Design to Code with Layout-as-Thought*.
|
13 |
It was specifically designed to evaluate layout fidelity in **webpage design-to-code generation**.
|
14 |
|
15 |
The dataset consists of **128 webpage screenshots** and their corresponding **HTML/CSS code**, manually curated from the Common Crawl corpus. Unlike prior datasets, CC-HARD emphasizes:
|
|
|
29 |
- `image`: A high-resolution PNG screenshot of a real-world webpage design
|
30 |
- `text`: The corresponding HTML/CSS code used to render that design
|
31 |
|
|
|
|
|
32 |
|
33 |
+
## Dataset Comparison: CC-HARD vs. Design2Code-HARD
|
34 |
|
35 |
We compare **CC-HARD** with **Design2Code-HARD** dataset across multiple structural and content dimensions.
|
36 |
The following table summarizes the key statistics, as reported in the *LaTCoder* paper (Table 2):
|
|
|
45 |
| Avg. Unique Tags | 23 ± 5 | 27 ± 5 |
|
46 |
|
47 |
|
48 |
+
### Key Differences and Analysis
|
49 |
|
50 |
As discussed in the paper:
|
51 |
|
|
|
57 |
As a result, models that perform well on Design2Code-HARD often struggle on CC-HARD — a trend clearly shown in the benchmark results.
|
58 |
This highlights the **increased layout sensitivity and real-world difficulty** embedded in CC-HARD, making it a more suitable testbed for evaluating layout-aware design-to-code systems.
|
59 |
|
60 |
+
---
|
61 |
|
62 |
## 🧾 Citation
|
63 |
|
|
|
77 |
keywords = {code generation, design to code, ui automation},
|
78 |
location = {Toronto ON, Canada},
|
79 |
series = {KDD '25}
|
80 |
+
}
|
|
|
|