messages listlengths 3 3 |
|---|
[
{
"role": "system",
"content": "You are an expert web designer. Create a polished, self-contained HTML website from the client brief. Think through the design before writing the HTML."
},
{
"role": "user",
"content": "so we run this tiny theater in bushwick — eighty seats, basement, kinetic art ... |
[
{
"role": "system",
"content": "You are an expert web designer. Create a polished, self-contained HTML website from the client brief. Think through the design before writing the HTML."
},
{
"role": "user",
"content": "need a site for our kyoto sashimono atelier, nail-free joinery, feels quiet an... |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
[{"role":"system","content":"You are an expert web designer. Create a polished, self-contained HTML (...TRUNCATED) |
MAD — Selected Web Design Samples
This public dataset contains ten selected supervised fine-tuning examples from the Material-Aware Design (MAD) research project.
Each JSONL row is one independent conversational training record with:
- a web-design system instruction;
- a client brief;
- a design rationale enclosed in a Qwen-style
<think>block; and - one complete, self-contained vanilla HTML/CSS/JavaScript implementation.
Format
qwen35_requested_sites_unlimited.jsonl uses a messages array compatible
with conversational SFT pipelines. The examples are retained at full length;
they are not token-truncated.
Scope
This is a small, deliberately selected public sample, not the complete MAD training corpus. The underlying large corpus remains private.
- Downloads last month
- 21