Update README.md

README.md

@@ -31,292 +31,39 @@ tags:
 size_categories:
 - 10K<n<100K
 ---
-<!DOCTYPE html>
-<html lang="en">
-<head>
-    <meta charset="UTF-8">
-    <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <title>SWE-Long Dataset Card</title>
-    <style>
-        * {
-            box-sizing: border-box;
-            margin: 0;
-            padding: 0;
-        }
+# Dataset Description
 
-        body {
-            font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, Oxygen, Ubuntu, sans-serif;
-            background: #f5f5f7;
-            color: #1d1d1f;
-            line-height: 1.6;
-            padding: 2rem 1rem;
-        }
+This dataset is a heavily filtered version of all the SWE-bench/SWE-smith-lang datasets (expect php) merged together (originally 88,118 rows total), it aims to improve benchmark results on DeepSWE-style problems, benchmarks, and general coding skills.
 
-        .container {
-            max-width: 900px;
-            margin: 0 auto;
-            background: #fff;
-            border-radius: 12px;
-            box-shadow: 0 4px 20px rgba(0, 0, 0, 0.08);
-            overflow: hidden;
-        }
+It is specifically filtered for rows with complex/long code problems in the original datasets, having an average row size of around 94kb and a total size of 6.12GB
 
-        header {
-            background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
-            color: #fff;
-            padding: 2.5rem 2rem;
-        }
+### Dataset Details
 
-        header h1 {
-            font-size: 2rem;
-            margin-bottom: 0.5rem;
-        }
+- **Curated by:** MoreThought
+- **Funded by:** MoreThought
+- **Shared by:** MoreThought
+- **License:** MIT
 
-        header .subtitle {
-            opacity: 0.9;
-            font-size: 1.05rem;
-        }
+### Dataset Sources
 
-        section {
-            padding: 1.75rem 2rem;
-            border-bottom: 1px solid #eaeaea;
-        }
+- **Repositorys:**
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-py
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-go
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-rs
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-ts
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-js
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-cpp
+- https://huggingface.co/datasets/SWE-bench/SWE-smith-java
+- 
+- **Paper:**
+- https://huggingface.co/papers/2504.21798
 
-        section:last-child {
-            border-bottom: none;
-        }
+## Uses
 
-        h2 {
-            font-size: 1.3rem;
-            margin-bottom: 1rem;
-            color: #333;
-            display: flex;
-            align-items: center;
-            gap: 0.5rem;
-        }
+- Improving benchmark results
 
-        h2::before {
-            content: "";
-            display: inline-block;
-            width: 4px;
-            height: 1.1em;
-            background: #667eea;
-            border-radius: 2px;
-        }
+- Improving general coding capabilities
 
-        p {
-            margin-bottom: 0.75rem;
-            color: #444;
-        }
+- Training software engineering/coding agents
 
-        table {
-            width: 100%;
-            border-collapse: collapse;
-            margin: 1rem 0;
-            font-size: 0.95rem;
-            background: #fafafa;
-            border-radius: 8px;
-            overflow: hidden;
-            border: 1px solid #eaeaea;
-        }
-
-        th, td {
-            padding: 0.75rem 1rem;
-            text-align: left;
-            border-bottom: 1px solid #eaeaea;
-        }
-
-        th {
-            background: #f0f0f4;
-            font-weight: 600;
-            color: #333;
-        }
-
-        tbody tr:last-child td {
-            border-bottom: none;
-        }
-
-        tbody tr:hover {
-            background: #f5f5fa;
-        }
-
-        a {
-            color: #667eea;
-            text-decoration: none;
-        }
-
-        a:hover {
-            text-decoration: underline;
-        }
-
-        .stat-grid {
-            display: grid;
-            grid-template-columns: repeat(auto-fit, minmax(180px, 1fr));
-            gap: 1rem;
-            margin: 1rem 0;
-        }
-
-        .stat-card {
-            background: linear-gradient(135deg, #f5f5fa 0%, #fff 100%);
-            border: 1px solid #eaeaea;
-            border-radius: 8px;
-            padding: 1.25rem;
-            text-align: center;
-        }
-
-        .stat-card .value {
-            font-size: 1.5rem;
-            font-weight: 700;
-            color: #667eea;
-            margin-bottom: 0.25rem;
-        }
-
-        .stat-card .label {
-            font-size: 0.85rem;
-            color: #666;
-            text-transform: uppercase;
-            letter-spacing: 0.5px;
-        }
-
-        .tag {
-            display: inline-block;
-            background: #e8e8ff;
-            color: #667eea;
-            padding: 0.2rem 0.6rem;
-            border-radius: 4px;
-            font-size: 0.85rem;
-            margin-right: 0.3rem;
-            margin-bottom: 0.3rem;
-        }
-
-        ul {
-            list-style: none;
-        }
-
-        ul li {
-            padding: 0.4rem 0;
-            color: #444;
-        }
-
-        ul li::before {
-            content: "▸";
-            color: #667eea;
-            margin-right: 0.5rem;
-        }
-    </style>
-</head>
-<body>
-    <div class="container">
-        <header>
-            <h1>SWE-Long Dataset</h1>
-            <p class="subtitle">A filtered, high-complexity subset of SWE-bench for deep code reasoning</p>
-        </header>
-
-        <section>
-            <h2>Overview</h2>
-            <p>
-                This dataset is a heavily filtered consolidation of the SWE-bench and SWE-smith-lang datasets
-                (excluding the PHP language split). The original combined corpus contained
-                <strong>88,118 rows</strong> across multiple programming languages.
-            </p>
-            <p>
-                From that pool, only rows containing <strong>complex and long code problems</strong> were retained.
-                The goal is to strengthen model performance on DeepSWE-style tasks, large-scale coding benchmarks,
-                and extended software engineering scenarios where long-context reasoning matters most.
-            </p>
-
-            <div class="stat-grid">
-                <div class="stat-card">
-                    <div class="value">6.12 GB</div>
-                    <div class="label">Total Size</div>
-                </div>
-                <div class="stat-card">
-                    <div class="value">~94 KB</div>
-                    <div class="label">Avg. Row Size</div>
-                </div>
-                <div class="stat-card">
-                    <div class="value">7</div>
-                    <div class="label">Languages</div>
-                </div>
-                <div class="stat-card">
-                    <div class="value">MIT</div>
-                    <div class="label">License</div>
-                </div>
-            </div>
-        </section>
-
-        <section>
-            <h2>Dataset Details</h2>
-            <table>
-                <thead>
-                    <tr>
-                        <th style="width:30%;">Attribute</th>
-                        <th>Value</th>
-                    </tr>
-                </thead>
-                <tbody>
-                    <tr><td>Curated by</td><td>MoreThought</td></tr>
-                    <tr><td>Funded by</td><td>MoreThought</td></tr>
-                    <tr><td>Shared by</td><td>MoreThought</td></tr>
-                    <tr><td>License</td><td>MIT</td></tr>
-                    <tr><td>Task Type</td><td>Code reasoning, software engineering, benchmark training</td></tr>
-                    <tr><td>Filtering Criteria</td><td>Long code problems; high average row size (~94 KB)</td></tr>
-                </tbody>
-            </table>
-        </section>
-
-        <section>
-            <h2>Source Repositories</h2>
-            <p>
-                Each source below is one SWE-smith language split. All splits were merged except for PHP,
-                then filtered down to the longest and most structurally complex problems.
-            </p>
-            <table>
-                <thead>
-                    <tr>
-                        <th style="width:20%;">Language</th>
-                        <th>Repository</th>
-                    </tr>
-                </thead>
-                <tbody>
-                    <tr><td>Python</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-py" target="_blank" rel="noopener">SWE-bench/SWE-smith-py</a></td></tr>
-                    <tr><td>Go</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-go" target="_blank" rel="noopener">SWE-bench/SWE-smith-go</a></td></tr>
-                    <tr><td>Rust</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-rs" target="_blank" rel="noopener">SWE-bench/SWE-smith-rs</a></td></tr>
-                    <tr><td>TypeScript</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-ts" target="_blank" rel="noopener">SWE-bench/SWE-smith-ts</a></td></tr>
-                    <tr><td>JavaScript</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-js" target="_blank" rel="noopener">SWE-bench/SWE-smith-js</a></td></tr>
-                    <tr><td>C++</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-cpp" target="_blank" rel="noopener">SWE-bench/SWE-smith-cpp</a></td></tr>
-                    <tr><td>Java</td><td><a href="https://huggingface.co/datasets/SWE-bench/SWE-smith-java" target="_blank" rel="noopener">SWE-bench/SWE-smith-java</a></td></tr>
-                </tbody>
-            </table>
-
-            <h2 style="margin-top: 1.5rem;">Reference Paper</h2>
-            <p>
-                <a href="https://huggingface.co/papers/2504.21798" target="_blank" rel="noopener">
-                    SWE-smith: Scaling Data for Software Engineering Agents (2504.21798)
-                </a>
-            </p>
-        </section>
-
-        <section>
-            <h2>Intended Uses</h2>
-            <p>
-                This dataset is designed for training and fine-tuning models that need exposure to
-                realistic, long-horizon software engineering problems.
-            </p>
-            <ul>
-                <li><strong>Benchmark improvement</strong> — Boost performance on DeepSWE and similar long-context SWE benchmarks.</li>
-                <li><strong>General coding capability</strong> — Strengthen reasoning across multi-file, multi-function codebases.</li>
-                <li><strong>SWE agent training</strong> — Train autonomous software engineering and code-repair agents.</li>
-                <li><strong>Long-context coding</strong> — Help models handle inputs of 50K–200K+ tokens with structured reasoning.</li>
-            </ul>
-            <p style="margin-top: 1rem;">
-                <span class="tag">long-context</span>
-                <span class="tag">software-engineering</span>
-                <span class="tag">multi-language</span>
-                <span class="tag">code-reasoning</span>
-                <span class="tag">agent-training</span>
-            </p>
-        </section>
-    </div>
-</body>
-</html>
+- Improving long-context coding