Skip to content

[codex] Update LexBench landing page #65

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
waple0820 merged 1 commit into from
Jun 22, 2026
+79 −145
Merged

[codex] Update LexBench landing page #65

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Binary file modified landing/assets/img/task-distribution.png
View file
Open in desktop
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
224 changes: 79 additions & 145 deletions landing/index.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@
<meta property="og:type" content="website">
<meta property="og:site_name" content="LexBench-Browser">
<meta property="og:title" content="LexBench-Browser — Real-World Browser Agent Benchmark">
<meta property="og:description" content="210 public browser-agent tasks across 107 real websites, with multilingual long-tail workflows, stepwise judging, and reproducible leaderboards.">
<meta property="og:description" content="377 live-web browser-agent tasks across 200+ websites, with multilingual authenticated workflows, stepwise judging, and reproducible leaderboards.">
<meta property="og:url" content="https://lexmount.github.io/browseruse-agent-bench/">
<meta property="og:image" content="https://lexmount.github.io/browseruse-agent-bench/assets/img/hero.png">
<meta name="twitter:card" content="summary_large_image">
Expand Down Expand Up @@ -39,7 +39,6 @@
<a href="#agents">Agents</a>
<a href="#community">Community</a>
<a href="#leaderboard">Leaderboard</a>
<a href="#citation">Cite</a>
<a class="topnav-cta" href="https://github.com/lexmount/browseruse-agent-bench" target="_blank" rel="noopener">GitHub →</a>
</div>
</div>
Expand Down Expand Up @@ -140,43 +139,44 @@ <h2 class="section-title">Overview</h2>
<div class="container">
<h2 class="section-title">Tasks</h2>
<p class="section-lede">
The first dataset, <strong>LexBench-Browser</strong>, focuses on browser-agent tasks
that resemble actual user workflows: search, e-commerce, video, social, academic and
tool-use across both English and Chinese websites, with varying difficulty tiers.
<strong>LexBench-Browser</strong> is a live-web benchmark for realistic browser-agent
workflows across Chinese and English websites. It covers information retrieval and
state-changing task execution, including authenticated and safety-sensitive scenarios
that prior benchmarks often under-represent.
</p>

<div class="stat-grid">
<div class="stat-card">
<div class="stat-num">210</div>
<div class="stat-num">377</div>
<div class="stat-label">Tasks total</div>
<div class="stat-sub">across 107 distinct websites</div>
<div class="stat-sub">across 200+ target websites</div>
</div>
<div class="stat-card">
<div class="stat-num">4</div>
<div class="stat-label">Reference agents</div>
<div class="stat-sub">browser-use · deepbrowse · Agent-TARS · skyvern</div>
<div class="stat-num">2</div>
<div class="stat-label">Task types</div>
<div class="stat-sub">294 information retrieval · 83 task execution</div>
</div>
<div class="stat-card">
<div class="stat-num">15</div>
<div class="stat-num">10</div>
<div class="stat-label">Models evaluated</div>
<div class="stat-sub">GPT · Claude · Gemini · Doubao · Kimi · Qwen · DeepSeek · MiniMax</div>
<div class="stat-sub">BU · GLM · GPT · Claude · Gemini · Doubao · Kimi · Qwen · MiniMax</div>
</div>
<div class="stat-card">
<div class="stat-num">3</div>
<div class="stat-label">Browser backends</div>
<div class="stat-sub">Chrome-Local · Lexmount Cloud · AgentBay</div>
<div class="stat-num">167</div>
<div class="stat-label">Login-required tasks</div>
<div class="stat-sub">plus 25 safety-testing tasks</div>
</div>
</div>

<figure class="wide-fig">
<img src="assets/img/task-distribution.png" alt="Task distribution diagram showing the categorical and difficulty breakdown of LexBench-Browser tasks" loading="lazy">
<figcaption>Categorical and difficulty distribution across the LexBench-Browser task pool.</figcaption>
<img src="assets/img/task-distribution.png" alt="Task distribution diagram showing LexBench-Browser domains and task types" loading="lazy">
<figcaption>Domain and task-type distribution across the LexBench-Browser task pool.</figcaption>
</figure>
<div class="muted-card">
<strong>Login-gated and operation-tier tasks are coming soon.</strong>
The next release adds tasks that require account context, multi-step transactions
and safety-sensitive flows — to evaluate Agents on the parts of the web that
English-only benchmarks miss.
<strong>Authenticated and operational workflows are included.</strong>
The benchmark contains 281 Chinese and 96 English tasks, with live sessions, account
context, multi-step operations and safety-sensitive flows evaluated under the same
browser-agent protocol.
</div>
</div>
</section>
Expand Down Expand Up @@ -332,13 +332,13 @@ <h2 class="section-title">Community</h2>
<div class="container">
<h2 class="section-title">Leaderboard</h2>
<p class="section-lede">
Live results across all reference agents and models. Click any header to sort. Showing the
top 15 entries by success rate.
Snapshot results for browser-use on LexBench-Browser across 10 models. Click any header to
sort. Showing all 10 entries by success rate.
</p>
<div class="lb-meta">
<span class="lb-pill">benchmark · LexBench-Browser</span>
<span class="lb-pill">judge · gpt-5.4 (per-task threshold)</span>
<span class="lb-pill">snapshot · 2026-04-29</span>
<span class="lb-pill">snapshot · 2026-06-16</span>
</div>
<div class="lb-wrap">
<table class="lb-table" id="lb-table">
Expand All @@ -354,172 +354,106 @@ <h2 class="section-title">Leaderboard</h2>
</tr>
</thead>
<tbody>
<tr data-rank="1" data-agent="browser-use" data-model="claude-opus-4-7" data-browser="Lexmount" data-success="58.0" data-steps="14.20" data-e2e="205.8">
<tr data-rank="1" data-agent="browser-use" data-model="dmx-claude-opus-4-8-thinking" data-browser="Lexmount" data-success="75.3" data-steps="16.90" data-e2e="522.4">
<td class="num">1</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">claude-opus-4-7</code></td>
<td><code class="cell-mono">dmx-claude-opus-4-8-thinking</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>58.0</strong></td>
<td class="num">14.2</td>
<td class="num">205.8</td>
<td class="num"><strong>75.3</strong></td>
<td class="num">16.9</td>
<td class="num">522.4</td>
</tr>
<tr data-rank="2" data-agent="browser-use" data-model="kimi-k2.5" data-browser="Lexmount" data-success="58.0" data-steps="24.70" data-e2e="280.1">
<tr data-rank="2" data-agent="browser-use" data-model="gpt-5.5" data-browser="Lexmount" data-success="74.3" data-steps="14.00" data-e2e="212.0">
<td class="num">2</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">kimi-k2.5</code></td>
<td><code class="cell-mono">gpt-5.5</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>58.0</strong></td>
<td class="num">24.7</td>
<td class="num">280.1</td>
<td class="num"><strong>74.3</strong></td>
<td class="num">14.0</td>
<td class="num">212.0</td>
</tr>
<tr data-rank="3" data-agent="browser-use" data-model="gemini-3.1-pro-preview" data-browser="Lexmount" data-success="56.0" data-steps="13.90" data-e2e="149.0">
<tr data-rank="3" data-agent="browser-use" data-model="bu-2-0" data-browser="Lexmount" data-success="68.7" data-steps="17.50" data-e2e="323.8">
<td class="num">3</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">gemini-3.1-pro-preview</code></td>
<td><code class="cell-mono">bu-2-0</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>56.0</strong></td>
<td class="num">13.9</td>
<td class="num">149.0</td>
<td class="num"><strong>68.7</strong></td>
<td class="num">17.5</td>
<td class="num">323.8</td>
</tr>
<tr data-rank="4" data-agent="browser-use" data-model="gpt-5.5" data-browser="Lexmount" data-success="54.0" data-steps="14.10" data-e2e="273.0">
<tr data-rank="4" data-agent="browser-use" data-model="glm-5.1" data-browser="Lexmount" data-success="68.2" data-steps="22.00" data-e2e="589.7">
<td class="num">4</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">gpt-5.5</code></td>
<td><code class="cell-mono">glm-5.1</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>54.0</strong></td>
<td class="num">14.1</td>
<td class="num">273.0</td>
<td class="num"><strong>68.2</strong></td>
<td class="num">22.0</td>
<td class="num">589.7</td>
</tr>
<tr data-rank="5" data-agent="browser-use" data-model="gemini-3.1-pro-preview" data-browser="Chrome-Local" data-success="54.0" data-steps="14.00" data-e2e="163.3">
<tr data-rank="5" data-agent="browser-use" data-model="qwen3.7-max" data-browser="Lexmount" data-success="65.8" data-steps="18.80" data-e2e="487.6">
<td class="num">5</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">gemini-3.1-pro-preview</code></td>
<td><span class="browser-pill">Chrome-Local</span></td>
<td class="num"><strong>54.0</strong></td>
<td class="num">14.0</td>
<td class="num">163.3</td>
</tr>
<tr data-rank="6" data-agent="browser-use" data-model="kimi-k2.6" data-browser="Lexmount" data-success="52.0" data-steps="30.30" data-e2e="447.6">
<td class="num">6</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">kimi-k2.6</code></td>
<td><code class="cell-mono">qwen3.7-max</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>52.0</strong></td>
<td class="num">30.3</td>
<td class="num">447.6</td>
<td class="num"><strong>65.8</strong></td>
<td class="num">18.8</td>
<td class="num">487.6</td>
</tr>
<tr data-rank="7" data-agent="browser-use" data-model="bu-2-0" data-browser="Chrome-Local" data-success="48.0" data-steps="20.80" data-e2e="136.5">
<td class="num">7</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">bu-2-0</code></td>
<td><span class="browser-pill">Chrome-Local</span></td>
<td class="num"><strong>48.0</strong></td>
<td class="num">20.8</td>
<td class="num">136.5</td>
</tr>
<tr data-rank="8" data-agent="browser-use" data-model="MiniMax-M2.7" data-browser="Chrome-Local" data-success="42.0" data-steps="27.50" data-e2e="413.1">
<td class="num">8</td>
<tr data-rank="6" data-agent="browser-use" data-model="gemini-3.1-pro-preview" data-browser="Lexmount" data-success="63.9" data-steps="12.60" data-e2e="368.6">
<td class="num">6</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">MiniMax-M2.7</code></td>
<td><span class="browser-pill">Chrome-Local</span></td>
<td class="num"><strong>42.0</strong></td>
<td class="num">27.5</td>
<td class="num">413.1</td>
</tr>
<tr data-rank="9" data-agent="Agent-TARS" data-model="gemini-3.1-pro-preview" data-browser="Lexmount" data-success="40.0" data-steps="18.40" data-e2e="121.1">
<td class="num">9</td>
<td><span class="cell-strong">Agent-TARS</span></td>
<td><code class="cell-mono">gemini-3.1-pro-preview</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>40.0</strong></td>
<td class="num">18.4</td>
<td class="num">121.1</td>
<td class="num"><strong>63.9</strong></td>
<td class="num">12.6</td>
<td class="num">368.6</td>
</tr>
<tr data-rank="10" data-agent="browser-use" data-model="bu-2-0" data-browser="Lexmount" data-success="40.0" data-steps="23.40" data-e2e="350.7">
<td class="num">10</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">bu-2-0</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>40.0</strong></td>
<td class="num">23.4</td>
<td class="num">350.7</td>
</tr>
<tr data-rank="11" data-agent="browser-use" data-model="gemini-2.5-pro" data-browser="Lexmount" data-success="40.0" data-steps="18.70" data-e2e="279.2">
<td class="num">11</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">gemini-2.5-pro</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>40.0</strong></td>
<td class="num">18.7</td>
<td class="num">279.2</td>
</tr>
<tr data-rank="12" data-agent="browser-use" data-model="qwen3.5-plus" data-browser="Lexmount" data-success="40.0" data-steps="23.70" data-e2e="326.7">
<td class="num">12</td>
<tr data-rank="7" data-agent="browser-use" data-model="doubao-seed-2-0-pro" data-browser="Lexmount" data-success="61.8" data-steps="17.70" data-e2e="595.3">
<td class="num">7</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">qwen3.5-plus</code></td>
<td><code class="cell-mono">doubao-seed-2-0-pro</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>40.0</strong></td>
<td class="num">23.7</td>
<td class="num">326.7</td>
<td class="num"><strong>61.8</strong></td>
<td class="num">17.7</td>
<td class="num">595.3</td>
</tr>
<tr data-rank="13" data-agent="browser-use" data-model="MiniMax-M2.5" data-browser="Lexmount" data-success="38.0" data-steps="27.40" data-e2e="354.3">
<td class="num">13</td>
<tr data-rank="8" data-agent="browser-use" data-model="gemini-3.5-flash" data-browser="Lexmount" data-success="58.6" data-steps="26.90" data-e2e="376.4">
<td class="num">8</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">MiniMax-M2.5</code></td>
<td><code class="cell-mono">gemini-3.5-flash</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>38.0</strong></td>
<td class="num">27.4</td>
<td class="num">354.3</td>
<td class="num"><strong>58.6</strong></td>
<td class="num">26.9</td>
<td class="num">376.4</td>
</tr>
<tr data-rank="14" data-agent="browser-use" data-model="MiniMax-M2.7" data-browser="Lexmount" data-success="36.0" data-steps="22.40" data-e2e="408.9">
<td class="num">14</td>
<tr data-rank="9" data-agent="browser-use" data-model="kimi-k2.6" data-browser="Lexmount" data-success="53.3" data-steps="23.60" data-e2e="433.4">
<td class="num">9</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">MiniMax-M2.7</code></td>
<td><code class="cell-mono">kimi-k2.6</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>36.0</strong></td>
<td class="num">22.4</td>
<td class="num">408.9</td>
<td class="num"><strong>53.3</strong></td>
<td class="num">23.6</td>
<td class="num">433.4</td>
</tr>
<tr data-rank="15" data-agent="browser-use" data-model="doubao-seed-2-0-pro" data-browser="Lexmount" data-success="36.0" data-steps="17.30" data-e2e="385.3">
<td class="num">15</td>
<tr data-rank="10" data-agent="browser-use" data-model="MiniMax-M3" data-browser="Lexmount" data-success="40.1" data-steps="20.10" data-e2e="584.5">
<td class="num">10</td>
<td><span class="cell-strong">browser-use</span></td>
<td><code class="cell-mono">doubao-seed-2-0-pro</code></td>
<td><code class="cell-mono">MiniMax-M3</code></td>
<td><span class="browser-pill">Lexmount</span></td>
<td class="num"><strong>36.0</strong></td>
<td class="num">17.3</td>
<td class="num">385.3</td>
<td class="num"><strong>40.1</strong></td>
<td class="num">20.1</td>
<td class="num">584.5</td>
</tr>
</tbody>
</table>
</div>
<p class="muted small">
Live source: leaderboard server at the team intranet. Data is automatically pulled from
Snapshot source:
<code>experiments/{benchmark}/{split}/{agent}/{model_id}/{ts}/</code> run dirs.
</p>
</div>
</section>

<section id="citation" class="section section-alt">
<div class="container">
<h2 class="section-title">Cite</h2>
<p class="section-lede">If you use LexBench-Browser in your work, please cite:</p>
<div class="bibtex">
<button class="copy-btn" type="button" data-target="#bibtex-block">Copy</button>
<pre><code id="bibtex-block">@misc{lexbench_browser_2026,
title = {LexBench-Browser: A Real-World Browser Agent Benchmark with Long-Tail and Multilingual Tasks},
author = {Lexmount Research and Collaborators},
year = {2026},
howpublished = {\url{https://lexmount.github.io/browseruse-agent-bench/}},
note = {Open benchmark; v1.0 reference release}
}</code></pre>
</div>
<p class="muted small">
Acknowledgements: integration scaffolding adapted patterns from browser-use, skyvern,
Agent-TARS and deepbrowse upstream codebases.
</p>
</div>
</section>

<footer class="site-footer">
<div class="container footer-grid">
<div class="footer-col">
Expand Down
Loading