feat(leaderboard): add Qwen results and time/token costs

rpent/benchmarks/docs.css

@@ -8,3 +8,5 @@
 .rpent-static-leaderboard .table-wrap table{min-width:900px}
 html[data-theme="dark"] .rpent-static-leaderboard{--rpent-fallback-border:#494052;color:var(--pst-color-text-base,#f4f0fa)}
 .rpent-static-leaderboard .memory-context{margin:1rem 0}
+.rpent-static-leaderboard .cost-section .table-wrap table{min-width:340px}
+.rpent-static-leaderboard .cost-method small{display:block;font-size:12px;margin-top:4px}

rpent/benchmarks/embed.html

@@ -6,10 +6,11 @@
       <div class="header-links"><a href="https://github.com/RLinf/RPent" target="_blank" rel="noopener">GitHub</a><a href="https://arxiv.org/abs/2607.08448v4" target="_blank" rel="noopener" data-i18n="paper">论文</a><button id="language" type="button" lang="en" aria-label="Switch to English">English</button></div>
     </div>
     <div class="title-row"><h1 data-i18n="title">RPent Leaderboard</h1></div>
-    <nav class="bench-nav" aria-label="Leaderboards"><a href="#libero-pro">LIBERO-PRO</a><a href="#standard-libero">LIBERO</a><a href="#robocasa">RoboCasa365</a><a href="#robotwin">RoboTwin</a></nav>
+    <nav class="bench-nav" aria-label="Leaderboards"><a href="#libero-pro">LIBERO-PRO</a><a href="#standard-libero">LIBERO</a><a href="#robocasa">RoboCasa365</a><a href="#robotwin">RoboTwin</a><a href="#time-token-costs">Time &amp; Token Costs</a></nav>
   </header>
   <main id="main">
     <div id="leaderboards"></div>
+    <section id="time-token-costs" class="cost-section"></section>
   </main>
   <footer><span>RPent Leaderboard</span><a href="#top" data-i18n="backTop">返回顶部</a></footer>
   <div id="chart-tooltip" class="chart-tooltip" role="tooltip" hidden></div>

rpent/benchmarks/leaderboard.js

@@ -32,10 +32,12 @@
     "backTop": "Back to top ↑",
     "view": "Evaluation",
     "rate": "Success rate",
-    "method": "Method / model",
+    "method": "Method",
+    "meanTime": "Mean time / episode (s)",
+    "outputTokens": "Total output tokens",
     "unreported": "Not reported",
     "modelUnknown": "Model not reported",
-    "noReasoning": "no reasoning",
+    "noReasoning": "no-reasoning",
     "table": "Results table",
     "download": "Download CSV",
     "print": "Print page",
@@ -50,10 +52,12 @@
     "backTop": "返回顶部 ↑",
     "view": "评测范围",
     "rate": "成功率",
-    "method": "方法 / 模型",
+    "method": "方法",
+    "meanTime": "平均每回合耗时（秒）",
+    "outputTokens": "总输出 token",
     "unreported": "未报告",
     "modelUnknown": "未报告模型",
-    "noReasoning": "无推理",
+    "noReasoning": "no-reasoning",
     "table": "结果明细表",
     "download": "下载 CSV",
     "print": "打印页面",
@@ -117,7 +121,7 @@
     const rows=d.configurations.filter(c=>columns.some(v=>v.record_ids.some(id=>records.get(id).configuration_id===c.id&&records.get(id).rate!==null)));
     return `<details class="comparison-matrix" data-details="${s.id}"><summary>${t('allScores')} <span>${rows.length} ${lang==='en'?'configurations':'配置'}</span></summary><div class="table-wrap" tabindex="0" role="region" aria-label="${h(s.name+' '+t('allScores'))}"><table><thead><tr><th>${t('method')}</th>${columns.map(v=>`<th class="rate">${h(tr(v.label))}</th>`).join('')}</tr></thead><tbody>${rows.map(c=>{
       const sample=columns.flatMap(allRows).find(r=>r.configuration_id===c.id&&r.rate!==null);
-      return `<tr data-method="${c.id}"><th scope="row"><div class="matrix-label"><span>${h(label(sample))}<small>${h(detailLabel(sample))}</small></span></div></th>${columns.map(v=>{
+      return `<tr data-method="${c.id}"><th scope="row"><div class="matrix-label"><span>${h(label(sample))}<small> ${h(detailLabel(sample))}</small></span></div></th>${columns.map(v=>{
         const r=allRows(v).find(r=>r.configuration_id===c.id);
         return `<td class="rate"${r?` data-matrix-record="${r.id}"`:''} title="${h(tr(r?.evaluation_note)||tr(v.label))}">${r?.rate!=null?r.rate+'%':`<span class="not-reported" aria-label="${t('unreported')}">—</span>`}</td>`;
       }).join('')}</tr>`;
@@ -136,13 +140,24 @@
   function setupTables(element,key) {
     window.RPentTables.setup(element,{key,language:lang,icon:asset('arrow-down.svg'),download:asset('download.svg')});
   }
+  function renderCosts() {
+    const element=document.getElementById('time-token-costs');
+    const groups=[['libero-pro','LIBERO-PRO'],['robocasa','RoboCasa365'],['robotwin','RoboTwin']];
+    element.innerHTML=`<div class="section-inner"><h2>Time &amp; Token Costs</h2>${groups.map(([id,name])=>{
+      const rows=d.cost_results.filter(r=>r.benchmark_id===id);
+      return `<section class="cost-group" data-cost-group="${id}"><h3>${name}</h3><div class="table-wrap" tabindex="0" role="region" aria-label="${name} Time &amp; Token Costs"><table data-sort-key="costs-${id}" data-sort-direction="ascending" data-highlight-best="false"><thead><tr><th>${t('method')}</th><th class="rate">${t('meanTime')}</th><th class="rate">${t('outputTokens')}</th></tr></thead><tbody>${rows.map(r=>{
+        return `<tr data-cost-record="${r.id}" data-method="${r.configuration_id}"><td><div class="matrix-label"><span>${h(label(r))}<small> ${h(detailLabel(r))}</small></span></div></td><td class="rate" data-value="${r.mean_elapsed_seconds}">${r.mean_elapsed_seconds.toLocaleString('en-US')}</td><td class="rate" data-value="${r.total_output_tokens}">${r.total_output_tokens.toLocaleString('en-US')}</td></tr>`;
+      }).join('')}</tbody></table></div></section>`;
+    }).join('')}</div>`;
+    setupTables(element,'costs');
+  }
   function render() {
     (options.embedded?document.host:document.documentElement).lang=lang==='en'?'en':'zh-CN';
     if(!options.embedded)document.title='RPent Leaderboard';
     document.querySelectorAll('[data-i18n]').forEach(el=>el.textContent=t(el.dataset.i18n));
     const language=document.getElementById('language');language.textContent=lang==='en'?'中文':'English';language.lang=lang==='en'?'zh-CN':'en';language.setAttribute('aria-label',lang==='en'?'切换为中文':'Switch to English');
     document.getElementById('leaderboards').innerHTML=d.sections.filter(s=>!s.parent).map(s=>`<section id="${s.id}" class="benchmark-band"><div class="benchmark-shell"><h2 class="benchmark-wordmark">${h(s.name)}</h2><div id="panel-${s.id}"></div>${d.sections.filter(c=>c.parent===s.id).map(c=>`<section id="${c.id}" class="benchmark-subsection"><div id="panel-${c.id}"></div></section>`).join('')}</div></section>`).join('');
-    d.sections.forEach(renderSection);hideTooltip();
+    d.sections.forEach(renderSection);renderCosts();hideTooltip();
   }
   function downloadCSV(rows, name) {
     const keys=['record_id','benchmark','view','method','model','planner','perception_model','reasoning','effort','success_rate_percent','successes','episodes','status','evaluation_note'];