Add accuracy retention to the opening and the benchmark tables

README.md

@@ -12,11 +12,11 @@ tags:
 
 <img src="thinklite-banner.png?v=2" alt="Thinklite-27B: Qwen 3.8 27B made efficient" style="width:100%;max-width:100%;height:auto;display:block;margin-bottom:1em;" />
 
-**Qwen3.8-27B that thinks with half the tokens, with average accuracy within half a point.** At
+**Qwen3.8-27B that thinks with half the tokens and keeps 99.6% of its accuracy.** At
 `reasoning_effort="xhigh"`, Thinklite-27B uses 47% less reasoning than Qwen3.8-27B on average over six benchmarks and
-writes half as many output tokens, so its time-to-final-output (TTFO) is about 2x shorter, at an average accuracy of
-89.8 against the base model's 90.2. At `reasoning_effort="low"` it uses 28% less reasoning, at an average accuracy 2.2
-points below the base model.
+writes half as many output tokens, so its time-to-final-output (TTFO) is about 2x shorter, while it keeps 99.6% of
+the base model's average accuracy (89.8 against 90.2). At `reasoning_effort="low"` it uses 28% less reasoning and
+keeps 97.5% of the base model's accuracy (2.2 points below).
 
 <style>
 .tl-table { display:table !important; width:100% !important; table-layout:fixed !important; border-collapse:separate; border-spacing:0; overflow:hidden; border:1px solid rgba(246,246,243,0.16); border-radius:0; background:#090909; font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,sans-serif; font-size:13px; color:rgba(246,246,243,0.80); }
@@ -99,6 +99,8 @@ tokenizer and `reasoning_effort` control.
   and [ukisai/Swift-1.5-Qwen3.8-27b](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b) (revision `b4c84d42`).
 - vLLM 0.30.0, bf16 weights, FP8 KV cache, temperature 1.0, top_p 0.95, top_k 20, 64 concurrent requests, one NVIDIA
   RTX PRO 6000 Blackwell (96 GB). All answers were scored together on one machine.
+- Accuracy retention is a model's accuracy divided by Qwen3.8-27B's on the same benchmark; its Average divides
+  the average accuracies. Values above 100% mean the model scored higher than the base model.
 - One sample per question (two on AIME 2025). Answers that hit their cap were rerun once with a 131,072-token cap;
   truncated or unparseable answers count as incorrect.
 
@@ -121,6 +123,10 @@ tokenizer and `reasoning_effort` control.
     <tr><td class="name">ThinkingCap-Qwen3.8-27B</td><td>97.0</td><td><strong>100.0</strong></td><td>86.5</td><td>90.9</td><td><strong>84.7</strong></td><td><strong>89.0</strong></td><td>91.37</td></tr>
     <tr><td class="name">Swift-1.5-Qwen3.8-27b</td><td>96.9</td><td><strong>100.0</strong></td><td><strong>86.6</strong></td><td>90.9</td><td><strong>84.7</strong></td><td>87.6</td><td>91.12</td></tr>
     <tr><td class="name">Qwen3.8-27B</td><td><strong>97.2</strong></td><td>93.3</td><td>86.3</td><td><strong>91.5</strong></td><td>84.0</td><td>88.7</td><td>90.17</td></tr>
+    <tr><td class="section" colspan="8">Accuracy retention vs Qwen3.8-27B</td></tr>
+    <tr><td class="name tl">Thinklite-27B</td><td class="tl">99.6%</td><td class="tl">101.8%</td><td class="tl">97.8%</td><td class="tl">97.4%</td><td class="tl"><strong>100.9%</strong></td><td class="tl">100.2%</td><td class="tl">99.6%</td></tr>
+    <tr><td class="name">ThinkingCap-Qwen3.8-27B</td><td><strong>99.8%</strong></td><td><strong>107.1%</strong></td><td>100.2%</td><td><strong>99.4%</strong></td><td><strong>100.9%</strong></td><td><strong>100.4%</strong></td><td><strong>101.3%</strong></td></tr>
+    <tr><td class="name">Swift-1.5-Qwen3.8-27b</td><td>99.7%</td><td><strong>107.1%</strong></td><td><strong>100.3%</strong></td><td><strong>99.4%</strong></td><td><strong>100.9%</strong></td><td>98.7%</td><td>101.1%</td></tr>
     <tr><td class="section" colspan="8">Reasoning tokens vs Qwen3.8-27B</td></tr>
     <tr><td class="name tl">Thinklite-27B</td><td class="tl">&#8595; 63%</td><td class="tl"><strong>&#8595; 43%</strong></td><td class="tl"><strong>&#8595; 69%</strong></td><td class="tl">&#8595; 41%</td><td class="tl"><strong>&#8595; 38%</strong></td><td class="tl">&#8595; 28%</td><td class="tl">&#8595; 47.0%</td></tr>
     <tr><td class="name">ThinkingCap-Qwen3.8-27B</td><td><strong>&#8595; 68%</strong></td><td>&#8595; 33%</td><td>&#8595; 58%</td><td><strong>&#8595; 51%</strong></td><td>&#8595; 20%</td><td><strong>&#8595; 44%</strong></td><td>&#8595; 45.8%</td></tr>
@@ -156,6 +162,10 @@ tokenizer and `reasoning_effort` control.
     <tr><td class="name">ThinkingCap-Qwen3.8-27B</td><td>97.2</td><td>86.7</td><td><strong>84.8</strong></td><td>85.8</td><td>72.5</td><td><strong>91.5</strong></td><td>86.40</td></tr>
     <tr><td class="name">Swift-1.5-Qwen3.8-27b</td><td>97.1</td><td><strong>90.0</strong></td><td>84.7</td><td>86.0</td><td><strong>73.3</strong></td><td>91.2</td><td>87.04</td></tr>
     <tr><td class="name">Qwen3.8-27B</td><td><strong>97.4</strong></td><td><strong>90.0</strong></td><td>84.4</td><td><strong>86.1</strong></td><td>72.5</td><td>90.8</td><td>86.88</td></tr>
+    <tr><td class="section" colspan="8">Accuracy retention vs Qwen3.8-27B</td></tr>
+    <tr><td class="name tl">Thinklite-27B</td><td class="tl">99.3%</td><td class="tl">90.7%</td><td class="tl">98.7%</td><td class="tl">97.9%</td><td class="tl">97.9%</td><td class="tl">100.4%</td><td class="tl">97.5%</td></tr>
+    <tr><td class="name">ThinkingCap-Qwen3.8-27B</td><td><strong>99.8%</strong></td><td>96.3%</td><td><strong>100.5%</strong></td><td>99.6%</td><td>100.0%</td><td><strong>100.7%</strong></td><td>99.4%</td></tr>
+    <tr><td class="name">Swift-1.5-Qwen3.8-27b</td><td>99.7%</td><td><strong>100.0%</strong></td><td>100.4%</td><td><strong>99.8%</strong></td><td><strong>101.0%</strong></td><td>100.4%</td><td><strong>100.2%</strong></td></tr>
     <tr><td class="section" colspan="8">Reasoning tokens vs Qwen3.8-27B</td></tr>
     <tr><td class="name tl">Thinklite-27B</td><td class="tl"><strong>&#8595; 21%</strong></td><td class="tl"><strong>&#8595; 46%</strong></td><td class="tl"><strong>&#8595; 23%</strong></td><td class="tl"><strong>&#8595; 41%</strong></td><td class="tl"><strong>&#8595; 17%</strong></td><td class="tl"><strong>&#8595; 19%</strong></td><td class="tl">&#8595; 28.0%</td></tr>
     <tr><td class="name">ThinkingCap-Qwen3.8-27B</td><td>&#8595; 12%</td><td>&#8595; 23%</td><td>&#8595; 18%</td><td>&#8595; 19%</td><td>&#8595; 15%</td><td>&#8595; 12%</td><td>&#8595; 16.4%</td></tr>