add benchmark table
Browse files
README.md
CHANGED
|
@@ -23,6 +23,304 @@ pipeline_tag: image-text-to-text
|
|
| 23 |
|
| 24 |
**ZwZ-4B** is a fine-grained multimodal perception model built upon [Qwen3-VL-4B](https://huggingface.co/Qwen/Qwen3-VL-4B). It is trained using **Region-to-Image Distillation (R2I)** combined with reinforcement learning, enabling superior fine-grained visual understanding in a single forward pass — no inference-time zooming or tool calling required. ZwZ-4B achieves state-of-the-art performance on fine-grained perception benchmarks among open-source models of comparable size.
|
| 25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
<div align=center>
|
| 27 |
<img src="gp_avg_comparison.png" width="90%" alt="avg_comparison"/>
|
| 28 |
</div>
|
|
|
|
| 23 |
|
| 24 |
**ZwZ-4B** is a fine-grained multimodal perception model built upon [Qwen3-VL-4B](https://huggingface.co/Qwen/Qwen3-VL-4B). It is trained using **Region-to-Image Distillation (R2I)** combined with reinforcement learning, enabling superior fine-grained visual understanding in a single forward pass — no inference-time zooming or tool calling required. ZwZ-4B achieves state-of-the-art performance on fine-grained perception benchmarks among open-source models of comparable size.
|
| 25 |
|
| 26 |
+
<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;max-width:780px;margin:0 auto;padding:12px 0">
|
| 27 |
+
|
| 28 |
+
<p style="font-size:11px;color:#555;margin-bottom:10px">
|
| 29 |
+
<table style="width:100%;border-collapse:collapse;font-size:10px">
|
| 30 |
+
<thead>
|
| 31 |
+
<tr>
|
| 32 |
+
<th style="padding:6px 4px;text-align:left;font-weight:600;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:10px" rowspan="2">Models</th>
|
| 33 |
+
<th style="padding:6px 4px;text-align:center;font-weight:600;border-bottom:1px solid #7c3aed;color:#7c3aed;font-size:10px" colspan="8">General Perception</th>
|
| 34 |
+
<th style="padding:6px 4px;text-align:center;font-weight:600;border-bottom:1px solid #7c3aed;color:#7c3aed;font-size:10px" colspan="2">Specific Perception</th>
|
| 35 |
+
<th style="padding:6px 4px;text-align:center;font-weight:600;border-bottom:1px solid #7c3aed;color:#7c3aed;font-size:10px" colspan="2">OOD Generalization</th>
|
| 36 |
+
<th style="padding:6px 4px;text-align:center;font-weight:600;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:10px" rowspan="2">Avg</th>
|
| 37 |
+
</tr>
|
| 38 |
+
<tr>
|
| 39 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">ZoomBench</th>
|
| 40 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">HR-4K</th>
|
| 41 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">HR-8K</th>
|
| 42 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">VStar</th>
|
| 43 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">CV-B.</th>
|
| 44 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">MME-RW-en</th>
|
| 45 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">MME-RW-cn</th>
|
| 46 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px;background:rgba(0,180,0,0.1)">GP-Avg</th>
|
| 47 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">CountQA</th>
|
| 48 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">ColorB.</th>
|
| 49 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">MMStar</th>
|
| 50 |
+
<th style="padding:5px 4px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size:9px">BabyVision</th>
|
| 51 |
+
</tr>
|
| 52 |
+
</thead>
|
| 53 |
+
<tbody>
|
| 54 |
+
|
| 55 |
+
<!-- Closed-Source Models -->
|
| 56 |
+
<tr><td colspan="14" style="padding:5px 8px;font-weight:600;font-size:9px;color:#7c3aed;border-bottom:1px solid rgba(124,58,237,0.2);background:rgba(124,58,237,0.1);font-style:italic">Closed-Source Models</td></tr>
|
| 57 |
+
|
| 58 |
+
<tr style="background:rgba(255,240,150,0.15)">
|
| 59 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">GPT-5.1</td>
|
| 60 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">47.22</td>
|
| 61 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">67.00</td>
|
| 62 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">65.25</td>
|
| 63 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">70.16</td>
|
| 64 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">84.22</td>
|
| 65 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">64.04</td>
|
| 66 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">55.57</td>
|
| 67 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">64.78</td>
|
| 68 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">31.41</td>
|
| 69 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">83.43</td>
|
| 70 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">71.60</td>
|
| 71 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">13.92</td>
|
| 72 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">59.44</td>
|
| 73 |
+
</tr>
|
| 74 |
+
|
| 75 |
+
<tr style="background:rgba(255,240,150,0.15)">
|
| 76 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Gemini-3-Flash</td>
|
| 77 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">59.29</td>
|
| 78 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">87.88</td>
|
| 79 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">85.00</td>
|
| 80 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">86.39</td>
|
| 81 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">89.57</td>
|
| 82 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">74.86</td>
|
| 83 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">72.62</td>
|
| 84 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">79.37</td>
|
| 85 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">66.88</td>
|
| 86 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">85.47</td>
|
| 87 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">83.60</td>
|
| 88 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">34.51</td>
|
| 89 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">75.10</td>
|
| 90 |
+
</tr>
|
| 91 |
+
|
| 92 |
+
<!-- Open-Source Models -->
|
| 93 |
+
<tr><td colspan="14" style="padding:5px 8px;font-weight:600;font-size:9px;color:#7c3aed;border-bottom:1px solid rgba(124,58,237,0.2);background:rgba(124,58,237,0.1);font-style:italic">Open-Source Models</td></tr>
|
| 94 |
+
|
| 95 |
+
<tr style="background:rgba(180,180,180,0.08)">
|
| 96 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Qwen3-VL-2B</td>
|
| 97 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">41.30</td>
|
| 98 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">71.75</td>
|
| 99 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">70.12</td>
|
| 100 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">72.77</td>
|
| 101 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">78.94</td>
|
| 102 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">59.52</td>
|
| 103 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">60.77</td>
|
| 104 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">65.02</td>
|
| 105 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">22.19</td>
|
| 106 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">76.86</td>
|
| 107 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">60.4</td>
|
| 108 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">12.11</td>
|
| 109 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">56.98</td>
|
| 110 |
+
</tr>
|
| 111 |
+
|
| 112 |
+
<tr style="background:rgba(180,180,180,0.08)">
|
| 113 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Qwen3-VL-4B</td>
|
| 114 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">40.24</td>
|
| 115 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">78.25</td>
|
| 116 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">72.88</td>
|
| 117 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">80.10</td>
|
| 118 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">84.95</td>
|
| 119 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">63.47</td>
|
| 120 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">63.63</td>
|
| 121 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">69.07</td>
|
| 122 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">28.14</td>
|
| 123 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">81.63</td>
|
| 124 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">69.73</td>
|
| 125 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">13.66</td>
|
| 126 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">61.52</td>
|
| 127 |
+
</tr>
|
| 128 |
+
|
| 129 |
+
<tr style="background:rgba(180,180,180,0.08)">
|
| 130 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Qwen2.5-VL-7B</td>
|
| 131 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">42.49</td>
|
| 132 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">71.62</td>
|
| 133 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">67.88</td>
|
| 134 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">78.53</td>
|
| 135 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">75.34</td>
|
| 136 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">60.80</td>
|
| 137 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">58.30</td>
|
| 138 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">64.99</td>
|
| 139 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">18.91</td>
|
| 140 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">76.36</td>
|
| 141 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">61.93</td>
|
| 142 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">12.89</td>
|
| 143 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">56.82</td>
|
| 144 |
+
</tr>
|
| 145 |
+
|
| 146 |
+
<tr style="background:rgba(180,180,180,0.08)">
|
| 147 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Qwen3-VL-8B</td>
|
| 148 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">37.87</td>
|
| 149 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">78.88</td>
|
| 150 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">74.63</td>
|
| 151 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">86.39</td>
|
| 152 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">85.44</td>
|
| 153 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">65.96</td>
|
| 154 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">66.67</td>
|
| 155 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">70.83</td>
|
| 156 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">28.99</td>
|
| 157 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">82.77</td>
|
| 158 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">70.93</td>
|
| 159 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">12.89</td>
|
| 160 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">62.86</td>
|
| 161 |
+
</tr>
|
| 162 |
+
|
| 163 |
+
<tr>
|
| 164 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">MiMo-VL-7B-RL</td>
|
| 165 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">45.09</td>
|
| 166 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">74.38</td>
|
| 167 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">72.88</td>
|
| 168 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">81.15</td>
|
| 169 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">84.31</td>
|
| 170 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">63.40</td>
|
| 171 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">59.78</td>
|
| 172 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">68.71</td>
|
| 173 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">28.27</td>
|
| 174 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">82.80</td>
|
| 175 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">73.53</td>
|
| 176 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">16.24</td>
|
| 177 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">61.98</td>
|
| 178 |
+
</tr>
|
| 179 |
+
|
| 180 |
+
<tr>
|
| 181 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">MiniCPM-V-4.5 (9B)</td>
|
| 182 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">42.60</td>
|
| 183 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">69.88</td>
|
| 184 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">63.62</td>
|
| 185 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">70.16</td>
|
| 186 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">80.25</td>
|
| 187 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">58.16</td>
|
| 188 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">56.23</td>
|
| 189 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">62.99</td>
|
| 190 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">23.43</td>
|
| 191 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">79.75</td>
|
| 192 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">67.87</td>
|
| 193 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">14.95</td>
|
| 194 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">56.99</td>
|
| 195 |
+
</tr>
|
| 196 |
+
|
| 197 |
+
<tr>
|
| 198 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">GLM-4.5V (108B)</td>
|
| 199 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">49.23</td>
|
| 200 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">81.63</td>
|
| 201 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">74.88</td>
|
| 202 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">83.25</td>
|
| 203 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">87.59</td>
|
| 204 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">66.04</td>
|
| 205 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">60.71</td>
|
| 206 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">71.90</td>
|
| 207 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">35.93</td>
|
| 208 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">84.59</td>
|
| 209 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">75.87</td>
|
| 210 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">15.72</td>
|
| 211 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">65.04</td>
|
| 212 |
+
</tr>
|
| 213 |
+
|
| 214 |
+
<tr>
|
| 215 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Qwen3-VL-235B-A22B</td>
|
| 216 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">49.11</td>
|
| 217 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>84.50</strong></td>
|
| 218 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>81.62</u></td>
|
| 219 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">87.96</td>
|
| 220 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">86.72</td>
|
| 221 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">67.07</td>
|
| 222 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">65.29</td>
|
| 223 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">74.61</td>
|
| 224 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>40.58</u></td>
|
| 225 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>85.62</u></td>
|
| 226 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>76.33</u></td>
|
| 227 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>18.30</u></td>
|
| 228 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">67.55</td>
|
| 229 |
+
</tr>
|
| 230 |
+
|
| 231 |
+
<tr>
|
| 232 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">Kimi-K2.5 (1T)</td>
|
| 233 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>56.33</u></td>
|
| 234 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">81.87</td>
|
| 235 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">75.38</td>
|
| 236 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">85.86</td>
|
| 237 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>89.18</strong></td>
|
| 238 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>71.51</strong></td>
|
| 239 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>68.40</u></td>
|
| 240 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">75.50</td>
|
| 241 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>52.81</strong></td>
|
| 242 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>86.61</strong></td>
|
| 243 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>81.80</strong></td>
|
| 244 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>33.25</strong></td>
|
| 245 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>71.18</strong></td>
|
| 246 |
+
</tr>
|
| 247 |
+
|
| 248 |
+
<!-- Our Models -->
|
| 249 |
+
<tr><td colspan="14" style="padding:5px 8px;font-weight:600;font-size:9px;color:#7c3aed;border-bottom:1px solid rgba(124,58,237,0.2);background:rgba(124,58,237,0.1);font-style:italic">Our Models</td></tr>
|
| 250 |
+
|
| 251 |
+
<tr style="background:rgba(100,130,255,0.06)">
|
| 252 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>ZwZ-2B (Ours)</strong></td>
|
| 253 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">53.49</td>
|
| 254 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">77.00</td>
|
| 255 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">75.38</td>
|
| 256 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">82.72</td>
|
| 257 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">83.36</td>
|
| 258 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">65.61</td>
|
| 259 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">65.39</td>
|
| 260 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">71.85</td>
|
| 261 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">21.60</td>
|
| 262 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">79.37</td>
|
| 263 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">63.40</td>
|
| 264 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">17.78</td>
|
| 265 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">62.28</td>
|
| 266 |
+
</tr>
|
| 267 |
+
|
| 268 |
+
<tr style="background:rgba(100,130,255,0.06)">
|
| 269 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>ZwZ-4B (Ours)</strong></td>
|
| 270 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">55.74</td>
|
| 271 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">81.75</td>
|
| 272 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">79.50</td>
|
| 273 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>92.67</strong></td>
|
| 274 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><u>87.90</u></td>
|
| 275 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">68.52</td>
|
| 276 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">68.09</td>
|
| 277 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)"><u>76.31</u></td>
|
| 278 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">30.82</td>
|
| 279 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">83.08</td>
|
| 280 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">71.13</td>
|
| 281 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">16.24</td>
|
| 282 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">66.86</td>
|
| 283 |
+
</tr>
|
| 284 |
+
|
| 285 |
+
<tr style="background:rgba(100,130,255,0.06)">
|
| 286 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px"><strong>ZwZ-7B (Ours)</strong></td>
|
| 287 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">55.62</td>
|
| 288 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">75.38</td>
|
| 289 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">73.25</td>
|
| 290 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">88.48</td>
|
| 291 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">79.83</td>
|
| 292 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">66.21</td>
|
| 293 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">66.96</td>
|
| 294 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px;background:rgba(0,180,0,0.08)">72.25</td>
|
| 295 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">20.72</td>
|
| 296 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">80.82</td>
|
| 297 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">63.40</td>
|
| 298 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">15.98</td>
|
| 299 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-size:10px">62.42</td>
|
| 300 |
+
</tr>
|
| 301 |
+
|
| 302 |
+
<tr style="background:rgba(100,130,255,0.06)">
|
| 303 |
+
<td style="padding:4px 4px;padding-left:14px;border-bottom:2px solid #7c3aed;font-size:10px"><strong>ZwZ-8B (Ours)</strong></td>
|
| 304 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><strong>58.11</strong></td>
|
| 305 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><u>84.38</u></td>
|
| 306 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><strong>82.00</strong></td>
|
| 307 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><u>91.10</u></td>
|
| 308 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px">87.40</td>
|
| 309 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><u>69.87</u></td>
|
| 310 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><strong>70.59</strong></td>
|
| 311 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px;background:rgba(0,180,0,0.08)"><strong>77.64</strong></td>
|
| 312 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px">32.40</td>
|
| 313 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px">83.59</td>
|
| 314 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px">73.13</td>
|
| 315 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px">16.75</td>
|
| 316 |
+
<td style="padding:4px 4px;text-align:center;border-bottom:2px solid #7c3aed;font-size:10px"><u>68.12</u></td>
|
| 317 |
+
</tr>
|
| 318 |
+
|
| 319 |
+
</tbody>
|
| 320 |
+
</table>
|
| 321 |
+
</div>
|
| 322 |
+
|
| 323 |
+
|
| 324 |
<div align=center>
|
| 325 |
<img src="gp_avg_comparison.png" width="90%" alt="avg_comparison"/>
|
| 326 |
</div>
|