-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
375 lines (352 loc) · 36.9 KB
/
index.html
File metadata and controls
375 lines (352 loc) · 36.9 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8" />
<title>Chart2Code</title>
<meta name="description" content="From Charts to Code: A Hierarchical Benchmark for Multimodal Models" />
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1" />
<meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1, user-scalable=no" />
<link rel="icon" href="https://github.com/CSU-JPG/Chart2Code.github.io/assets/logo_2.png" type="image/png">
<link rel="shortcut icon" href="favicon_mm.ico" type="image/x-icon" />
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Lobster&display=swap" rel="stylesheet">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Nunito:ital@0;1&family=Pacifico&family=Poppins:wght@500&display=swap" rel="stylesheet">
<link href="https://fonts.googleapis.com/css2?family=Playfair+Display:ital,wght@0,400..900;1,400..900&family=Raleway:wght@400;500&family=Lora:ital@0;1&display=swap" rel="stylesheet">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.5.2/css/all.min.css" integrity="sha512-SnH5WK+bZxgPHs44uWIX+LLJAJ9/2PkPKZ5QiAj6Ta86w+fsb2TkcmfRyVX3pBnMFcV7oQPJkl9QevSCWr3W6A==" crossorigin="anonymous" referrerpolicy="no-referrer" />
<link rel="stylesheet" href="css/normalize.css" />
<link rel="stylesheet" href="css/fonts.css" />
<link rel="stylesheet" href="css/styles.css" />
<link rel="stylesheet" href="css/leaderboard.css" />
<script async src="https://www.googletagmanager.com/gtag/js?id=G-H9XFCMDPNS"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() { dataLayer.push(arguments); }
gtag("js", new Date());
gtag("config", "G-H9XFCMDPNS");
</script>
</head>
<body>
<div class="page-container">
<section class="title-section">
<div class="content-wrapper title-wrapper">
<div class="title-header">
<h1 class="chart2code-title">
<span class="word-chart">Chart</span><span class="word-2">2</span><span class="word-code">Code</span>
</h1>
</div>
<h2 class="subtitle">From Charts to Code: A Hierarchical Benchmark for Multimodal Models</h2>
<div class="authors">
<span class="author-block"><a href="#">Jiahao Tang</a><sup>1</sup>,</span>
<span class="author-block"><a href="https://zhaohengyuan1.github.io/">Henry Hengyuan Zhao</a><sup>2</sup>,</span>
<span class="author-block"><a href="#">Lijian Wu</a><sup>1</sup>,</span>
<span class="author-block"><a href="#">Yifei Tao</a><sup>3</sup>,</span>
<span class="author-block"><a href="https://scholar.google.com/citations?user=RLVSYY0AAAAJ&hl=en">Dongxing Mao</a><sup>1</sup>,</span><br>
<span class="author-block"><a href="#">Yang Wan</a><sup>1</sup>,</span>
<span class="author-block"><a href="https://scholar.google.com/citations?user=l18d7kcAAAAJ&hl=en">Jingru Tan</a><sup>1</sup>,</span>
<span class="author-block"><a href="https://minzeng1990.github.io/">Min Zeng</a><sup>1</sup>,</span>
<span class="author-block"><a href="https://scholar.google.com.hk/citations?user=w47WJE4AAAAJ&hl=en">Min Li</a><sup>1</sup>,</span>
<span class="author-block"><a href="https://fingerrec.github.io/">Alex Jinpeng Wang</a><sup>1</sup></span>
</div>
<div class="affiliations">
<span class="author-block"><a href="https://github.com/CSU-JPG"><sup>1</sup>CSU-JPG, </a>Central South University,</span>
<span class="author-block"><sup>2</sup>National University of Singapore,</span>
<span class="author-block"><sup>3</sup>Nanyang Technological University,</span>
</div>
<div class="button-group">
<a href="https://arxiv.org/abs/2510.17932"><button class="outline multimodal"><i class="fas fa-file-pdf"></i> Paper </button></a>
<a href="https://github.com/CSU-JPG/Chart2Code"><button class="outline multimodal"><i class="fa-brands fa-github"></i> Code </button></a>
<a href="https://huggingface.co/datasets/CSU-JPG/Chart2Code"><button class="outline multimodal"><i class="fas fa-images"></i> Data </button></a>
</div>
</div>
</section>
<section class="main-container">
<div class="content-wrapper">
<div class="content-box">
<h2 class="text-title">What's new with Chart2Code benchmark</h2>
<p class="text-content">TL;DR: We introduce Chart2Code, a new benchmark designed to evaluate chart generation capabilities of LMMs under progressively challenging conditions. There are five tasks in the Chart2Code benchmark.</p>
<div class="leaderboard-container">
<div class="tabs-container">
<div class="tab-links">
<button class="tab-link active" data-tab="dr-sample">Level 1: Direct Reproduction</button>
<button class="tab-link" data-tab="crd-sample">Level 1: Custom Raw Darta</button>
<button class="tab-link" data-tab="cfd-sample">Level 1: Custom Figure Data</button>
<button class="tab-link" data-tab="level2-sample">Level 2</button>
<button class="tab-link" data-tab="level3-sample">Level 3</button>
</div>
<div class="tab-content">
<div id="dr-sample" class="tab-pane active">
<img src="./assets/illustration_DR.png" alt="Direct Reproduction Example" class="tab-image">
</div>
<div id="crd-sample" class="tab-pane">
<img src="./assets/illustration_CRD.png" alt="Customized Raw Data Example" class="tab-image">
</div>
<div id="cfd-sample" class="tab-pane">
<img src="./assets/illustration_CFD.png" alt="Customized Figure Data Example" class="tab-image">
</div>
<div id="level2-sample" class="tab-pane">
<img src="./assets/illustration_level2.png" alt="Level 2 Example" class="tab-image">
</div>
<div id="level3-sample" class="tab-pane">
<img src="./assets/illustration_level3.png" alt="Level 3 Example" class="tab-image">
</div>
</div>
</div>
</div>
<h2 class="text-title">Overview</h2>
<p class="text-content">Chart2Code is a new benchmark for evaluating the chart understanding and code generation capabilities of large multimodal models. Chart2Code is explicitly designed from a user-driven perspective, capturing diverse real-world scenarios and progressively increasing task difficulty.</p>
<img src="./assets/figure1.png" class="responsive-image" alt="Chart2Code Overview"/>
<p class="text-content">To our knowledge, Chart2Code is the <b>first hierarchical benchmark</b> that reflects practical chart2code usage while systematically scaling task complexity. It consists of three levels as illustrated in above figure:
<br>Level1(<b>Chart Reproduction</b>) reproduces charts from a reference figure and user query;
<br>Level2(<b>Chart Editing</b>) involves complex modifications such as changing chart types or adding elements;
<br>Level3(<b>Long-Table to Chart Generation</b>) requires models to transform long, information-dense tables into faithful charts following user instructions.
</p>
<h2 class="text-title">Data Statistic</h2>
<img src="./assets/figure2.png" class="responsive-image" alt="Data Statistics"/>
<p class="text-content">In total, Chart2Code contains 2,023 tasks across 22 chart types, paired with multi-level evaluation metrics that assess both code correctness and the visual fidelity of rendered charts.</p>
<h2 class="text-title">Benchmark Comparison</h2>
<img src="./assets/figure3.png" class="responsive-image" alt="Benchmark Comparison Table"/>
<p class="text-content text-center">Table 1: Chart2Code is a unique benchmark featuring a more comprehensive set of tasks that better reflect real-world scenarios.</p>
<p class="text-content"><b>Comparison of existing chart-to-code benchmarks.</b>: Chart2Code is a new benchmark designed to rigorously evaluate chart generation capabilities of LMMs under progressively challenging conditions. This hierarchical design reflects real-world usage while progressively increasing difficulty, and its distinctions from prior benchmarks are highlighted in the above table.</p>
<h2 class="text-title">Human-Model Performance Comparison</h2>
<div class="leaderboard-container">
<div class="tabs-container">
<div class="tab-links">
<button class="tab-link active" data-tab="dr-sample">Level 1: Direct Reproduction</button>
<button class="tab-link" data-tab="crd-sample">Level 1: Custom Raw Darta</button>
<button class="tab-link" data-tab="cfd-sample">Level 1: Custom Figure Data</button>
<button class="tab-link" data-tab="level2-sample">Level 2 Example</button>
<button class="tab-link" data-tab="level3-sample">Level 3 Example</button>
</div>
<div class="tab-content">
<div id="dr-sample" class="tab-pane active">
<img src="assets/human_model_comparison_DR.png" alt="Direct Reproduction Example" class="tab-image">
</div>
<div id="crd-sample" class="tab-pane">
<img src="assets/human_model_comparison_CRD.png" alt="Customized Raw Data Example" class="tab-image">
</div>
<div id="cfd-sample" class="tab-pane">
<img src="assets/human_model_comparison_CFD.png" alt="Customized Figure Data Example" class="tab-image">
</div>
<div id="level2-sample" class="tab-pane">
<img src="assets/human_model_comparison_level2.png" alt="Level 2 Example" class="tab-image">
</div>
<div id="level3-sample" class="tab-pane">
<img src="assets/human_model_comparison_level3.png" alt="Level 3 Example" class="tab-image">
</div>
</div>
</div>
</div>
<div class="leaderboard-container">
<div class="tabs-container">
<div class="tab-links">
<button class="tab-link active" data-tab="level1">Level 1: Chart Reproduction</button>
<button class="tab-link" data-tab="level2">Level 2: Chart Editing</button>
<button class="tab-link" data-tab="level3">Level 3: Long-Table to Chart</button>
</div>
<div class="tab-content">
<div id="level1" class="tab-pane active">
<table class="sortable-table">
<thead>
<tr>
<th rowspan="2"><strong>Model</strong></th>
<th colspan="3"><strong>Direct Reproduction(DR)</strong></th>
<th colspan="3"><strong>Customize Raw Data(CRD)</strong></th>
<th colspan="3"><strong>Customize Figure Data(CFD)</strong></th>
</tr>
<tr>
<th><strong>Exec.Rate</strong></th>
<th><strong>LLM-Score</strong></th>
<th><strong>LMM-Score</strong></th>
<th><strong>Exec.Rate</strong></th>
<th><strong>LLM-Score</strong></th>
<th><strong>LMM-Score</strong></th>
<th><strong>Exec.Rate</strong></th>
<th><strong>LLM-Score</strong></th>
<th><strong>LMM-Score</strong></th>
</tr>
</thead>
<tbody>
<tr class="group-header"><td colspan="10"><em><strong>Proprietary</strong></em></td></tr>
<tr><td>Gemini-2.5-Pro</td><td>90.4</td><td>0.6286</td><td><strong>0.3807</strong></td><td><strong>100</strong></td><td><strong>0.6763</strong></td><td>0.2661</td><td>87.04</td><td><strong>0.6145</strong></td><td>0.2214</td></tr>
<tr><td>Claude-Sonnet-4</td><td><strong>96.38</strong></td><td>0.5629</td><td>0.2553</td><td>97.2</td><td>0.4878</td><td>0.236</td><td>88.89</td><td>0.5538</td><td>0.2273</td></tr>
<tr><td>GPT-5</td><td>87.48</td><td>0.6334</td><td>0.3575</td><td>94.4</td><td>0.6070</td><td>0.2238</td><td>85.19</td><td>0.6082</td><td>0.2382</td></tr>
<tr><td>Seed-1.5-VL</td><td>85.81</td><td>0.5536</td><td>0.2341</td><td>97.2</td><td>0.6325</td><td>0.2662</td><td>65.74</td><td>0.5756</td><td>0.1962</td></tr>
<tr><td>Seed-1.6-VL</td><td>84.70</td><td>0.5237</td><td>0.2301</td><td>94.4</td><td>0.6525</td><td>0.2503</td><td>83.96</td><td>0.5978</td><td>0.2075</td></tr>
<tr class="group-header"><td colspan="10"><em><strong>Open-Source LMMs (non-thinking)</strong></em></td></tr>
<tr><td>LLaVA-OV-Qwen2-7B-SI</td><td>32.82</td><td>0.1820</td><td>0.0154</td><td>11.11</td><td>0.4225</td><td>0.1550</td><td>0</td><td>-</td><td>-</td></tr>
<tr><td>LLaVA-OV-Qwen2-7B-OV</td><td>11.13</td><td>0.2651</td><td>0.0376</td><td>5.56</td><td>0.4213</td><td>0.0825</td><td>0</td><td>-</td><td>-</td></tr>
<tr><td>DeepSeek-VL-7B</td><td>48.68</td><td>0.2854</td><td>0.0431</td><td>61.11</td><td>0.5374</td><td>0.1114</td><td>10.19</td><td>0.2539</td><td>0.0145</td></tr>
<tr><td>kimi-VL-A3B</td><td>68.85</td><td>0.4409</td><td>0.1374</td><td>72.22</td><td>0.5887</td><td>0.2081</td><td>61.11</td><td>0.4641</td><td>0.1379</td></tr>
<tr><td>Qwen2-VL-7B</td><td>64.39</td><td>0.3364</td><td>0.0664</td><td>75.00</td><td>0.5950</td><td>0.1367</td><td>30.56</td><td>0.4235</td><td>0.0519</td></tr>
<tr><td>Qwen2-VL-72B</td><td>75.66</td><td>0.4368</td><td>0.1207</td><td>80.56</td><td>0.6082</td><td>0.1628</td><td>51.85</td><td>0.5518</td><td>0.1373</td></tr>
<tr><td>InternVL-2.5-8B</td><td>66.89</td><td>0.3348</td><td>0.0723</td><td>80.56</td><td>0.5712</td><td>0.1183</td><td>37.74</td><td>0.5715</td><td>0.0568</td></tr>
<tr><td>InternVL-2.5-38B</td><td>86.23</td><td>0.4577</td><td>0.1463</td><td>0</td><td>-</td><td>-</td><td>0</td><td>-</td><td>-</td></tr>
<tr><td>InternVL-3-8B</td><td>66.34</td><td>0.4371</td><td>0.1389</td><td>86.11</td><td>0.6169</td><td>0.1732</td><td>57.41</td><td>0.4450</td><td>0.1028</td></tr>
<tr><td>GLM-4V-9B</td><td>72.18</td><td>0.2881</td><td>0.0459</td><td>66.67</td><td>0.5628</td><td>0.1183</td><td>44.74</td><td>0.2904</td><td>0.0130</td></tr>
<tr><td>Intern-VL-3.5-8B</td><td>66.34</td><td>0.4371</td><td>0.1389</td><td>86.11</td><td>0.6169</td><td>0.1732</td><td>57.41</td><td>0.4450</td><td>0.1028</td></tr>
<tr><td>MiMo-VL-7B-RL</td><td>37.83</td><td>0.5439</td><td>0.2316</td><td>69.44</td><td>0.6068</td><td>0.2421</td><td>41.67</td><td>0.4962</td><td>0.1407</td></tr>
<tr><td>MiMo-VL-7B-SFT</td><td>44.65</td><td>0.4959</td><td>0.1983</td><td>69.44</td><td>0.6237</td><td>0.1852</td><td>46.30</td><td>0.5155</td><td>0.1732</td></tr>
<tr><td>Qwen2.5-VL-7B</td><td>65.64</td><td>0.4197</td><td>0.0994</td><td>75.00</td><td>0.5952</td><td>0.1515</td><td>44.44</td><td>0.5952</td><td>0.0910</td></tr>
<tr><td>Qwen2.5-VL-72B</td><td>65.36</td><td>0.5118</td><td>0.1893</td><td>100</td><td>0.6273</td><td>0.1989</td><td>37.96</td><td>0.5532</td><td>0.1688</td></tr>
<tr><td>Molmo-7B-D</td><td>34.77</td><td>0.2164</td><td>0.0943</td><td>4.55</td><td>0.2400</td><td><strong>0.4600</strong></td><td>0.97</td><td>0.0500</td><td><strong>0.4100</strong></td></tr>
<tr><td>Qwen3-30B</td><td>64.67</td><td>0.5293</td><td>0.2531</td><td>77.78</td><td>0.2546</td><td>0.2368</td><td>70.37</td><td>0.2412</td><td>0.2698</td></tr>
<tr class="group-header"><td colspan="10"><em><strong>Open-Source LMMs (thinking)</strong></em></td></tr>
<tr><td>MiMo-VL-7B-RL</td><td>55.77</td><td>0.5261</td><td>0.2294</td><td>69.44</td><td>0.6053</td><td>0.2582</td><td>33.33</td><td>0.5807</td><td>0.2172</td></tr>
<tr><td>MiMo-VL-7B-SFT</td><td>50.35</td><td><strong>0.6555</strong></td><td>0.2130</td><td>86.11</td><td>0.6644</td><td>0.2248</td><td>38.89</td><td>0.5578</td><td>0.1455</td></tr>
<tr><td>Qwen3-30B</td><td>45.06</td><td>0.5582</td><td>0.2730</td><td>72.22</td><td>0.3367</td><td>0.3368</td><td>39.81</td><td>0.3185</td><td>0.2780</td></tr>
</tbody>
</table>
</div>
<div id="level2" class="tab-pane">
<table class="sortable-table">
<thead>
<tr>
<th rowspan="2"><strong>Model</strong></th>
<th rowspan="2"><strong>Exec.<br>Rate</strong></th>
<th colspan="9"><strong>Code-Level</strong></th>
<th rowspan="2"><strong>Chart-Level<br>LMM-Score</strong></th>
</tr>
<tr>
<th><strong>Color</strong></th><th><strong>Grid</strong></th><th><strong>Layout</strong></th><th><strong>Legend</strong></th><th><strong>Visual</strong></th><th><strong>Data</strong></th><th><strong>Text</strong></th><th><strong>Type</strong></th><th><strong>LLM-Score</strong></th>
</tr>
</thead>
<tbody>
<tr class="group-header"><td colspan="12"><em><strong>Proprietary</strong></em></td></tr>
<tr><td>Gemini-2.5-Pro</td><td>90.30</td><td><strong>0.6217</strong></td><td><strong>0.8842</strong></td><td><strong>0.9613</strong></td><td><strong>0.5093</strong></td><td><strong>0.5170</strong></td><td><strong>0.7560</strong></td><td>0.6330</td><td><strong>0.9636</strong></td><td><strong>0.5742</strong></td><td>0.2459</td></tr>
<tr><td>Claude-Sonnet-4</td><td><strong>91.19</strong></td><td>0.5737</td><td>0.8110</td><td>0.9587</td><td>0.4714</td><td>0.4776</td><td>0.6736</td><td>0.5869</td><td>0.9563</td><td>0.5317</td><td>0.2147</td></tr>
<tr><td>GPT-5</td><td>90.58</td><td>0.5812</td><td>0.8467</td><td>0.9499</td><td>0.4835</td><td>0.4815</td><td>0.7047</td><td>0.6096</td><td>0.9581</td><td>0.5663</td><td><strong>0.2506</strong></td></tr>
<tr><td>Seed-1.5-VL</td><td>63.17</td><td>0.5106</td><td>0.8230</td><td>0.9538</td><td>0.4408</td><td>0.4582</td><td>0.6983</td><td>0.7166</td><td>0.9400</td><td>0.5126</td><td>0.1975</td></tr>
<tr><td>Seed-1.6-VL</td><td>72.38</td><td>0.5277</td><td>0.8013</td><td>0.9471</td><td>0.4714</td><td>0.4453</td><td>0.6884</td><td><strong>0.7312</strong></td><td>0.9431</td><td>0.5151</td><td>0.1863</td></tr>
<tr class="group-header"><td colspan="12"><em><strong>Open-Source LMMs (non-thinking)</strong></em></td></tr>
<tr><td>LLaVA-OV-Qwen2-7B-SI</td><td>1.19</td><td>0.3507</td><td>0.6964</td><td>0.7833</td><td>0.4074</td><td>0.3002</td><td>0.5249</td><td>0.4871</td><td>0.7889</td><td>0.3157</td><td>0.0875</td></tr>
<tr><td>LLaVA-OV-Qwen2-7B-OV</td><td>2.57</td><td>0.3163</td><td>0.6013</td><td>0.6863</td><td>0.4488</td><td>0.2030</td><td>0.5685</td><td>0.4928</td><td>0.8154</td><td>0.3512</td><td>0.0366</td></tr>
<tr><td>DeepSeek-VL-7B</td><td>21.68</td><td>0.2523</td><td>0.6206</td><td>0.7350</td><td>0.2436</td><td>0.1820</td><td>0.4031</td><td>0.4538</td><td>0.7922</td><td>0.2583</td><td>0.0433</td></tr>
<tr><td>kimi-VL-A3B</td><td>49.5</td><td>0.3901</td><td>0.7270</td><td>0.9074</td><td>0.3411</td><td>0.3196</td><td>0.5724</td><td>0.5913</td><td>0.9033</td><td>0.3701</td><td>0.1039</td></tr>
<tr><td>Qwen2-VL-7B</td><td>24.95</td><td>0.2846</td><td>0.5825</td><td>0.7711</td><td>0.2723</td><td>0.2385</td><td>0.4693</td><td>0.4883</td><td>0.8141</td><td>0.3181</td><td>0.0780</td></tr>
<tr><td>Qwen2-VL-72B</td><td>55.05</td><td>0.4013</td><td>0.7704</td><td>0.9044</td><td>0.3464</td><td>0.3345</td><td>0.6086</td><td>0.5744</td><td>0.9098</td><td>0.3928</td><td>0.1140</td></tr>
<tr><td>InternVL-2.5-8B</td><td>21.29</td><td>0.3341</td><td>0.7002</td><td>0.8362</td><td>0.3148</td><td>0.2955</td><td>0.5421</td><td>0.5530</td><td>0.8536</td><td>0.3344</td><td>0.0869</td></tr>
<tr><td>InternVL-2.5-38B</td><td>68.22</td><td>0.4544</td><td>0.7902</td><td>0.9405</td><td>0.4146</td><td>0.3745</td><td>0.6334</td><td>0.6361</td><td>0.9338</td><td>0.4311</td><td>0.1367</td></tr>
<tr><td>InternVL-3-8B</td><td>4.55</td><td>0.3491</td><td>0.5914</td><td>0.9447</td><td>0.3389</td><td>0.3645</td><td>0.5561</td><td>0.5421</td><td>0.8556</td><td>0.3419</td><td>0.0943</td></tr>
<tr><td>InternVL-3-38B</td><td>67.43</td><td>0.4720</td><td>0.7853</td><td>0.9410</td><td>0.4133</td><td>0.3994</td><td>0.6525</td><td>0.6538</td><td>0.9235</td><td>0.4528</td><td>0.1476</td></tr>
<tr><td>GLM-4V-9B</td><td>10.69</td><td>0.2011</td><td>0.6910</td><td>0.7794</td><td>0.2357</td><td>0.2196</td><td>0.4604</td><td>0.5003</td><td>0.7472</td><td>0.2953</td><td>0.0770</td></tr>
<tr><td>Intern-VL-3.5-8B</td><td>27.23</td><td>0.4015</td><td>0.7350</td><td>0.9056</td><td>0.3566</td><td>0.3718</td><td>0.6121</td><td>0.6505</td><td>0.8998</td><td>0.3964</td><td>0.1466</td></tr>
<tr><td>MiMo-VL-7B-RL</td><td>20.59</td><td>0.4378</td><td>0.8462</td><td>0.9205</td><td>0.4201</td><td>0.4231</td><td>0.6505</td><td>0.6666</td><td>0.9200</td><td>0.4615</td><td>0.1573</td></tr>
<tr><td>MiMo-VL-7B-SFT</td><td>21.88</td><td>0.4325</td><td>0.7506</td><td>0.8941</td><td>0.3823</td><td>0.4035</td><td>0.6431</td><td>0.6564</td><td>0.9405</td><td>0.4459</td><td>0.1399</td></tr>
<tr><td>Qwen2.5-VL-7B</td><td>33.36</td><td>0.3524</td><td>0.7374</td><td>0.8592</td><td>0.3296</td><td>0.3302</td><td>0.5944</td><td>0.5780</td><td>0.8887</td><td>0.3603</td><td>0.0974</td></tr>
<tr><td>Qwen2.5-VL-72B</td><td>71.49</td><td>0.5018</td><td>0.8229</td><td>0.9509</td><td>0.4467</td><td>0.4242</td><td>0.6673</td><td>0.6815</td><td>0.9348</td><td>0.4739</td><td>0.1684</td></tr>
<tr><td>Molmo-7B-D</td><td>0.99</td><td>0.2471</td><td>0.8152</td><td>0.5636</td><td>0.1000</td><td>0.2275</td><td>0.3477</td><td>0.3082</td><td>0.3476</td><td>0.3488</td><td>0.1347</td></tr>
<tr><td>Qwen3-30B</td><td>41.39</td><td>0.54</td><td>0.8174</td><td>0.9587</td><td>0.4623</td><td>0.4501</td><td>0.6911</td><td>0.7084</td><td>0.9384</td><td>0.3611</td><td>0.2257</td></tr>
<tr class="group-header"><td colspan="12"><em><strong>Open-Source LMMs (thinking)</strong></em></td></tr>
<tr><td>MiMo-VL-7B-RL</td><td>27.62</td><td>0.5076</td><td>0.7560</td><td>0.9449</td><td>0.4109</td><td>0.4379</td><td>0.7006</td><td>0.6859</td><td>0.9446</td><td>0.4819</td><td>0.1737</td></tr>
<tr><td>MiMo-VL-7B-SFT</td><td>24.16</td><td>0.4562</td><td>0.7404</td><td>0.9286</td><td>0.3686</td><td>0.3980</td><td>0.6812</td><td>0.6617</td><td>0.9385</td><td>0.4496</td><td>0.1774</td></tr>
<tr><td>Qwen3-30B</td><td>42.38</td><td>0.5213</td><td>0.8248</td><td>0.9549</td><td>0.4718</td><td>0.4453</td><td>0.6924</td><td>0.7046</td><td>0.9403</td><td>0.4947</td><td>0.2362</td></tr>
</tbody>
</table>
</div>
<div id="level3" class="tab-pane">
<table class="sortable-table">
<thead>
<tr>
<th rowspan="2"><strong>Model</strong></th>
<th rowspan="2"><strong>Exec.<br>Rate</strong></th>
<th colspan="9"><strong>Code-Level</strong></th>
<th rowspan="2"><strong>Figure-Level<br>LMM-Score</strong></th>
</tr>
<tr>
<th><strong>Color</strong></th><th><strong>Grid</strong></th><th><strong>Layout</strong></th><th><strong>Legend</strong></th><th><strong>Visual</strong></th><th><strong>Data</strong></th><th><strong>Text</strong></th><th><strong>Type</strong></th><th><strong>LLM-Score</strong></th>
</tr>
</thead>
<tbody>
<tr class="group-header"><td colspan="12"><em><strong>Proprietary</strong></em></td></tr>
<tr><td>Gemini-2.5-Pro</td><td>29.33</td><td><strong>0.7276</strong></td><td><strong>0.9733</strong></td><td>1.0000</td><td>0.7727</td><td><strong>0.6701</strong></td><td>0.7880</td><td><strong>0.8291</strong></td><td>0.9470</td><td>0.3516</td><td>0.0361</td></tr>
<tr><td>Claude-Sonnet-4</td><td><strong>38.00</strong></td><td>0.5676</td><td>0.7963</td><td>1.0000</td><td>0.8148</td><td>0.3731</td><td>0.5881</td><td>0.7175</td><td>0.9062</td><td><strong>0.5125</strong></td><td>0.007</td></tr>
<tr><td>GPT-5</td><td><strong>38.00</strong></td><td>0.5676</td><td>0.7963</td><td>1.0000</td><td>0.8148</td><td>0.3731</td><td>0.5881</td><td>0.7175</td><td>0.9062</td><td><strong>0.5125</strong></td><td>0.0362</td></tr>
<tr><td>Seed-1.5-VL</td><td>18.67</td><td>0.7252</td><td>0.8929</td><td>1.0000</td><td><strong>0.8869</strong></td><td>0.5502</td><td>0.7182</td><td>0.7804</td><td><strong>0.9690</strong></td><td>0.0000</td><td><strong>0.0611</strong></td></tr>
<tr><td>Seed-1.6-VL</td><td>40.00</td><td>0.7030</td><td>0.8833</td><td>1.0000</td><td>0.7972</td><td>0.5396</td><td><strong>0.7956</strong></td><td>0.8128</td><td>0.9244</td><td>0.0000</td><td>0.0547</td></tr>
</tbody>
</table>
</div>
</div>
</div>
</div>
<h2 class="text-title">Results Analysis</h2>
<img src="assets/figure7.png" class="responsive-image" alt="Correlation of model performance">
<p class="text-content text-center">Correlation of the model performance (i.e., LMM-score) on different manually annotated difficulties (i.e., Easy, Medium, Hard) on Level 1, 2, 3, respectively.</p>
<img src="assets/figure8.png" class="responsive-image" alt="Model generalization analysis">
<p class="text-content">Left: Both proprietary and open-source models generalize well on Level 1 and Level 2 tasks when calculating the LLM-score for predicted code assessment. Right: Proprietary models tend to obtain higher LMM-scores on the Level 1 task rather than the Level 2, while open-source models perform poorly on both tasks (scores are lower than 0.5).</p>
<img src="assets/figure9.png" class="responsive-image" alt="Performance analysis on different task cases">
<p class="text-content text-center">Analysis of model performance on different task cases with LLM-score and LMM-score.</p>
<h2 class="text-title">Example of Data and Error Cases</h2>
<div class="leaderboard-container">
<div class="tabs-container">
<div class="tab-links">
<button class="tab-link active" data-tab="dr-sample">Level 1: Direct Reproduction</button>
<button class="tab-link" data-tab="crd-sample">Level 1: Custom Raw Darta</button>
<button class="tab-link" data-tab="cfd-sample">Level 1: Custom Figure</button>
<button class="tab-link" data-tab="level2-sample">Level 2 Example</button>
<button class="tab-link" data-tab="level3-sample">Level 3 Example</button>
<button class="tab-link" data-tab="error-cases">Error Cases</button>
</div>
<div class="tab-content">
<div id="dr-sample" class="tab-pane active">
<h3 class="text-title">An Example of Level 1: Direct Reproduction</h3>
<img src="assets/DR_sample.png" alt="Direct Reproduction Example" class="tab-image">
</div>
<div id="crd-sample" class="tab-pane">
<h3 class="text-title">An Example of Level 1: Customized Text-Format Table Data</h3>
<img src="assets/CRD_sample.png" alt="Customized Raw Data Example" class="tab-image">
</div>
<div id="cfd-sample" class="tab-pane">
<h3 class="text-title">An Example of Level 1: Figure-Format Table Data</h3>
<img src="assets/CFD_sample.png" alt="Customized Figure Data Example" class="tab-image">
</div>
<div id="level2-sample" class="tab-pane">
<h3 class="text-title">An Example of Level 2</h3>
<img src="assets/level2_sample.png" alt="Level 2 Example" class="tab-image">
</div>
<div id="level3-sample" class="tab-pane">
<h3 class="text-title">An Example of Level 3</h3>
<img src="assets/level3_sample.png" alt="Level 3 Example" class="tab-image">
</div>
<div id="error-cases" class="tab-pane">
<h3 class="text-title">Error Cases Visualization</h3>
<img src="assets/error_analyse.png" alt="Error Analysis 1" class="tab-image" style="margin-bottom: 20px;">
<img src="assets/error_analyse_2.png" alt="Error Analysis 2" class="tab-image">
</div>
</div>
</div>
</div>
<h2 class="text-title">Citation</h2>
<pre id="citation"><code>
@misc{tang2025chartscodehierarchicalbenchmark,
title={From Charts to Code: A Hierarchical Benchmark for Multimodal Models},
author={Jiahao Tang and Henry Hengyuan Zhao and Lijian Wu and Yifei Tao and Dongxing Mao and Yang Wan and Jingru Tan and Min Zeng and Min Li and Alex Jinpeng Wang},
year={2025},
eprint={2510.17932},
archivePrefix={arXiv},
primaryClass={cs.SE},
url={https://arxiv.org/abs/2510.17932},
}
</code></pre>
<p class="acknowledgements">
<b>Acknowledge:</b> Thanks to Carlos & John for this webpage template. Also thanks to the SWE-bench team and their benchmark <a href="https://www.swebench.com/multimodal.html">https://www.swebench.com/multimodal.html</a>.
</p>
<p class="acknowledgements">
<b>Template Usage:</b> If you would like to use this website template for your own leaderboard, please <span style="color:brown">send Carlos & John an email requesting permission.</span> If granted, please make sure to acknowledge the SWE-bench team and link to this leaderboard on the home page of the website.
</p>
<div class="logo-group">
<a href="https://www.csu.edu.cn/"><img src="./assets/csu.gif" /></a>
<a href="https://www.nus.edu.sg/"><img src="https://www.nus.edu.sg/images/default-source/base/logo.png" alt="NUS Logo" /></a>
<a href="https://www.ntu.edu.sg//"><img src="./assets/NTU.png" alt="NUS Logo" /></a>
</div>
</div>
</div>
</section>
</div>
<script src="css/leaderboard.js"></script>
</body>
</html>