Update README.md
Browse files
README.md
CHANGED
@@ -48,6 +48,281 @@ print(evidence)
|
|
48 |
# evidence:
|
49 |
```
|
50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
## About
|
52 |
|
53 |
*Built by Dien X. Tran*
|
|
|
48 |
# evidence:
|
49 |
```
|
50 |
|
51 |
+
## **Evaluation Results**
|
52 |
+
|
53 |
+
We evaluate the model's performance in the QA-based approaches section without handling token lengths exceeding 512.
|
54 |
+
|
55 |
+
<table>
|
56 |
+
<thead>
|
57 |
+
<tr>
|
58 |
+
<th colspan="2">Method</th>
|
59 |
+
<th colspan="4">ISE-DSC01</th>
|
60 |
+
</tr>
|
61 |
+
<tr>
|
62 |
+
<th>ER</th>
|
63 |
+
<th>VC</th>
|
64 |
+
<th>Strict Acc</th>
|
65 |
+
<th>VC Acc</th>
|
66 |
+
<th>ER Acc</th>
|
67 |
+
<th>Time (s)</th>
|
68 |
+
</tr>
|
69 |
+
</thead>
|
70 |
+
<tbody>
|
71 |
+
<tr>
|
72 |
+
<td rowspan="3">TF-IDF</td>
|
73 |
+
<td>InfoXLM<sub>large</sub></td>
|
74 |
+
<td>73.59</td>
|
75 |
+
<td>78.08</td>
|
76 |
+
<td>76.61</td>
|
77 |
+
<td>378</td>
|
78 |
+
</tr>
|
79 |
+
<tr>
|
80 |
+
<td>XLM-R<sub>large</sub></td>
|
81 |
+
<td>75.61</td>
|
82 |
+
<td>80.50</td>
|
83 |
+
<td>78.58</td>
|
84 |
+
<td>366</td>
|
85 |
+
</tr>
|
86 |
+
<tr>
|
87 |
+
<td>Ernie-M<sub>large</sub></td>
|
88 |
+
<td>78.19</td>
|
89 |
+
<td>81.69</td>
|
90 |
+
<td>80.65</td>
|
91 |
+
<td>403</td>
|
92 |
+
</tr>
|
93 |
+
<tr>
|
94 |
+
<td rowspan="3">BM25</td>
|
95 |
+
<td>InfoXLM<sub>large</sub></td>
|
96 |
+
<td>72.09</td>
|
97 |
+
<td>77.37</td>
|
98 |
+
<td>75.04</td>
|
99 |
+
<td>320</td>
|
100 |
+
</tr>
|
101 |
+
<tr>
|
102 |
+
<td>XLM-R<sub>large</sub></td>
|
103 |
+
<td>73.94</td>
|
104 |
+
<td>79.37</td>
|
105 |
+
<td>76.95</td>
|
106 |
+
<td>333</td>
|
107 |
+
</tr>
|
108 |
+
<tr>
|
109 |
+
<td>Ernie-M<sub>large</sub></td>
|
110 |
+
<td>76.58</td>
|
111 |
+
<td>80.76</td>
|
112 |
+
<td>79.02</td>
|
113 |
+
<td>381</td>
|
114 |
+
</tr>
|
115 |
+
<tr>
|
116 |
+
<td rowspan="3">SBert</td>
|
117 |
+
<td>InfoXLM<sub>large</sub></td>
|
118 |
+
<td>71.20</td>
|
119 |
+
<td>76.59</td>
|
120 |
+
<td>74.15</td>
|
121 |
+
<td>915</td>
|
122 |
+
</tr>
|
123 |
+
<tr>
|
124 |
+
<td>XLM-R<sub>large</sub></td>
|
125 |
+
<td>72.85</td>
|
126 |
+
<td>78.78</td>
|
127 |
+
<td>75.89</td>
|
128 |
+
<td>835</td>
|
129 |
+
</tr>
|
130 |
+
<tr>
|
131 |
+
<td>Ernie-M<sub>large</sub></td>
|
132 |
+
<td>75.46</td>
|
133 |
+
<td>79.89</td>
|
134 |
+
<td>77.91</td>
|
135 |
+
<td>920</td>
|
136 |
+
</tr>
|
137 |
+
<tr>
|
138 |
+
<th colspan="1">QA-based approaches</th>
|
139 |
+
<th colspan="1">VC</th>
|
140 |
+
<th colspan="4"></th>
|
141 |
+
</tr>
|
142 |
+
<tr>
|
143 |
+
<td rowspan="3">ViMRC<sub>large</sub></td>
|
144 |
+
<td>InfoXLM<sub>large</sub></td>
|
145 |
+
<td>54.36</td>
|
146 |
+
<td>64.14</td>
|
147 |
+
<td>56.84</td>
|
148 |
+
<td>9798</td>
|
149 |
+
</tr>
|
150 |
+
<tr>
|
151 |
+
<td>XLM-R<sub>large</sub></td>
|
152 |
+
<td>53.98</td>
|
153 |
+
<td>66.70</td>
|
154 |
+
<td>57.77</td>
|
155 |
+
<td>9809</td>
|
156 |
+
</tr>
|
157 |
+
<tr>
|
158 |
+
<td>Ernie-M<sub>large</sub></td>
|
159 |
+
<td>56.62</td>
|
160 |
+
<td>62.19</td>
|
161 |
+
<td>58.91</td>
|
162 |
+
<td>9833</td>
|
163 |
+
</tr>
|
164 |
+
<tr>
|
165 |
+
<td rowspan="3">InfoXLM<sub>large</sub></td>
|
166 |
+
<td>InfoXLM<sub>large</sub></td>
|
167 |
+
<td>53.50</td>
|
168 |
+
<td>63.83</td>
|
169 |
+
<td>56.17</td>
|
170 |
+
<td>10057</td>
|
171 |
+
</tr>
|
172 |
+
<tr>
|
173 |
+
<td>XLM-R<sub>large</sub></td>
|
174 |
+
<td>53.32</td>
|
175 |
+
<td>66.70</td>
|
176 |
+
<td>57.25</td>
|
177 |
+
<td>10066</td>
|
178 |
+
</tr>
|
179 |
+
<tr>
|
180 |
+
<td>Ernie-M<sub>large</sub></td>
|
181 |
+
<td>56.34</td>
|
182 |
+
<td>62.36</td>
|
183 |
+
<td>58.69</td>
|
184 |
+
<td>10078</td>
|
185 |
+
</tr>
|
186 |
+
<tr>
|
187 |
+
<th colspan="2">LLM</th>
|
188 |
+
<th colspan="4"></th>
|
189 |
+
</tr>
|
190 |
+
<tr>
|
191 |
+
<td colspan="2">Qwen2.5-1.5B-Instruct</td>
|
192 |
+
<td>59.23</td>
|
193 |
+
<td>66.68</td>
|
194 |
+
<td>65.51</td>
|
195 |
+
<td>19780</td>
|
196 |
+
</tr>
|
197 |
+
<tr>
|
198 |
+
<td colspan="2">Qwen2.5-3B-Instruct</td>
|
199 |
+
<td>60.87</td>
|
200 |
+
<td>66.92</td>
|
201 |
+
<td>66.10</td>
|
202 |
+
<td>31284</td>
|
203 |
+
</tr>
|
204 |
+
<tr>
|
205 |
+
<th colspan="1">LLM</th>
|
206 |
+
<th colspan="1">VC</th>
|
207 |
+
<th colspan="4"></th>
|
208 |
+
</tr>
|
209 |
+
<tr>
|
210 |
+
<td rowspan="3">Qwen2.5-1.5B-Instruct</td>
|
211 |
+
<td>InfoXLM<sub>large</sub></td>
|
212 |
+
<td>64.40</td>
|
213 |
+
<td>68.37</td>
|
214 |
+
<td>66.49</td>
|
215 |
+
<td>19970</td>
|
216 |
+
</tr>
|
217 |
+
<tr>
|
218 |
+
<td>XLM-R<sub>large</sub></td>
|
219 |
+
<td>64.66</td>
|
220 |
+
<td>69.63</td>
|
221 |
+
<td>66.72</td>
|
222 |
+
<td>19976</td>
|
223 |
+
</tr>
|
224 |
+
<tr>
|
225 |
+
<td>Ernie-M<sub>large</sub></td>
|
226 |
+
<td>65.70</td>
|
227 |
+
<td>68.37</td>
|
228 |
+
<td>67.33</td>
|
229 |
+
<td>20003</td>
|
230 |
+
</tr>
|
231 |
+
<tr>
|
232 |
+
<td rowspan="3">Qwen2.5-3B-Instruct</td>
|
233 |
+
<td>InfoXLM<sub>large</sub></td>
|
234 |
+
<td>65.72</td>
|
235 |
+
<td>69.66</td>
|
236 |
+
<td>67.51</td>
|
237 |
+
<td>31477</td>
|
238 |
+
</tr>
|
239 |
+
<tr>
|
240 |
+
<td>XLM-R<sub>large</sub></td>
|
241 |
+
<td>66.12</td>
|
242 |
+
<td>70.44</td>
|
243 |
+
<td>67.83</td>
|
244 |
+
<td>31483</td>
|
245 |
+
</tr>
|
246 |
+
<tr>
|
247 |
+
<td>Ernie-M<sub>large</sub></td>
|
248 |
+
<td>67.48</td>
|
249 |
+
<td>70.77</td>
|
250 |
+
<td>68.75</td>
|
251 |
+
<td>31512</td>
|
252 |
+
</tr>
|
253 |
+
<tr>
|
254 |
+
<th colspan="1">SER Faster (ours)</th>
|
255 |
+
<th colspan="1">TVC (ours)</th>
|
256 |
+
<th colspan="4"></th>
|
257 |
+
</tr>
|
258 |
+
<tr>
|
259 |
+
<td>TF-IDF + ViMRC<sub>large</sub></td>
|
260 |
+
<td>Ernie-M<sub>large</sub></td>
|
261 |
+
<td style="color:blue">78.32</td>
|
262 |
+
<td style="color:blue">81.91</td>
|
263 |
+
<td style="color:blue">80.26</td>
|
264 |
+
<td style="color:blue">995</td>
|
265 |
+
</tr>
|
266 |
+
<tr>
|
267 |
+
<td>TF-IDF + InfoXLM<sub>large</sub></td>
|
268 |
+
<td>Ernie-M<sub>large</sub></td>
|
269 |
+
<td style="color:blue">78.37</td>
|
270 |
+
<td style="color:blue">81.91</td>
|
271 |
+
<td style="color:blue">80.32</td>
|
272 |
+
<td style="color:blue">925</td>
|
273 |
+
</tr>
|
274 |
+
<tr>
|
275 |
+
<th colspan="1">SER (ours)</th>
|
276 |
+
<th colspan="1">TVC (ours)</th>
|
277 |
+
<th colspan="4"></th>
|
278 |
+
</tr>
|
279 |
+
<tr>
|
280 |
+
<td rowspan="3">TF-IDF + ViMRC<sub>large</sub></td>
|
281 |
+
<td>InfoXLM<sub>large</sub></td>
|
282 |
+
<td>75.13</td>
|
283 |
+
<td>79.54</td>
|
284 |
+
<td>76.87</td>
|
285 |
+
<td>5191</td>
|
286 |
+
</tr>
|
287 |
+
<tr>
|
288 |
+
<td>XLM-R<sub>large</sub></td>
|
289 |
+
<td>76.71</td>
|
290 |
+
<td>81.65</td>
|
291 |
+
<td>78.91</td>
|
292 |
+
<td>5219</td>
|
293 |
+
</tr>
|
294 |
+
<tr>
|
295 |
+
<td>Ernie-M<sub>large</sub></td>
|
296 |
+
<td><strong>78.97</strong></td>
|
297 |
+
<td><strong>82.54</strong></td>
|
298 |
+
<td><strong>80.91</strong></td>
|
299 |
+
<td>5225</td>
|
300 |
+
</tr>
|
301 |
+
<tr>
|
302 |
+
<td rowspan="3">TF-IDF + InfoXLM<sub>large</sub></td>
|
303 |
+
<td>InfoXLM<sub>large</sub></td>
|
304 |
+
<td>75.13</td>
|
305 |
+
<td>79.60</td>
|
306 |
+
<td>76.87</td>
|
307 |
+
<td>5175</td>
|
308 |
+
</tr>
|
309 |
+
<tr>
|
310 |
+
<td>XLM-R<sub>large</sub></td>
|
311 |
+
<td>76.74</td>
|
312 |
+
<td>81.71</td>
|
313 |
+
<td>78.95</td>
|
314 |
+
<td>5200</td>
|
315 |
+
</tr>
|
316 |
+
<tr>
|
317 |
+
<td>Ernie-M<sub>large</sub></td>
|
318 |
+
<td><strong>78.97</strong></td>
|
319 |
+
<td>82.49</td>
|
320 |
+
<td><strong>80.91</strong></td>
|
321 |
+
<td>5297</td>
|
322 |
+
</tr>
|
323 |
+
</tbody>
|
324 |
+
</table>
|
325 |
+
|
326 |
## About
|
327 |
|
328 |
*Built by Dien X. Tran*
|