Update README.md
Browse files
README.md
CHANGED
|
@@ -85,6 +85,277 @@ print(evidence)
|
|
| 85 |
|
| 86 |
## **Evaluation Results**
|
| 87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
**SemViQA-QATC** plays a crucial role in the **SemViQA** system by enhancing accuracy in evidence extraction. When integrated into a pipeline, this model helps determine whether a claim is supported or refuted based on retrieved evidence.
|
| 89 |
|
| 90 |
## **Citation**
|
|
|
|
| 85 |
|
| 86 |
## **Evaluation Results**
|
| 87 |
|
| 88 |
+
<table>
|
| 89 |
+
<thead>
|
| 90 |
+
<tr>
|
| 91 |
+
<th colspan="2">Method</th>
|
| 92 |
+
<th colspan="4">ISE-DSC01</th>
|
| 93 |
+
</tr>
|
| 94 |
+
<tr>
|
| 95 |
+
<th>ER</th>
|
| 96 |
+
<th>VC</th>
|
| 97 |
+
<th>Strict Acc</th>
|
| 98 |
+
<th>VC Acc</th>
|
| 99 |
+
<th>ER Acc</th>
|
| 100 |
+
<th>Time (s)</th>
|
| 101 |
+
</tr>
|
| 102 |
+
</thead>
|
| 103 |
+
<tbody>
|
| 104 |
+
<tr>
|
| 105 |
+
<td rowspan="3">TF-IDF</td>
|
| 106 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 107 |
+
<td>73.59</td>
|
| 108 |
+
<td>78.08</td>
|
| 109 |
+
<td>76.61</td>
|
| 110 |
+
<td>378</td>
|
| 111 |
+
</tr>
|
| 112 |
+
<tr>
|
| 113 |
+
<td>XLM-R<sub>large</sub></td>
|
| 114 |
+
<td>75.61</td>
|
| 115 |
+
<td>80.50</td>
|
| 116 |
+
<td>78.58</td>
|
| 117 |
+
<td>366</td>
|
| 118 |
+
</tr>
|
| 119 |
+
<tr>
|
| 120 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 121 |
+
<td>78.19</td>
|
| 122 |
+
<td>81.69</td>
|
| 123 |
+
<td>80.65</td>
|
| 124 |
+
<td>403</td>
|
| 125 |
+
</tr>
|
| 126 |
+
<tr>
|
| 127 |
+
<td rowspan="3">BM25</td>
|
| 128 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 129 |
+
<td>72.09</td>
|
| 130 |
+
<td>77.37</td>
|
| 131 |
+
<td>75.04</td>
|
| 132 |
+
<td>320</td>
|
| 133 |
+
</tr>
|
| 134 |
+
<tr>
|
| 135 |
+
<td>XLM-R<sub>large</sub></td>
|
| 136 |
+
<td>73.94</td>
|
| 137 |
+
<td>79.37</td>
|
| 138 |
+
<td>76.95</td>
|
| 139 |
+
<td>333</td>
|
| 140 |
+
</tr>
|
| 141 |
+
<tr>
|
| 142 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 143 |
+
<td>76.58</td>
|
| 144 |
+
<td>80.76</td>
|
| 145 |
+
<td>79.02</td>
|
| 146 |
+
<td>381</td>
|
| 147 |
+
</tr>
|
| 148 |
+
<tr>
|
| 149 |
+
<td rowspan="3">SBert</td>
|
| 150 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 151 |
+
<td>71.20</td>
|
| 152 |
+
<td>76.59</td>
|
| 153 |
+
<td>74.15</td>
|
| 154 |
+
<td>915</td>
|
| 155 |
+
</tr>
|
| 156 |
+
<tr>
|
| 157 |
+
<td>XLM-R<sub>large</sub></td>
|
| 158 |
+
<td>72.85</td>
|
| 159 |
+
<td>78.78</td>
|
| 160 |
+
<td>75.89</td>
|
| 161 |
+
<td>835</td>
|
| 162 |
+
</tr>
|
| 163 |
+
<tr>
|
| 164 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 165 |
+
<td>75.46</td>
|
| 166 |
+
<td>79.89</td>
|
| 167 |
+
<td>77.91</td>
|
| 168 |
+
<td>920</td>
|
| 169 |
+
</tr>
|
| 170 |
+
<tr>
|
| 171 |
+
<th colspan="1">QA-based approaches</th>
|
| 172 |
+
<th colspan="1">VC</th>
|
| 173 |
+
<th colspan="4"></th>
|
| 174 |
+
</tr>
|
| 175 |
+
<tr>
|
| 176 |
+
<td rowspan="3">ViMRC<sub>large</sub></td>
|
| 177 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 178 |
+
<td>54.36</td>
|
| 179 |
+
<td>64.14</td>
|
| 180 |
+
<td>56.84</td>
|
| 181 |
+
<td>9798</td>
|
| 182 |
+
</tr>
|
| 183 |
+
<tr>
|
| 184 |
+
<td>XLM-R<sub>large</sub></td>
|
| 185 |
+
<td>53.98</td>
|
| 186 |
+
<td>66.70</td>
|
| 187 |
+
<td>57.77</td>
|
| 188 |
+
<td>9809</td>
|
| 189 |
+
</tr>
|
| 190 |
+
<tr>
|
| 191 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 192 |
+
<td>56.62</td>
|
| 193 |
+
<td>62.19</td>
|
| 194 |
+
<td>58.91</td>
|
| 195 |
+
<td>9833</td>
|
| 196 |
+
</tr>
|
| 197 |
+
<tr>
|
| 198 |
+
<td rowspan="3">InfoXLM<sub>large</sub></td>
|
| 199 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 200 |
+
<td>53.50</td>
|
| 201 |
+
<td>63.83</td>
|
| 202 |
+
<td>56.17</td>
|
| 203 |
+
<td>10057</td>
|
| 204 |
+
</tr>
|
| 205 |
+
<tr>
|
| 206 |
+
<td>XLM-R<sub>large</sub></td>
|
| 207 |
+
<td>53.32</td>
|
| 208 |
+
<td>66.70</td>
|
| 209 |
+
<td>57.25</td>
|
| 210 |
+
<td>10066</td>
|
| 211 |
+
</tr>
|
| 212 |
+
<tr>
|
| 213 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 214 |
+
<td>56.34</td>
|
| 215 |
+
<td>62.36</td>
|
| 216 |
+
<td>58.69</td>
|
| 217 |
+
<td>10078</td>
|
| 218 |
+
</tr>
|
| 219 |
+
<tr>
|
| 220 |
+
<th colspan="2">LLM</th>
|
| 221 |
+
<th colspan="4"></th>
|
| 222 |
+
</tr>
|
| 223 |
+
<tr>
|
| 224 |
+
<td colspan="2">Qwen2.5-1.5B-Instruct</td>
|
| 225 |
+
<td>59.23</td>
|
| 226 |
+
<td>66.68</td>
|
| 227 |
+
<td>65.51</td>
|
| 228 |
+
<td>19780</td>
|
| 229 |
+
</tr>
|
| 230 |
+
<tr>
|
| 231 |
+
<td colspan="2">Qwen2.5-3B-Instruct</td>
|
| 232 |
+
<td>60.87</td>
|
| 233 |
+
<td>66.92</td>
|
| 234 |
+
<td>66.10</td>
|
| 235 |
+
<td>31284</td>
|
| 236 |
+
</tr>
|
| 237 |
+
<tr>
|
| 238 |
+
<th colspan="1">LLM</th>
|
| 239 |
+
<th colspan="1">VC</th>
|
| 240 |
+
<th colspan="4"></th>
|
| 241 |
+
</tr>
|
| 242 |
+
<tr>
|
| 243 |
+
<td rowspan="3">Qwen2.5-1.5B-Instruct</td>
|
| 244 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 245 |
+
<td>64.40</td>
|
| 246 |
+
<td>68.37</td>
|
| 247 |
+
<td>66.49</td>
|
| 248 |
+
<td>19970</td>
|
| 249 |
+
</tr>
|
| 250 |
+
<tr>
|
| 251 |
+
<td>XLM-R<sub>large</sub></td>
|
| 252 |
+
<td>64.66</td>
|
| 253 |
+
<td>69.63</td>
|
| 254 |
+
<td>66.72</td>
|
| 255 |
+
<td>19976</td>
|
| 256 |
+
</tr>
|
| 257 |
+
<tr>
|
| 258 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 259 |
+
<td>65.70</td>
|
| 260 |
+
<td>68.37</td>
|
| 261 |
+
<td>67.33</td>
|
| 262 |
+
<td>20003</td>
|
| 263 |
+
</tr>
|
| 264 |
+
<tr>
|
| 265 |
+
<td rowspan="3">Qwen2.5-3B-Instruct</td>
|
| 266 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 267 |
+
<td>65.72</td>
|
| 268 |
+
<td>69.66</td>
|
| 269 |
+
<td>67.51</td>
|
| 270 |
+
<td>31477</td>
|
| 271 |
+
</tr>
|
| 272 |
+
<tr>
|
| 273 |
+
<td>XLM-R<sub>large</sub></td>
|
| 274 |
+
<td>66.12</td>
|
| 275 |
+
<td>70.44</td>
|
| 276 |
+
<td>67.83</td>
|
| 277 |
+
<td>31483</td>
|
| 278 |
+
</tr>
|
| 279 |
+
<tr>
|
| 280 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 281 |
+
<td>67.48</td>
|
| 282 |
+
<td>70.77</td>
|
| 283 |
+
<td>68.75</td>
|
| 284 |
+
<td>31512</td>
|
| 285 |
+
</tr>
|
| 286 |
+
<tr>
|
| 287 |
+
<th colspan="1">SER Faster (ours)</th>
|
| 288 |
+
<th colspan="1">TVC (ours)</th>
|
| 289 |
+
<th colspan="4"></th>
|
| 290 |
+
</tr>
|
| 291 |
+
<tr>
|
| 292 |
+
<td>TF-IDF + ViMRC<sub>large</sub></td>
|
| 293 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 294 |
+
<td style="color:blue">78.32</td>
|
| 295 |
+
<td style="color:blue">81.91</td>
|
| 296 |
+
<td style="color:blue">80.26</td>
|
| 297 |
+
<td style="color:blue">995</td>
|
| 298 |
+
</tr>
|
| 299 |
+
<tr>
|
| 300 |
+
<td>TF-IDF + InfoXLM<sub>large</sub></td>
|
| 301 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 302 |
+
<td style="color:blue">78.37</td>
|
| 303 |
+
<td style="color:blue">81.91</td>
|
| 304 |
+
<td style="color:blue">80.32</td>
|
| 305 |
+
<td style="color:blue">925</td>
|
| 306 |
+
</tr>
|
| 307 |
+
<tr>
|
| 308 |
+
<th colspan="1">SER (ours)</th>
|
| 309 |
+
<th colspan="1">TVC (ours)</th>
|
| 310 |
+
<th colspan="4"></th>
|
| 311 |
+
</tr>
|
| 312 |
+
<tr>
|
| 313 |
+
<td rowspan="3">TF-IDF + ViMRC<sub>large</sub></td>
|
| 314 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 315 |
+
<td>75.13</td>
|
| 316 |
+
<td>79.54</td>
|
| 317 |
+
<td>76.87</td>
|
| 318 |
+
<td>5191</td>
|
| 319 |
+
</tr>
|
| 320 |
+
<tr>
|
| 321 |
+
<td>XLM-R<sub>large</sub></td>
|
| 322 |
+
<td>76.71</td>
|
| 323 |
+
<td>81.65</td>
|
| 324 |
+
<td>78.91</td>
|
| 325 |
+
<td>5219</td>
|
| 326 |
+
</tr>
|
| 327 |
+
<tr>
|
| 328 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 329 |
+
<td><strong>78.97</strong></td>
|
| 330 |
+
<td><strong>82.54</strong></td>
|
| 331 |
+
<td><strong>80.91</strong></td>
|
| 332 |
+
<td>5225</td>
|
| 333 |
+
</tr>
|
| 334 |
+
<tr>
|
| 335 |
+
<td rowspan="3">TF-IDF + InfoXLM<sub>large</sub></td>
|
| 336 |
+
<td>InfoXLM<sub>large</sub></td>
|
| 337 |
+
<td>75.13</td>
|
| 338 |
+
<td>79.60</td>
|
| 339 |
+
<td>76.87</td>
|
| 340 |
+
<td>5175</td>
|
| 341 |
+
</tr>
|
| 342 |
+
<tr>
|
| 343 |
+
<td>XLM-R<sub>large</sub></td>
|
| 344 |
+
<td>76.74</td>
|
| 345 |
+
<td>81.71</td>
|
| 346 |
+
<td>78.95</td>
|
| 347 |
+
<td>5200</td>
|
| 348 |
+
</tr>
|
| 349 |
+
<tr>
|
| 350 |
+
<td>Ernie-M<sub>large</sub></td>
|
| 351 |
+
<td><strong>78.97</strong></td>
|
| 352 |
+
<td>82.49</td>
|
| 353 |
+
<td><strong>80.91</strong></td>
|
| 354 |
+
<td>5297</td>
|
| 355 |
+
</tr>
|
| 356 |
+
</tbody>
|
| 357 |
+
</table>
|
| 358 |
+
|
| 359 |
**SemViQA-QATC** plays a crucial role in the **SemViQA** system by enhancing accuracy in evidence extraction. When integrated into a pipeline, this model helps determine whether a claim is supported or refuted based on retrieved evidence.
|
| 360 |
|
| 361 |
## **Citation**
|