add scores
Browse files
README.md
CHANGED
|
@@ -108,7 +108,159 @@ Use the code below to get started with the model.
|
|
| 108 |
|
| 109 |
## Evaluation
|
| 110 |
|
| 111 |
-
<
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 112 |
|
| 113 |
### Testing Data, Factors & Metrics
|
| 114 |
|
|
|
|
| 108 |
|
| 109 |
## Evaluation
|
| 110 |
|
| 111 |
+
<table border="1" style="width: 100%; text-align: center; border-collapse: collapse;">
|
| 112 |
+
<colgroup>
|
| 113 |
+
<col style="width: 10%;">
|
| 114 |
+
<col style="width: 10%;">
|
| 115 |
+
<col style="width: 7%;">
|
| 116 |
+
<col style="width: 7%;">
|
| 117 |
+
<col style="width: 7%;">
|
| 118 |
+
<col style="background-color: rgba(80, 15, 213, 0.5); width: 7%;">
|
| 119 |
+
</colgroup>
|
| 120 |
+
<thead>
|
| 121 |
+
<tr>
|
| 122 |
+
<th>Category</th>
|
| 123 |
+
<th>Benchmark</th>
|
| 124 |
+
<th>Llama-3.2-1B</th>
|
| 125 |
+
<th>Qwen2.5-1.5B</th>
|
| 126 |
+
<th>SmolLM2-1.7B</th>
|
| 127 |
+
<th>Falcon3-1B-Instruct</th>
|
| 128 |
+
</tr>
|
| 129 |
+
</thead>
|
| 130 |
+
<tbody>
|
| 131 |
+
<tr>
|
| 132 |
+
<td rowspan="3">General</td>
|
| 133 |
+
<td>MMLU (5-shot)</td>
|
| 134 |
+
<td>23.4</td>
|
| 135 |
+
<td><b>58.4</b></td>
|
| 136 |
+
<td>48.4</td>
|
| 137 |
+
<td>43.9</td>
|
| 138 |
+
</tr>
|
| 139 |
+
<tr>
|
| 140 |
+
<td>MMLU-PRO (5-shot)</td>
|
| 141 |
+
<td>11.3</td>
|
| 142 |
+
<td><b>21.3</b></td>
|
| 143 |
+
<td>17.2</td>
|
| 144 |
+
<td>18.6</td>
|
| 145 |
+
</tr>
|
| 146 |
+
<tr>
|
| 147 |
+
<td>IFEval</td>
|
| 148 |
+
<td><b>55.8</b></td>
|
| 149 |
+
<td>44.4</td>
|
| 150 |
+
<td>53.0</td>
|
| 151 |
+
<td>54.4</td>
|
| 152 |
+
</tr>
|
| 153 |
+
<tr>
|
| 154 |
+
<td rowspan="3">Math</td>
|
| 155 |
+
<td>GSM8K (5-shot)</td>
|
| 156 |
+
<td>37.4</td>
|
| 157 |
+
<td><b>57.2</b></td>
|
| 158 |
+
<td>43.4</td>
|
| 159 |
+
<td>38.6</td>
|
| 160 |
+
</tr>
|
| 161 |
+
<tr>
|
| 162 |
+
<td>GSM8K (8-shot, COT)</td>
|
| 163 |
+
<td>35.6</td>
|
| 164 |
+
<td><b>62.2</b></td>
|
| 165 |
+
<td>47.2</td>
|
| 166 |
+
<td>41.8</td>
|
| 167 |
+
</tr>
|
| 168 |
+
<tr>
|
| 169 |
+
<td>MATH Lvl-5 (4-shot)</td>
|
| 170 |
+
<td><b>3.9</b></td>
|
| 171 |
+
<td>0.2</td>
|
| 172 |
+
<td>0.1</td>
|
| 173 |
+
<td>1.0</td>
|
| 174 |
+
</tr>
|
| 175 |
+
<tr>
|
| 176 |
+
<td rowspan="6">Reasoning</td>
|
| 177 |
+
<td>Arc Challenge (25-shot)</td>
|
| 178 |
+
<td>34.1</td>
|
| 179 |
+
<td>47.0</td>
|
| 180 |
+
<td><b>47.6</b></td>
|
| 181 |
+
<td>45.9</td>
|
| 182 |
+
</tr>
|
| 183 |
+
<tr>
|
| 184 |
+
<td>GPQA (0-shot)</td>
|
| 185 |
+
<td>25.3</td>
|
| 186 |
+
<td><b>29.6</b></td>
|
| 187 |
+
<td>28.7</td>
|
| 188 |
+
<td>26.5</td>
|
| 189 |
+
</tr>
|
| 190 |
+
<tr>
|
| 191 |
+
<td>GPQA (0-shot, COT)</td>
|
| 192 |
+
<td>13.2</td>
|
| 193 |
+
<td>9.2</td>
|
| 194 |
+
<td>16.0</td>
|
| 195 |
+
<td><b>21.3</b></td>
|
| 196 |
+
</tr>
|
| 197 |
+
<tr>
|
| 198 |
+
<td>MUSR (0-shot)</td>
|
| 199 |
+
<td>32.4</td>
|
| 200 |
+
<td>36.8</td>
|
| 201 |
+
<td>33.0</td>
|
| 202 |
+
<td><b>40.7</b></td>
|
| 203 |
+
</tr>
|
| 204 |
+
<tr>
|
| 205 |
+
<td>BBH (3-shot)</td>
|
| 206 |
+
<td>30.3</td>
|
| 207 |
+
<td><b>38.5</b></td>
|
| 208 |
+
<td>33.1</td>
|
| 209 |
+
<td>35.1</td>
|
| 210 |
+
</tr>
|
| 211 |
+
<tr>
|
| 212 |
+
<td>BBH (3-shot, COT)</td>
|
| 213 |
+
<td>0.0</td>
|
| 214 |
+
<td>20.3</td>
|
| 215 |
+
<td>0.8</td>
|
| 216 |
+
<td><b>30.5</b></td>
|
| 217 |
+
</tr>
|
| 218 |
+
<tr>
|
| 219 |
+
<td rowspan="5">CommonSense Understanding</td>
|
| 220 |
+
<td>PIQA (0-shot)</td>
|
| 221 |
+
<td>72.1</td>
|
| 222 |
+
<td>73.2</td>
|
| 223 |
+
<td><b>74.4</b></td>
|
| 224 |
+
<td>72.0</td>
|
| 225 |
+
</tr>
|
| 226 |
+
<tr>
|
| 227 |
+
<td>SciQ (0-shot)</td>
|
| 228 |
+
<td>61.8</td>
|
| 229 |
+
<td>69.5</td>
|
| 230 |
+
<td>71.4</td>
|
| 231 |
+
<td><b>86.8</b></td>
|
| 232 |
+
</tr>
|
| 233 |
+
<tr>
|
| 234 |
+
<td>Winogrande (0-shot)</td>
|
| 235 |
+
<td>-</td>
|
| 236 |
+
<td>-</td>
|
| 237 |
+
<td>-</td>
|
| 238 |
+
<td><b>60.2</b></td>
|
| 239 |
+
</tr>
|
| 240 |
+
<tr>
|
| 241 |
+
<td>OpenbookQA (0-shot)</td>
|
| 242 |
+
<td>40.2</td>
|
| 243 |
+
<td>40.4</td>
|
| 244 |
+
<td><b>42.8</b></td>
|
| 245 |
+
<td>40.0</td>
|
| 246 |
+
</tr>
|
| 247 |
+
<tr>
|
| 248 |
+
<td>MT-Bench (avg)</td>
|
| 249 |
+
<td>5.4</td>
|
| 250 |
+
<td><b>7.1</b></td>
|
| 251 |
+
<td>6.1</td>
|
| 252 |
+
<td>5.5</td>
|
| 253 |
+
</tr>
|
| 254 |
+
<tr>
|
| 255 |
+
<td rowspan="1">Instructions following</td>
|
| 256 |
+
<td>Alapaca (WC)</td>
|
| 257 |
+
<td><b>8.6</b></td>
|
| 258 |
+
<td><b>8.6</b></td>
|
| 259 |
+
<td>5.4</td>
|
| 260 |
+
<td>6.1</td>
|
| 261 |
+
</tr>
|
| 262 |
+
</tbody>
|
| 263 |
+
</table>
|
| 264 |
|
| 265 |
### Testing Data, Factors & Metrics
|
| 266 |
|