Update README.md
Browse files
README.md
CHANGED
|
@@ -185,5 +185,85 @@ lm_eval \
|
|
| 185 |
### Accuracy
|
| 186 |
|
| 187 |
#### Open LLM Leaderboard evaluation scores
|
| 188 |
-
|
| 189 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185 |
### Accuracy
|
| 186 |
|
| 187 |
#### Open LLM Leaderboard evaluation scores
|
| 188 |
+
<table>
|
| 189 |
+
<tr>
|
| 190 |
+
<td><strong>Benchmark</strong>
|
| 191 |
+
</td>
|
| 192 |
+
<td><strong>Phi-3.5-mini-instruct</strong>
|
| 193 |
+
</td>
|
| 194 |
+
<td><strong>Phi-3.5-mini-instruct-FP8-KV(this model)</strong>
|
| 195 |
+
</td>
|
| 196 |
+
<td><strong>Recovery</strong>
|
| 197 |
+
</td>
|
| 198 |
+
</tr>
|
| 199 |
+
<tr>
|
| 200 |
+
<td>MMLU (5-shot)
|
| 201 |
+
</td>
|
| 202 |
+
<td>68.81
|
| 203 |
+
</td>
|
| 204 |
+
<td>68.56
|
| 205 |
+
</td>
|
| 206 |
+
<td>99.64%
|
| 207 |
+
</td>
|
| 208 |
+
</tr>
|
| 209 |
+
<tr>
|
| 210 |
+
<td>ARC Challenge (25-shot, acc_norm)
|
| 211 |
+
</td>
|
| 212 |
+
<td>64.68
|
| 213 |
+
</td>
|
| 214 |
+
<td>64.51
|
| 215 |
+
</td>
|
| 216 |
+
<td>99.74%
|
| 217 |
+
</td>
|
| 218 |
+
</tr>
|
| 219 |
+
<tr>
|
| 220 |
+
<td>GSM-8K (5-shot, strict-match)
|
| 221 |
+
</td>
|
| 222 |
+
<td>78.24
|
| 223 |
+
</td>
|
| 224 |
+
<td>77.26
|
| 225 |
+
</td>
|
| 226 |
+
<td>98.75%
|
| 227 |
+
</td>
|
| 228 |
+
</tr>
|
| 229 |
+
<tr>
|
| 230 |
+
<td>Hellaswag (10-shot, acc_norm)
|
| 231 |
+
</td>
|
| 232 |
+
<td>79.03
|
| 233 |
+
</td>
|
| 234 |
+
<td>78.88
|
| 235 |
+
</td>
|
| 236 |
+
<td>99.81%
|
| 237 |
+
</td>
|
| 238 |
+
</tr>
|
| 239 |
+
<tr>
|
| 240 |
+
<td>Winogrande (5-shot, acc)
|
| 241 |
+
</td>
|
| 242 |
+
<td>73.40
|
| 243 |
+
</td>
|
| 244 |
+
<td>73.80
|
| 245 |
+
</td>
|
| 246 |
+
<td>100.5%
|
| 247 |
+
</td>
|
| 248 |
+
</tr>
|
| 249 |
+
<tr>
|
| 250 |
+
<td>TruthfulQA (0-shot, mc2)
|
| 251 |
+
</td>
|
| 252 |
+
<td>56.39
|
| 253 |
+
</td>
|
| 254 |
+
<td>56.95
|
| 255 |
+
</td>
|
| 256 |
+
<td>100.9%
|
| 257 |
+
</td>
|
| 258 |
+
</tr>
|
| 259 |
+
<tr>
|
| 260 |
+
<td><strong>Average</strong>
|
| 261 |
+
</td>
|
| 262 |
+
<td><strong>70.09</strong>
|
| 263 |
+
</td>
|
| 264 |
+
<td><strong>70.00</strong>
|
| 265 |
+
</td>
|
| 266 |
+
<td><strong>99.89%</strong>
|
| 267 |
+
</td>
|
| 268 |
+
</tr>
|
| 269 |
+
</table>
|