Update README.md
Browse files
README.md
CHANGED
|
@@ -127,367 +127,385 @@ print(tokenizer.decode(output[0]))
|
|
| 127 |
|
| 128 |
# Evaluation
|
| 129 |
|
| 130 |
-
|
|
|
|
|
|
|
|
|
|
| 131 |
<table>
|
| 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 |
</table>
|
| 251 |
|
| 252 |
-
|
| 253 |
<table>
|
| 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 |
-
</tr>
|
| 370 |
-
<!-- Society & Culture -->
|
| 371 |
-
<tr>
|
| 372 |
-
<td rowspan="5"><b>Society & Culture</b></td>
|
| 373 |
-
<td><b>K-Refer*</b></td>
|
| 374 |
-
<td align="center">64.0</td>
|
| 375 |
-
<td align="center">53.6</td>
|
| 376 |
-
<td align="center">66.4</td>
|
| 377 |
-
<td align="center">71.6</td>
|
| 378 |
-
<td align="center">72.4</td>
|
| 379 |
-
<td align="center">43.2</td>
|
| 380 |
-
<td align="center"><b>89.6</b></td>
|
| 381 |
-
</tr>
|
| 382 |
-
<tr>
|
| 383 |
-
<td><b>K-Refer-Hard*</b></td>
|
| 384 |
-
<td align="center">67.1</td>
|
| 385 |
-
<td align="center">42.9</td>
|
| 386 |
-
<td align="center">61.4</td>
|
| 387 |
-
<td align="center">69.3</td>
|
| 388 |
-
<td align="center">65.7</td>
|
| 389 |
-
<td align="center">36.4</td>
|
| 390 |
-
<td align="center"><b>86.4</b></td>
|
| 391 |
-
</tr>
|
| 392 |
-
<tr>
|
| 393 |
-
<td><b>Ko-Sovereign*</b></td>
|
| 394 |
-
<td align="center">44.4</td>
|
| 395 |
-
<td align="center">35.8</td>
|
| 396 |
-
<td align="center">36.7</td>
|
| 397 |
-
<td align="center">46.9</td>
|
| 398 |
-
<td align="center"><b>49.8</b></td>
|
| 399 |
-
<td align="center">33.8</td>
|
| 400 |
-
<td align="center">56.3</td>
|
| 401 |
-
</tr>
|
| 402 |
-
<tr>
|
| 403 |
-
<td><b>HAERAE*</b></td>
|
| 404 |
-
<td align="center">61.3</td>
|
| 405 |
-
<td align="center">50.6</td>
|
| 406 |
-
<td align="center">70.8</td>
|
| 407 |
-
<td align="center">72.9</td>
|
| 408 |
-
<td align="center">68.4</td>
|
| 409 |
-
<td align="center">49.5</td>
|
| 410 |
-
<td align="center"><b>81.5</b></td>
|
| 411 |
-
</tr>
|
| 412 |
-
<tr>
|
| 413 |
-
<td><b>Avg.</b></td>
|
| 414 |
-
<td align="center">59.2</td>
|
| 415 |
-
<td align="center">45.7</td>
|
| 416 |
-
<td align="center">58.8</td>
|
| 417 |
-
<td align="center">65.2</td>
|
| 418 |
-
<td align="center">64.1</td>
|
| 419 |
-
<td align="center">40.7</td>
|
| 420 |
-
<td align="center"><b>78.4</b></td>
|
| 421 |
-
</tr>
|
| 422 |
-
<!-- Reasoning (Domain) -->
|
| 423 |
-
<tr>
|
| 424 |
-
<td rowspan="3"><b>Reasoning (Domain)</b></td>
|
| 425 |
-
<td><b>KMMLU</b></td>
|
| 426 |
-
<td align="center">43.5</td>
|
| 427 |
-
<td align="center">50.6</td>
|
| 428 |
-
<td align="center">45.1</td>
|
| 429 |
-
<td align="center">52.6</td>
|
| 430 |
-
<td align="center">55.4</td>
|
| 431 |
-
<td align="center">33.0</td>
|
| 432 |
-
<td align="center"><b>57.3</b></td>
|
| 433 |
-
</tr>
|
| 434 |
-
<tr>
|
| 435 |
-
<td><b>Ko-Sovereign*</b></td>
|
| 436 |
-
<td align="center">42.4</td>
|
| 437 |
-
<td align="center">42.5</td>
|
| 438 |
-
<td align="center">42.4</td>
|
| 439 |
-
<td align="center">45.6</td>
|
| 440 |
-
<td align="center">54.7</td>
|
| 441 |
-
<td align="center">36.7</td>
|
| 442 |
-
<td align="center"><b>58.0</b></td>
|
| 443 |
-
</tr>
|
| 444 |
-
<tr>
|
| 445 |
-
<td><b>Avg.</b></td>
|
| 446 |
-
<td align="center">43.0</td>
|
| 447 |
-
<td align="center">46.5</td>
|
| 448 |
-
<td align="center">43.8</td>
|
| 449 |
-
<td align="center">49.1</td>
|
| 450 |
-
<td align="center">55.1</td>
|
| 451 |
-
<td align="center">34.8</td>
|
| 452 |
-
<td align="center"><b>57.7</b></td>
|
| 453 |
-
</tr>
|
| 454 |
-
<!-- Instruction Following -->
|
| 455 |
-
<tr>
|
| 456 |
-
<td rowspan="3"><b>Instruction Following</b></td>
|
| 457 |
-
<td><b>Ko-IFEval*</b></td>
|
| 458 |
-
<td align="center">65.4</td>
|
| 459 |
-
<td align="center">75.9</td>
|
| 460 |
-
<td align="center">73.3</td>
|
| 461 |
-
<td align="center">69.1</td>
|
| 462 |
-
<td align="center"><b>83.6</b></td>
|
| 463 |
-
<td align="center">60.1</td>
|
| 464 |
-
<td align="center">82.0</td>
|
| 465 |
-
</tr>
|
| 466 |
-
<tr>
|
| 467 |
-
<td><b>Ko-MTBench</b></td>
|
| 468 |
-
<td align="center">74.0</td>
|
| 469 |
-
<td align="center">63.0</td>
|
| 470 |
-
<td align="center">74.0</td>
|
| 471 |
-
<td align="center">79.6</td>
|
| 472 |
-
<td align="center">71.0</td>
|
| 473 |
-
<td align="center">57.0</td>
|
| 474 |
-
<td align="center"><b>89.7</b></td>
|
| 475 |
-
</tr>
|
| 476 |
-
<tr>
|
| 477 |
-
<td><b>Avg.</b></td>
|
| 478 |
-
<td align="center">68.9</td>
|
| 479 |
-
<td align="center">69.4</td>
|
| 480 |
-
<td align="center">73.6</td>
|
| 481 |
-
<td align="center">74.4</td>
|
| 482 |
-
<td align="center">77.3</td>
|
| 483 |
-
<td align="center">58.5</td>
|
| 484 |
-
<td align="center"><b>85.9</b></td>
|
| 485 |
-
</tr>
|
| 486 |
-
</tbody>
|
| 487 |
</table>
|
| 488 |
|
| 489 |
`*` indicates KT proprietary evaluation resources.
|
| 490 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 491 |
<br>
|
| 492 |
|
| 493 |
# Usage
|
|
@@ -553,4 +571,5 @@ Mi:dm 2.0 is licensed under the [MIT License](./LICENSE).
|
|
| 553 |
### Contact
|
| 554 |
Mi:dm 2.0 Technical Inquiries: midm-llm@kt.com
|
| 555 |
|
| 556 |
-
<br>
|
|
|
|
|
|
| 127 |
|
| 128 |
# Evaluation
|
| 129 |
|
| 130 |
+
|
| 131 |
+
#### Korean
|
| 132 |
+
|
| 133 |
+
<!-- first half table-->
|
| 134 |
<table>
|
| 135 |
+
<tr>
|
| 136 |
+
<th rowspan="2">Model</th>
|
| 137 |
+
<th colspan="5" align="center">Society & Culture</th>
|
| 138 |
+
<th colspan="3" align="center">General Knowledge</th>
|
| 139 |
+
<th colspan="3" align="center">Instruction Following</th>
|
| 140 |
+
</tr>
|
| 141 |
+
<tr>
|
| 142 |
+
<th align="center">K-Refer<sup>*</sup></th>
|
| 143 |
+
<th align="center">K-Refer-Hard<sup>*</sup></th>
|
| 144 |
+
<th align="center">Ko-Sovereign<sup>*</sup></th>
|
| 145 |
+
<th align="center">HAERAE</th>
|
| 146 |
+
<th align="center">Avg.</th>
|
| 147 |
+
<th align="center">KMMLU</th>
|
| 148 |
+
<th align="center">Ko-Sovereign<sup>*</sup></th>
|
| 149 |
+
<th align="center">Avg.</th>
|
| 150 |
+
<th align="center">Ko-IFEval</th>
|
| 151 |
+
<th align="center">Ko-MTBench</th>
|
| 152 |
+
<th align="center">Avg.</th>
|
| 153 |
+
</tr>
|
| 154 |
+
|
| 155 |
+
<!-- Small Models -->
|
| 156 |
+
<tr>
|
| 157 |
+
<td><strong>Qwen3-4B</strong></td>
|
| 158 |
+
<td align="center">53.6</td>
|
| 159 |
+
<td align="center">42.9</td>
|
| 160 |
+
<td align="center">35.8</td>
|
| 161 |
+
<td align="center">50.6</td>
|
| 162 |
+
<td align="center">45.7</td>
|
| 163 |
+
<td align="center"><strong>50.6</strong></td>
|
| 164 |
+
<td align="center"><strong>42.5</strong></td>
|
| 165 |
+
<td align="center"><strong>46.5</strong></td>
|
| 166 |
+
<td align="center"><strong>75.9</strong></td>
|
| 167 |
+
<td align="center">63.0</td>
|
| 168 |
+
<td align="center">69.4</td>
|
| 169 |
+
</tr>
|
| 170 |
+
<tr>
|
| 171 |
+
<td><strong>Exaone-3.5-2.4B-inst</strong></td>
|
| 172 |
+
<td align="center">64.0</td>
|
| 173 |
+
<td align="center"><strong>67.1</strong></td>
|
| 174 |
+
<td align="center"><strong>44.4</strong></td>
|
| 175 |
+
<td align="center">61.3</td>
|
| 176 |
+
<td align="center"><strong>59.2</strong></td>
|
| 177 |
+
<td align="center">43.5</td>
|
| 178 |
+
<td align="center">42.4</td>
|
| 179 |
+
<td align="center">43.0</td>
|
| 180 |
+
<td align="center">65.4</td>
|
| 181 |
+
<td align="center"><u>74.0</u></td>
|
| 182 |
+
<td align="center">68.9</td>
|
| 183 |
+
</tr>
|
| 184 |
+
<tr>
|
| 185 |
+
<td><strong>Mi:dm 2.0-Mini-inst</strong></td>
|
| 186 |
+
<td align="center"><u>66.4</u></td>
|
| 187 |
+
<td align="center">61.4</td>
|
| 188 |
+
<td align="center">36.7</td>
|
| 189 |
+
<td align="center"><u>70.8</u></td>
|
| 190 |
+
<td align="center">58.8</td>
|
| 191 |
+
<td align="center">45.1</td>
|
| 192 |
+
<td align="center">42.4</td>
|
| 193 |
+
<td align="center">43.8</td>
|
| 194 |
+
<td align="center"><u>73.3</u></td>
|
| 195 |
+
<td align="center"><strong>74.0</strong></td>
|
| 196 |
+
<td align="center"><strong>73.6</strong></td>
|
| 197 |
+
</tr>
|
| 198 |
+
|
| 199 |
+
<!-- Spacer row -->
|
| 200 |
+
<tr><td colspan="13"> </td></tr>
|
| 201 |
+
|
| 202 |
+
<!-- Large Models -->
|
| 203 |
+
<tr>
|
| 204 |
+
<td><strong>Qwen3-14B</strong></td>
|
| 205 |
+
<td align="center"><u>72.4</u></td>
|
| 206 |
+
<td align="center">65.7</td>
|
| 207 |
+
<td align="center"><u>49.8</u></td>
|
| 208 |
+
<td align="center">68.4</td>
|
| 209 |
+
<td align="center">64.1</td>
|
| 210 |
+
<td align="center"><u>55.4</u></td>
|
| 211 |
+
<td align="center"><u>54.7</u></td>
|
| 212 |
+
<td align="center"><u>55.1</u></td>
|
| 213 |
+
<td align="center"><strong>83.6</strong></td>
|
| 214 |
+
<td align="center">71</td>
|
| 215 |
+
<td align="center"><u>77.3</u></td>
|
| 216 |
+
</tr>
|
| 217 |
+
<tr>
|
| 218 |
+
<td><strong>Llama-3.1-8B-inst</strong></td>
|
| 219 |
+
<td align="center">43.2</td>
|
| 220 |
+
<td align="center">36.4</td>
|
| 221 |
+
<td align="center">33.8</td>
|
| 222 |
+
<td align="center">49.5</td>
|
| 223 |
+
<td align="center">40.7</td>
|
| 224 |
+
<td align="center">33.0</td>
|
| 225 |
+
<td align="center">36.7</td>
|
| 226 |
+
<td align="center">34.8</td>
|
| 227 |
+
<td align="center">60.1</td>
|
| 228 |
+
<td align="center">57</td>
|
| 229 |
+
<td align="center">58.5</td>
|
| 230 |
+
</tr>
|
| 231 |
+
<tr>
|
| 232 |
+
<td><strong>Exaone-3.5-7.8B-inst</strong></td>
|
| 233 |
+
<td align="center">71.6</td>
|
| 234 |
+
<td align="center"><u>69.3</u></td>
|
| 235 |
+
<td align="center">46.9</td>
|
| 236 |
+
<td align="center"><u>72.9</u></td>
|
| 237 |
+
<td align="center"><u>65.2</u></td>
|
| 238 |
+
<td align="center">52.6</td>
|
| 239 |
+
<td align="center">45.6</td>
|
| 240 |
+
<td align="center">49.1</td>
|
| 241 |
+
<td align="center">69.1</td>
|
| 242 |
+
<td align="center"><u>79.6</u></td>
|
| 243 |
+
<td align="center">74.4</td>
|
| 244 |
+
</tr>
|
| 245 |
+
<tr>
|
| 246 |
+
<td><strong>Mi:dm 2.0-Base-inst</strong></td>
|
| 247 |
+
<td align="center"><strong>89.6</strong></td>
|
| 248 |
+
<td align="center"><strong>86.4</strong></td>
|
| 249 |
+
<td align="center"><strong>56.3</strong></td>
|
| 250 |
+
<td align="center"><strong>81.5</strong></td>
|
| 251 |
+
<td align="center"><strong>78.4</strong></td>
|
| 252 |
+
<td align="center"><strong>57.3</strong></td>
|
| 253 |
+
<td align="center"><strong>58.0</strong></td>
|
| 254 |
+
<td align="center"><strong>57.7</strong></td>
|
| 255 |
+
<td align="center"><u>82</u></td>
|
| 256 |
+
<td align="center"><strong>89.7</strong></td>
|
| 257 |
+
<td align="center"><strong>85.9</strong></td>
|
| 258 |
+
</tr>
|
| 259 |
</table>
|
| 260 |
|
| 261 |
+
<!-- second half table-->
|
| 262 |
<table>
|
| 263 |
+
<tr>
|
| 264 |
+
<th rowspan="2" align="center">Model</th>
|
| 265 |
+
<th colspan="5" align="center">Comprehension</th>
|
| 266 |
+
<th colspan="5" align="center">Reasoning</th>
|
| 267 |
+
</tr>
|
| 268 |
+
<tr>
|
| 269 |
+
<th align="center">K-Prag<sup>*</sup></th>
|
| 270 |
+
<th align="center">K-Refer-Hard<sup>*</sup></th>
|
| 271 |
+
<th align="center">Ko-Best</th>
|
| 272 |
+
<th align="center">Ko-Sovereign<sup>*</sup></th>
|
| 273 |
+
<th align="center">Avg.</th>
|
| 274 |
+
<th align="center">Ko-Winogrande</th>
|
| 275 |
+
<th align="center">Ko-Best</th>
|
| 276 |
+
<th align="center">LogicKor</th>
|
| 277 |
+
<th align="center">HRM8K</th>
|
| 278 |
+
<th align="center">Avg.</th>
|
| 279 |
+
</tr>
|
| 280 |
+
|
| 281 |
+
<!-- Small Models -->
|
| 282 |
+
<tr>
|
| 283 |
+
<td><strong>Qwen3-4B</strong></td>
|
| 284 |
+
<td align="center"><strong>73.9<strong></td>
|
| 285 |
+
<td align="center">56.7</td>
|
| 286 |
+
<td align="center"><strong>91.5</strong></td>
|
| 287 |
+
<td align="center"><strong>43.5</strong></td>
|
| 288 |
+
<td align="center"><strong>66.6</strong></td>
|
| 289 |
+
<td align="center"><strong>67.5</strong></td>
|
| 290 |
+
<td align="center"><strong>69.2</strong></td>
|
| 291 |
+
<td align="center">5.6</td>
|
| 292 |
+
<td align="center"><strong>56.7</strong></td>
|
| 293 |
+
<td align="center"><strong>43.8</strong></td>
|
| 294 |
+
</tr>
|
| 295 |
+
<tr>
|
| 296 |
+
<td><strong>Exaone-3.5-2.4B-inst</strong></td>
|
| 297 |
+
<td align="center">68.7</td>
|
| 298 |
+
<td align="center"><strong>58.5</strong></td>
|
| 299 |
+
<td align="center"><u>87.2</u></td>
|
| 300 |
+
<td align="center">38.0</td>
|
| 301 |
+
<td align="center"><u>62.5</u></td>
|
| 302 |
+
<td align="center">60.3</td>
|
| 303 |
+
<td align="center">64.1</td>
|
| 304 |
+
<td align="center">7.4</td>
|
| 305 |
+
<td align="center">38.5</td>
|
| 306 |
+
<td align="center">36.7</td>
|
| 307 |
+
</tr>
|
| 308 |
+
<tr>
|
| 309 |
+
<td><strong>Mi:dm 2.0-Mini-inst</strong></td>
|
| 310 |
+
<td align="center">69.5</td>
|
| 311 |
+
<td align="center">55.4</td>
|
| 312 |
+
<td align="center">80.5</td>
|
| 313 |
+
<td align="center">42.5</td>
|
| 314 |
+
<td align="center">61.9</td>
|
| 315 |
+
<td align="center"><u>61.7</u></td>
|
| 316 |
+
<td align="center"><u>64.5</u></td>
|
| 317 |
+
<td align="center"><strong>7.7</strong></td>
|
| 318 |
+
<td align="center"><u>39.9</u></td>
|
| 319 |
+
<td align="center"><u>37.4</u></td>
|
| 320 |
+
</tr>
|
| 321 |
+
|
| 322 |
+
<!-- Visual Spacer -->
|
| 323 |
+
<tr><td colspan="11"> </td></tr>
|
| 324 |
+
|
| 325 |
+
<!-- Large Models -->
|
| 326 |
+
<tr>
|
| 327 |
+
<td><strong>Qwen3-14B</strong></td>
|
| 328 |
+
<td align="center"><strong>86.7</strong></td>
|
| 329 |
+
<td align="center"><strong>74.0</strong></td>
|
| 330 |
+
<td align="center">93.9</td>
|
| 331 |
+
<td align="center">52.0</td>
|
| 332 |
+
<td align="center"><strong>76.8</strong></td>
|
| 333 |
+
<td align="center"><strong>77.2</strong></td>
|
| 334 |
+
<td align="center"><strong>75.4</strong></td>
|
| 335 |
+
<td align="center">6.4</td>
|
| 336 |
+
<td align="center"><strong>64.5</strong></td>
|
| 337 |
+
<td align="center"><strong>48.8</strong></td>
|
| 338 |
+
</tr>
|
| 339 |
+
<tr>
|
| 340 |
+
<td><strong>Llama-3.1-8B-inst</strong></td>
|
| 341 |
+
<td align="center">59.9</td>
|
| 342 |
+
<td align="center">48.6</td>
|
| 343 |
+
<td align="center">77.4</td>
|
| 344 |
+
<td align="center">31.5</td>
|
| 345 |
+
<td align="center">51.5</td>
|
| 346 |
+
<td align="center">40.1</td>
|
| 347 |
+
<td align="center">26.0</td>
|
| 348 |
+
<td align="center">2.4</td>
|
| 349 |
+
<td align="center">30.9</td>
|
| 350 |
+
<td align="center">19.8</td>
|
| 351 |
+
</tr>
|
| 352 |
+
<tr>
|
| 353 |
+
<td><strong>Exaone-3.5-7.8B-inst</strong></td>
|
| 354 |
+
<td align="center"><u>73.5</u></td>
|
| 355 |
+
<td align="center"><u>61.9</u></td>
|
| 356 |
+
<td align="center"><u>92.0</u></td>
|
| 357 |
+
<td align="center">44.0</td>
|
| 358 |
+
<td align="center">67.2</td>
|
| 359 |
+
<td align="center">64.6</td>
|
| 360 |
+
<td align="center">60.3</td>
|
| 361 |
+
<td align="center"><strong>8.6</strong></td>
|
| 362 |
+
<td align="center">49.7</td>
|
| 363 |
+
<td align="center">39.5</td>
|
| 364 |
+
</tr>
|
| 365 |
+
<tr>
|
| 366 |
+
<td><strong>Mi:dm 2.0-Base-inst</strong></td>
|
| 367 |
+
<td align="center"><u>86.5</u></td>
|
| 368 |
+
<td align="center"><u>70.8</u></td>
|
| 369 |
+
<td align="center"><strong>95.2</strong></td>
|
| 370 |
+
<td align="center"><strong>53.0</strong></td>
|
| 371 |
+
<td align="center"><u>76.1</u></td>
|
| 372 |
+
<td align="center"><u>75.1</u></td>
|
| 373 |
+
<td align="center"><u>73.0</u></td>
|
| 374 |
+
<td align="center"><strong>8.6</strong></td>
|
| 375 |
+
<td align="center"><u>52.9</u></td>
|
| 376 |
+
<td align="center"><u>44.8</u></td>
|
| 377 |
+
</tr>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 378 |
</table>
|
| 379 |
|
| 380 |
`*` indicates KT proprietary evaluation resources.
|
| 381 |
|
| 382 |
+
<br>
|
| 383 |
+
|
| 384 |
+
|
| 385 |
+
#### English
|
| 386 |
+
|
| 387 |
+
|
| 388 |
+
<table>
|
| 389 |
+
<tr>
|
| 390 |
+
<th rowspan="2" align="center">Model</th>
|
| 391 |
+
<th align="center">Instruction</th>
|
| 392 |
+
<th colspan="4" align="center">Reasoning</th>
|
| 393 |
+
<th align="center">Math</th>
|
| 394 |
+
<th align="center">Coding</th>
|
| 395 |
+
<th colspan="3" align="center">General Knowledge</th>
|
| 396 |
+
</tr>
|
| 397 |
+
<tr>
|
| 398 |
+
<th align="center">IFEval</th>
|
| 399 |
+
<th align="center">BBH</th>
|
| 400 |
+
<th align="center">GPQA</th>
|
| 401 |
+
<th align="center">MuSR</th>
|
| 402 |
+
<th align="center">Avg.</th>
|
| 403 |
+
<th align="center">GSM8K</th>
|
| 404 |
+
<th align="center">MBPP+</th>
|
| 405 |
+
<th align="center">MMLU-pro</th>
|
| 406 |
+
<th align="center">MMLU</th>
|
| 407 |
+
<th align="center">Avg.</th>
|
| 408 |
+
</tr>
|
| 409 |
+
|
| 410 |
+
<!-- Small Models -->
|
| 411 |
+
<tr>
|
| 412 |
+
<td><strong>Qwen3-4B</strong></td>
|
| 413 |
+
<td align="center"><u>79.7</u></td>
|
| 414 |
+
<td align="center"><strong>79.0</strong></td>
|
| 415 |
+
<td align="center"><u>39.8</u></td>
|
| 416 |
+
<td align="center"><strong>58.5</strong></td>
|
| 417 |
+
<td align="center"><strong>59.1</strong></td>
|
| 418 |
+
<td align="center"><strong>90.4</strong></td>
|
| 419 |
+
<td align="center"><u>62.4</u></td>
|
| 420 |
+
<td align="center">-</td>
|
| 421 |
+
<td align="center"><strong>73.3</strong></td>
|
| 422 |
+
<td align="center"><strong>73.3</strong></td>
|
| 423 |
+
</tr>
|
| 424 |
+
<tr>
|
| 425 |
+
<td><strong>Exaone-3.5-2.4B-inst</strong></td>
|
| 426 |
+
<td align="center"><strong>81.1</strong></td>
|
| 427 |
+
<td align="center">46.4</td>
|
| 428 |
+
<td align="center">28.1</td>
|
| 429 |
+
<td align="center">49.7</td>
|
| 430 |
+
<td align="center">41.4</td>
|
| 431 |
+
<td align="center"><u>82.5</u></td>
|
| 432 |
+
<td align="center">59.8</td>
|
| 433 |
+
<td align="center">-</td>
|
| 434 |
+
<td align="center">59.5</td>
|
| 435 |
+
<td align="center">59.5</td>
|
| 436 |
+
</tr>
|
| 437 |
+
<tr>
|
| 438 |
+
<td><strong>Mi:dm 2.0-Mini-inst</strong></td>
|
| 439 |
+
<td align="center">73.6</td>
|
| 440 |
+
<td align="center"><u>44.5</u></td>
|
| 441 |
+
<td align="center">26.6</td>
|
| 442 |
+
<td align="center"><u>51.7</u></td>
|
| 443 |
+
<td align="center"><u>40.9</u></td>
|
| 444 |
+
<td align="center">83.1</td>
|
| 445 |
+
<td align="center"><strong>60.9</strong></td>
|
| 446 |
+
<td align="center">-</td>
|
| 447 |
+
<td align="center">56.5</td>
|
| 448 |
+
<td align="center">56.5</td>
|
| 449 |
+
</tr>
|
| 450 |
+
|
| 451 |
+
<tr><td colspan="11"> </td></tr>
|
| 452 |
+
|
| 453 |
+
<!-- Large Models -->
|
| 454 |
+
<tr>
|
| 455 |
+
<td><strong>Qwen3-14B</strong></td>
|
| 456 |
+
<td align="center"><u>83.9</u></td>
|
| 457 |
+
<td align="center"><strong>83.4</strong></td>
|
| 458 |
+
<td align="center"><strong>49.8</strong></td>
|
| 459 |
+
<td align="center"><strong>57.7</strong></td>
|
| 460 |
+
<td align="center"><strong>63.6</strong></td>
|
| 461 |
+
<td align="center">88.0</td>
|
| 462 |
+
<td align="center">73.4</td>
|
| 463 |
+
<td align="center"><strong>70.5</strong></td>
|
| 464 |
+
<td align="center"><strong>82.7</strong></td>
|
| 465 |
+
<td align="center"><strong>76.6</strong></td>
|
| 466 |
+
</tr>
|
| 467 |
+
<tr>
|
| 468 |
+
<td><strong>Llama-3.1-8B-inst</strong></td>
|
| 469 |
+
<td align="center">79.9</td>
|
| 470 |
+
<td align="center"><u>60.3</u></td>
|
| 471 |
+
<td align="center">21.6</td>
|
| 472 |
+
<td align="center">50.3</td>
|
| 473 |
+
<td align="center">44.1</td>
|
| 474 |
+
<td align="center">81.2</td>
|
| 475 |
+
<td align="center"><strong>81.8</strong></td>
|
| 476 |
+
<td align="center">47.6</td>
|
| 477 |
+
<td align="center"><u>70.7</u></td>
|
| 478 |
+
<td align="center"><u>59.2</u></td>
|
| 479 |
+
</tr>
|
| 480 |
+
<tr>
|
| 481 |
+
<td><strong>Exaone-3.5-7.8B-inst</strong></td>
|
| 482 |
+
<td align="center">83.6</td>
|
| 483 |
+
<td align="center">50.1</td>
|
| 484 |
+
<td align="center"><u>33.1</u></td>
|
| 485 |
+
<td align="center"><u>51.2</u></td>
|
| 486 |
+
<td align="center"><u>44.8</u></td>
|
| 487 |
+
<td align="center">81.1</td>
|
| 488 |
+
<td align="center">79.4</td>
|
| 489 |
+
<td align="center">40.7</td>
|
| 490 |
+
<td align="center">69.0</td>
|
| 491 |
+
<td align="center">54.8</td>
|
| 492 |
+
</tr>
|
| 493 |
+
<tr>
|
| 494 |
+
<td><strong>Mi:dm 2.0-Base-inst</strong></td>
|
| 495 |
+
<td align="center"><strong>84.0</strong></td>
|
| 496 |
+
<td align="center">77.7</td>
|
| 497 |
+
<td align="center">33.5</td>
|
| 498 |
+
<td align="center">51.9</td>
|
| 499 |
+
<td align="center">54.4</td>
|
| 500 |
+
<td align="center"><strong>91.6</strong></td>
|
| 501 |
+
<td align="center"><u>77.5</u></td>
|
| 502 |
+
<td align="center"><u>53.3</u></td>
|
| 503 |
+
<td align="center">73.7</td>
|
| 504 |
+
<td align="center">63.5</td>
|
| 505 |
+
</tr>
|
| 506 |
+
</table>
|
| 507 |
+
|
| 508 |
+
|
| 509 |
<br>
|
| 510 |
|
| 511 |
# Usage
|
|
|
|
| 571 |
### Contact
|
| 572 |
Mi:dm 2.0 Technical Inquiries: midm-llm@kt.com
|
| 573 |
|
| 574 |
+
<br>
|
| 575 |
+
|