Update README.md
Browse files
README.md
CHANGED
|
@@ -200,274 +200,9 @@ remaining weights from the LLM-guided Token Alignment and unfreeze all parameter
|
|
| 200 |
|
| 201 |
### Document Understanding Results
|
| 202 |
|
| 203 |
-
<
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
<th style="background-color: #d3d3d3;"><b>8B-Model</b></th>
|
| 207 |
-
<th>MM1.5</th>
|
| 208 |
-
<th>InfoVQA</th>
|
| 209 |
-
<th>DeepForm</th>
|
| 210 |
-
<th>ChartQA</th>
|
| 211 |
-
<th>TextVQA<sub>Val</sub></th>
|
| 212 |
-
<th>WTQ</th>
|
| 213 |
-
<th>TabFact</th>
|
| 214 |
-
<th>FUNSD</th>
|
| 215 |
-
<th>SROIE</th>
|
| 216 |
-
<th>KLC</th>
|
| 217 |
-
</tr>
|
| 218 |
-
</thead>
|
| 219 |
-
<tbody>
|
| 220 |
-
<tr>
|
| 221 |
-
<td style="background-color: #d3d3d3;"><b>Score</b></td>
|
| 222 |
-
<td>71.9</td>
|
| 223 |
-
<td>-</td>
|
| 224 |
-
<td>-</td>
|
| 225 |
-
<td>55.6</td>
|
| 226 |
-
<td>74.1</td>
|
| 227 |
-
<td>-</td>
|
| 228 |
-
<td>-</td>
|
| 229 |
-
<td>-</td>
|
| 230 |
-
<td>-</td>
|
| 231 |
-
<td>-</td>
|
| 232 |
-
</tr>
|
| 233 |
-
<tr>
|
| 234 |
-
<td>MiniCPM-V</td>
|
| 235 |
-
<td>3B</td>
|
| 236 |
-
<td>COLM'24</td>
|
| 237 |
-
<td>71.9</td>
|
| 238 |
-
<td>-</td>
|
| 239 |
-
<td>-</td>
|
| 240 |
-
<td>55.6</td>
|
| 241 |
-
<td>74.1</td>
|
| 242 |
-
<td>-</td>
|
| 243 |
-
<td>-</td>
|
| 244 |
-
<td>-</td>
|
| 245 |
-
</tr>
|
| 246 |
-
<tr>
|
| 247 |
-
<td>Mini-Monkey</td>
|
| 248 |
-
<td>2B</td>
|
| 249 |
-
<td>ICLR'25</td>
|
| 250 |
-
<td>87.4</td>
|
| 251 |
-
<td>60.1</td>
|
| 252 |
-
<td>-</td>
|
| 253 |
-
<td>76.5</td>
|
| 254 |
-
<td>75.7</td>
|
| 255 |
-
<td>-</td>
|
| 256 |
-
<td>-</td>
|
| 257 |
-
<td>42.9</td>
|
| 258 |
-
<td>70.3</td>
|
| 259 |
-
<td>-</td>
|
| 260 |
-
</tr>
|
| 261 |
-
<tr>
|
| 262 |
-
<td>InternVL2.5</td>
|
| 263 |
-
<td>2B</td>
|
| 264 |
-
<td>arxiv'24</td>
|
| 265 |
-
<td>88.7</td>
|
| 266 |
-
<td>60.9</td>
|
| 267 |
-
<td>15.2</td>
|
| 268 |
-
<td>79.2</td>
|
| 269 |
-
<td>74.3</td>
|
| 270 |
-
<td>38.7</td>
|
| 271 |
-
<td>58.1</td>
|
| 272 |
-
<td>37.9</td>
|
| 273 |
-
<td>68.1</td>
|
| 274 |
-
<td>16.1</td>
|
| 275 |
-
</tr>
|
| 276 |
-
<tr style="background-color: #d3d3d3;">
|
| 277 |
-
<td style="background-color: #d3d3d3;"><b>TokenVL w/o TA</b></td>
|
| 278 |
-
<td style="background-color: #d3d3d3;">8B</td>
|
| 279 |
-
<td style="background-color: #d3d3d3;">-</td>
|
| 280 |
-
<td style="background-color: #d3d3d3;"><u>93.8</u></td>
|
| 281 |
-
<td style="background-color: #d3d3d3;">75.3</td>
|
| 282 |
-
<td style="background-color: #d3d3d3;"><u>72.4</u></td>
|
| 283 |
-
<td style="background-color: #d3d3d3;"><u>86.5</u></td>
|
| 284 |
-
<td style="background-color: #d3d3d3;"><u>79.3</u></td>
|
| 285 |
-
<td style="background-color: #d3d3d3;"><u>57.2</u></td>
|
| 286 |
-
<td style="background-color: #d3d3d3;"><u>83.6</u></td>
|
| 287 |
-
<td style="background-color: #d3d3d3;"><u>41.5</u></td>
|
| 288 |
-
<td style="background-color: #d3d3d3;"><u>79.0</u></td>
|
| 289 |
-
<td style="background-color: #d3d3d3;"><u>39.6</u></td>
|
| 290 |
-
</tr>
|
| 291 |
-
<tr style="background-color: #d3d3d3;">
|
| 292 |
-
<td style="background-color: #d3d3d3;"><b>TokenVL</b></td>
|
| 293 |
-
<td style="background-color: #d3d3d3;">8B</td>
|
| 294 |
-
<td style="background-color: #d3d3d3;">-</td>
|
| 295 |
-
<td style="background-color: #d3d3d3;"><b>94.2</b></td>
|
| 296 |
-
<td style="background-color: #d3d3d3;"><u>76.5</u></td>
|
| 297 |
-
<td style="background-color: #d3d3d3;"><b>72.9</b></td>
|
| 298 |
-
<td style="background-color: #d3d3d3;"><b>86.6</b></td>
|
| 299 |
-
<td style="background-color: #d3d3d3;"><b>79.9</b></td>
|
| 300 |
-
<td style="background-color: #d3d3d3;"><b>61.4</b></td>
|
| 301 |
-
<td style="background-color: #d3d3d3;"><b>85.2</b></td>
|
| 302 |
-
<td style="background-color: #d3d3d3;"><b>42.2</b></td>
|
| 303 |
-
<td style="background-color: #d3d3d3;"><b>81.9</b></td>
|
| 304 |
-
<td style="background-color: #d3d3d3;"><b>39.9</b></td>
|
| 305 |
-
</tr>
|
| 306 |
-
</tbody>
|
| 307 |
-
</table>
|
| 308 |
-
|
| 309 |
-
<table>
|
| 310 |
-
<thead>
|
| 311 |
-
<tr>
|
| 312 |
-
<th>Model</th>
|
| 313 |
-
<th>size</th>
|
| 314 |
-
<th>Venue</th>
|
| 315 |
-
<th>DocVQA</th>
|
| 316 |
-
<th>InfoVQA</th>
|
| 317 |
-
<th>DeepForm</th>
|
| 318 |
-
<th>ChartQA</th>
|
| 319 |
-
<th>TextVQA<sub>Val</sub></th>
|
| 320 |
-
<th>WTQ</th>
|
| 321 |
-
<th>TabFact</th>
|
| 322 |
-
<th>FUNSD</th>
|
| 323 |
-
<th>SROIE</th>
|
| 324 |
-
<th>KLC</th>
|
| 325 |
-
</tr>
|
| 326 |
-
</thead>
|
| 327 |
-
<tbody>
|
| 328 |
-
<tr>
|
| 329 |
-
<td>Claude-3.5 Sonnet</td>
|
| 330 |
-
<td colspan="2" style="background-color: #d3d3d3;">Closed-source model</td>
|
| 331 |
-
<td>88.5</td>
|
| 332 |
-
<td>59.1</td>
|
| 333 |
-
<td>31.4</td>
|
| 334 |
-
<td>51.8</td>
|
| 335 |
-
<td>71.4</td>
|
| 336 |
-
<td>47.1</td>
|
| 337 |
-
<td>53.5</td>
|
| 338 |
-
<td>-</td>
|
| 339 |
-
<td>-</td>
|
| 340 |
-
<td>24.8</td>
|
| 341 |
-
</tr>
|
| 342 |
-
<tr>
|
| 343 |
-
<td>GeminiPro-1.5</td>
|
| 344 |
-
<td colspan="2" style="background-color: #d3d3d3;">Closed-source model</td>
|
| 345 |
-
<td>91.2</td>
|
| 346 |
-
<td>73.9</td>
|
| 347 |
-
<td>32.2</td>
|
| 348 |
-
<td>34.7</td>
|
| 349 |
-
<td>80.4</td>
|
| 350 |
-
<td>50.3</td>
|
| 351 |
-
<td>71.2</td>
|
| 352 |
-
<td>-</td>
|
| 353 |
-
<td>-</td>
|
| 354 |
-
<td>24.1</td>
|
| 355 |
-
</tr>
|
| 356 |
-
<tr>
|
| 357 |
-
<td>GPT4o 20240806</td>
|
| 358 |
-
<td colspan="2" style="background-color: #d3d3d3;">Closed-source model</td>
|
| 359 |
-
<td>92.8</td>
|
| 360 |
-
<td>66.4</td>
|
| 361 |
-
<td>38.4</td>
|
| 362 |
-
<td>85.7</td>
|
| 363 |
-
<td>70.5</td>
|
| 364 |
-
<td>46.6</td>
|
| 365 |
-
<td>81.1</td>
|
| 366 |
-
<td>-</td>
|
| 367 |
-
<td>-</td>
|
| 368 |
-
<td>29.9</td>
|
| 369 |
-
</tr>
|
| 370 |
-
</tbody>
|
| 371 |
-
</table>
|
| 372 |
-
|
| 373 |
-
<table>
|
| 374 |
-
<thead>
|
| 375 |
-
<tr>
|
| 376 |
-
<th>Model</th>
|
| 377 |
-
<th>size</th>
|
| 378 |
-
<th>Venue</th>
|
| 379 |
-
<th>DocVQA</th>
|
| 380 |
-
<th>InfoVQA</th>
|
| 381 |
-
<th>DeepForm</th>
|
| 382 |
-
<th>ChartQA</th>
|
| 383 |
-
<th>TextVQA<sub>Val</sub></th>
|
| 384 |
-
<th>WTQ</th>
|
| 385 |
-
<th>TabFact</th>
|
| 386 |
-
<th>FUNSD</th>
|
| 387 |
-
<th>SROIE</th>
|
| 388 |
-
<th>KLC</th>
|
| 389 |
-
</tr>
|
| 390 |
-
</thead>
|
| 391 |
-
<tbody>
|
| 392 |
-
<tr>
|
| 393 |
-
<td>DocPeida</td>
|
| 394 |
-
<td>7B</td>
|
| 395 |
-
<td>arxiv'23</td>
|
| 396 |
-
<td>47.1</td>
|
| 397 |
-
<td>15.2</td>
|
| 398 |
-
<td>-</td>
|
| 399 |
-
<td>46.9</td>
|
| 400 |
-
<td>60.2</td>
|
| 401 |
-
<td>-</td>
|
| 402 |
-
<td>-</td>
|
| 403 |
-
<td>29.9</td>
|
| 404 |
-
<td>21.4</td>
|
| 405 |
-
<td>-</td>
|
| 406 |
-
</tr>
|
| 407 |
-
<tr>
|
| 408 |
-
<td>DocOwl</td>
|
| 409 |
-
<td>7B</td>
|
| 410 |
-
<td>arxiv'23</td>
|
| 411 |
-
<td>62.2</td>
|
| 412 |
-
<td>38.2</td>
|
| 413 |
-
<td>42.6</td>
|
| 414 |
-
<td>57.4</td>
|
| 415 |
-
<td>52.6</td>
|
| 416 |
-
<td>26.9</td>
|
| 417 |
-
<td>67.6</td>
|
| 418 |
-
<td>0.5</td>
|
| 419 |
-
<td>1.7</td>
|
| 420 |
-
<td>30.3</td>
|
| 421 |
-
</tr>
|
| 422 |
-
<tr>
|
| 423 |
-
<td>LLaVA1.5</td>
|
| 424 |
-
<td>7B</td>
|
| 425 |
-
<td>NeurIPS'23</td>
|
| 426 |
-
<td>-</td>
|
| 427 |
-
<td>-</td>
|
| 428 |
-
<td>-</td>
|
| 429 |
-
<td>9.3</td>
|
| 430 |
-
<td>-</td>
|
| 431 |
-
<td>-</td>
|
| 432 |
-
<td>-</td>
|
| 433 |
-
<td>0.2</td>
|
| 434 |
-
<td>1.7</td>
|
| 435 |
-
<td>-</td>
|
| 436 |
-
</tr>
|
| 437 |
-
<tr>
|
| 438 |
-
<td>UReader</td>
|
| 439 |
-
<td>7B</td>
|
| 440 |
-
<td>EMNLP'23</td>
|
| 441 |
-
<td>65.4</td>
|
| 442 |
-
<td>42.2</td>
|
| 443 |
-
<td>49.5</td>
|
| 444 |
-
<td>59.3</td>
|
| 445 |
-
<td>57.6</td>
|
| 446 |
-
<td>29.4</td>
|
| 447 |
-
<td>67.6</td>
|
| 448 |
-
<td>-</td>
|
| 449 |
-
<td>-</td>
|
| 450 |
-
<td>32.8</td>
|
| 451 |
-
</tr>
|
| 452 |
-
<tr>
|
| 453 |
-
<td>DocPeida</td>
|
| 454 |
-
<td>7B</td>
|
| 455 |
-
<td>arxiv'23</td>
|
| 456 |
-
<td>47.1</td>
|
| 457 |
-
<td>15.2</td>
|
| 458 |
-
<td>-</td>
|
| 459 |
-
<td>46.9</td>
|
| 460 |
-
<td>60.2</td>
|
| 461 |
-
<td>-</td>
|
| 462 |
-
<td>-</td>
|
| 463 |
-
<td>29.9</td>
|
| 464 |
-
<td>21.4</td>
|
| 465 |
-
<td>-</td>
|
| 466 |
-
</tr>
|
| 467 |
-
<!-- Add additional rows as needed -->
|
| 468 |
-
</tbody>
|
| 469 |
-
</table>
|
| 470 |
-
|
| 471 |
|
| 472 |
|
| 473 |
|
|
|
|
| 200 |
|
| 201 |
### Document Understanding Results
|
| 202 |
|
| 203 |
+
<div align="center">
|
| 204 |
+
<img width="1300" alt="image" src="https://cdn-uploads.huggingface.co/production/uploads/650d4a36cbd0c7d550d3b41b/Msfs1YkDQHq2-djhm6QqD.png">
|
| 205 |
+
</div>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 206 |
|
| 207 |
|
| 208 |
|