File size: 16,380 Bytes
b0c0df0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 |
{
"cells": [
{
"cell_type": "code",
"execution_count": 98,
"metadata": {},
"outputs": [],
"source": [
"model_1 = \"gpt-4o\"\n",
"model_2 = \"Qwen-VL-72B-Instruct\"\n",
"\n",
"model_1 = model_1.replace(\"-\", \"_\")\n",
"model_2 = model_2.replace(\"-\", \"_\")"
]
},
{
"cell_type": "code",
"execution_count": 99,
"metadata": {},
"outputs": [],
"source": [
"import datasets\n",
"\n",
"data = datasets.load_dataset(\"lmms-lab/LiveBenchDetailedResults\", \"2024-09\")"
]
},
{
"cell_type": "code",
"execution_count": 100,
"metadata": {},
"outputs": [],
"source": [
"model_data_1 = data[model_1].to_pandas()\n",
"model_data_2 = data[model_2].to_pandas()"
]
},
{
"cell_type": "code",
"execution_count": 101,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"Index(['id', 'images', 'question', 'ground_truth', 'criteria', 'subtask',\n",
" 'website', 'response', 'score', 'reason'],\n",
" dtype='object')"
]
},
"execution_count": 101,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model_data_1.columns"
]
},
{
"cell_type": "code",
"execution_count": 102,
"metadata": {},
"outputs": [],
"source": [
"same_data = [\"images\", \"question\", \"ground_truth\", \"criteria\", \"subtask\", \"website\"]\n",
"model_data_2.drop(columns=same_data, inplace=True)\n",
"merged_data = model_data_1.merge(model_data_2, on=\"id\", suffixes=(f\"_1\", f\"_2\"))"
]
},
{
"cell_type": "code",
"execution_count": 103,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>id</th>\n",
" <th>images</th>\n",
" <th>question</th>\n",
" <th>ground_truth</th>\n",
" <th>criteria</th>\n",
" <th>subtask</th>\n",
" <th>website</th>\n",
" <th>response_1</th>\n",
" <th>score_1</th>\n",
" <th>reason_1</th>\n",
" <th>response_2</th>\n",
" <th>score_2</th>\n",
" <th>reason_2</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>41</td>\n",
" <td>[{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH...</td>\n",
" <td>Identify each vulnerability's CVE identifier a...</td>\n",
" <td>1. CVE Identifier: CVE-2024-29824\\nDescription...</td>\n",
" <td>Total score: 10 points.\\n\\nScoring:\\n- For eac...</td>\n",
" <td>Concrete Recognition</td>\n",
" <td>technology</td>\n",
" <td>Here are the vulnerabilities along with their ...</td>\n",
" <td>10.00</td>\n",
" <td>The assistant response correctly identified al...</td>\n",
" <td>1. **CVE-2024-29824**\\n - **Description**: I...</td>\n",
" <td>10.00</td>\n",
" <td>The assistant response correctly identifies ea...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>274</td>\n",
" <td>[{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH...</td>\n",
" <td>List all the news headlines in the order they ...</td>\n",
" <td>1. 'I'm amazed I lived to tell the tale': Awar...</td>\n",
" <td>0-1 correct: 0 points, 2-4 correct: 4 points, ...</td>\n",
" <td>Concrete Recognition</td>\n",
" <td>artandculture</td>\n",
" <td>1. **‘I’m amazed I lived to tell the tale’: Aw...</td>\n",
" <td>10.00</td>\n",
" <td>The assistant response correctly matches all t...</td>\n",
" <td></td>\n",
" <td>0.00</td>\n",
" <td>No response</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>215</td>\n",
" <td>[{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH...</td>\n",
" <td>List all the news headlines related to the 202...</td>\n",
" <td>Here are the news headlines related to the 202...</td>\n",
" <td>The total score is 10 points. Listing each hea...</td>\n",
" <td>Concrete Recognition</td>\n",
" <td>sports</td>\n",
" <td>Here are the news headlines related to the 202...</td>\n",
" <td>4.29</td>\n",
" <td>The assistant response correctly identified th...</td>\n",
" <td>1. \"Who will be the No. 1 pick in the 2025 NBA...</td>\n",
" <td>8.58</td>\n",
" <td>The assistant response correctly identified 6 ...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>215</td>\n",
" <td>[{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH...</td>\n",
" <td>List all the news headlines related to the 202...</td>\n",
" <td>Here are the news headlines related to the 202...</td>\n",
" <td>The total score is 10 points. Listing each hea...</td>\n",
" <td>Concrete Recognition</td>\n",
" <td>sports</td>\n",
" <td>Here are the news headlines related to the 202...</td>\n",
" <td>4.29</td>\n",
" <td>The assistant response correctly identified th...</td>\n",
" <td>```markdown\\n# Interview with Abdoulaye Trésor...</td>\n",
" <td>10.00</td>\n",
" <td>The assistant's response accurately captures t...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>18</td>\n",
" <td>[{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH...</td>\n",
" <td>Identify and list the three formats provided b...</td>\n",
" <td>The three formats provided by the new annotati...</td>\n",
" <td>To score the response:\\n\\n- Award 10 points if...</td>\n",
" <td>Concrete Recognition</td>\n",
" <td>technology</td>\n",
" <td>The new `annotationlib` module in Python 3.14 ...</td>\n",
" <td>10.00</td>\n",
" <td>The assistant correctly identified and describ...</td>\n",
" <td>The three formats provided by the new annotati...</td>\n",
" <td>10.00</td>\n",
" <td>The assistant correctly identified and describ...</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" id images \\\n",
"0 41 [{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH... \n",
"1 274 [{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH... \n",
"2 215 [{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH... \n",
"3 215 [{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH... \n",
"4 18 [{'bytes': b'\\x89PNG\\r\\n\\x1a\\n\\x00\\x00\\x00\\rIH... \n",
"\n",
" question \\\n",
"0 Identify each vulnerability's CVE identifier a... \n",
"1 List all the news headlines in the order they ... \n",
"2 List all the news headlines related to the 202... \n",
"3 List all the news headlines related to the 202... \n",
"4 Identify and list the three formats provided b... \n",
"\n",
" ground_truth \\\n",
"0 1. CVE Identifier: CVE-2024-29824\\nDescription... \n",
"1 1. 'I'm amazed I lived to tell the tale': Awar... \n",
"2 Here are the news headlines related to the 202... \n",
"3 Here are the news headlines related to the 202... \n",
"4 The three formats provided by the new annotati... \n",
"\n",
" criteria subtask \\\n",
"0 Total score: 10 points.\\n\\nScoring:\\n- For eac... Concrete Recognition \n",
"1 0-1 correct: 0 points, 2-4 correct: 4 points, ... Concrete Recognition \n",
"2 The total score is 10 points. Listing each hea... Concrete Recognition \n",
"3 The total score is 10 points. Listing each hea... Concrete Recognition \n",
"4 To score the response:\\n\\n- Award 10 points if... Concrete Recognition \n",
"\n",
" website response_1 score_1 \\\n",
"0 technology Here are the vulnerabilities along with their ... 10.00 \n",
"1 artandculture 1. **‘I’m amazed I lived to tell the tale’: Aw... 10.00 \n",
"2 sports Here are the news headlines related to the 202... 4.29 \n",
"3 sports Here are the news headlines related to the 202... 4.29 \n",
"4 technology The new `annotationlib` module in Python 3.14 ... 10.00 \n",
"\n",
" reason_1 \\\n",
"0 The assistant response correctly identified al... \n",
"1 The assistant response correctly matches all t... \n",
"2 The assistant response correctly identified th... \n",
"3 The assistant response correctly identified th... \n",
"4 The assistant correctly identified and describ... \n",
"\n",
" response_2 score_2 \\\n",
"0 1. **CVE-2024-29824**\\n - **Description**: I... 10.00 \n",
"1 0.00 \n",
"2 1. \"Who will be the No. 1 pick in the 2025 NBA... 8.58 \n",
"3 ```markdown\\n# Interview with Abdoulaye Trésor... 10.00 \n",
"4 The three formats provided by the new annotati... 10.00 \n",
"\n",
" reason_2 \n",
"0 The assistant response correctly identifies ea... \n",
"1 No response \n",
"2 The assistant response correctly identified 6 ... \n",
"3 The assistant's response accurately captures t... \n",
"4 The assistant correctly identified and describ... "
]
},
"execution_count": 103,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"merged_data.head()"
]
},
{
"cell_type": "code",
"execution_count": 104,
"metadata": {},
"outputs": [],
"source": [
"delta = merged_data[\"score_1\"] - merged_data[\"score_2\"]"
]
},
{
"cell_type": "code",
"execution_count": 105,
"metadata": {},
"outputs": [],
"source": [
"merged_data[\"delta\"] = delta"
]
},
{
"cell_type": "code",
"execution_count": 106,
"metadata": {},
"outputs": [],
"source": [
"sorted_data = merged_data.sort_values(\"delta\", ascending=False)"
]
},
{
"cell_type": "code",
"execution_count": 107,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"Index(['id', 'images', 'question', 'ground_truth', 'criteria', 'subtask',\n",
" 'website', 'response_1', 'score_1', 'reason_1', 'response_2', 'score_2',\n",
" 'reason_2', 'delta'],\n",
" dtype='object')"
]
},
"execution_count": 107,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"sorted_data.columns"
]
},
{
"cell_type": "code",
"execution_count": 108,
"metadata": {},
"outputs": [],
"source": [
"features = data[model_1].features.copy()"
]
},
{
"cell_type": "code",
"execution_count": 109,
"metadata": {},
"outputs": [],
"source": [
"for feature in [\"score\", \"reason\", \"response\"]:\n",
" feat = features.pop(feature)\n",
" features[f\"{feature}_1\"] = feat\n",
" features[f\"{feature}_2\"] = feat"
]
},
{
"cell_type": "code",
"execution_count": 110,
"metadata": {},
"outputs": [],
"source": [
"features[\"delta\"] = datasets.Value(\"float32\")"
]
},
{
"cell_type": "code",
"execution_count": 111,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{'id': Value(dtype='int32', id=None),\n",
" 'images': Sequence(feature=Image(mode=None, decode=True, id=None), length=-1, id=None),\n",
" 'question': Value(dtype='string', id=None),\n",
" 'ground_truth': Value(dtype='string', id=None),\n",
" 'criteria': Value(dtype='string', id=None),\n",
" 'subtask': Value(dtype='string', id=None),\n",
" 'website': Value(dtype='string', id=None),\n",
" 'score_1': Value(dtype='float32', id=None),\n",
" 'score_2': Value(dtype='float32', id=None),\n",
" 'reason_1': Value(dtype='string', id=None),\n",
" 'reason_2': Value(dtype='string', id=None),\n",
" 'response_1': Value(dtype='string', id=None),\n",
" 'response_2': Value(dtype='string', id=None),\n",
" 'delta': Value(dtype='float32', id=None)}"
]
},
"execution_count": 111,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"features"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": 112,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Generating train split: 218 examples [00:00, 329.76 examples/s]\n"
]
}
],
"source": [
"def gen():\n",
" for i, row in sorted_data.iterrows():\n",
" yield row.to_dict()\n",
"\n",
"\n",
"final_data = datasets.Dataset.from_generator(gen, features=features)"
]
},
{
"cell_type": "code",
"execution_count": 113,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Map: 100%|██████████| 218/218 [00:00<00:00, 497.97 examples/s]it/s]\n",
"Creating parquet from Arrow format: 100%|██████████| 3/3 [00:00<00:00, 7.03ba/s]\n",
"Uploading the dataset shards: 100%|██████████| 1/1 [00:14<00:00, 14.44s/it]\n"
]
},
{
"data": {
"text/plain": [
"CommitInfo(commit_url='https://huggingface.co/datasets/lmms-lab/LiveBenchDetailedResultsComparison/commit/f0941a0afeb32f8f0aae1861e6ef658a237bb249', commit_message='Upload dataset', commit_description='', oid='f0941a0afeb32f8f0aae1861e6ef658a237bb249', pr_url=None, repo_url=RepoUrl('https://huggingface.co/datasets/lmms-lab/LiveBenchDetailedResultsComparison', endpoint='https://huggingface.co', repo_type='dataset', repo_id='lmms-lab/LiveBenchDetailedResultsComparison'), pr_revision=None, pr_num=None)"
]
},
"execution_count": 113,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"final_data.push_to_hub(\"lmms-lab/LiveBenchDetailedResultsComparison\", \"2024-09\", split=f\"{model_1}_vs_{model_2}\")"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "live_bench",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.7"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
|