Skip to content

Instantly share code, notes, and snippets.

@yoeven

yoeven/vocr_result.json

Created March 7, 2025 11:54
Show Gist options
  • Save yoeven/a6bccae45a7d133130a6e1486e0df4ac to your computer and use it in GitHub Desktop.
Save yoeven/a6bccae45a7d133130a6e1486e0df4ac to your computer and use it in GitHub Desktop.
Download ZIP
vocr_result.json
Raw
vocr_result.json
{
"success": true,
"context": {
"code_url": [
"https://github.com/mistralai/mistral-src"
],
"markdown": [
"https://mistral.ai/news/announcing-mistral-7b/"
]
},
"width": 1500,
"height": 1999,
"tags": [
"text",
"printing",
"paper",
"font",
"menu",
"document"
],
"has_text": true,
"sections": [
{
"text": "Mistral 7B\nAlbert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford,\nDevendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel,\nGuillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux,\nPierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,\nWilliam El Sayed\nMistral Al\nAbstract\nWe introduce Mistral 7B, a 7-billion-parameter language model engineered for\nsuperior performance and efficiency. Mistral 7B outperforms the best open 13B\nmodel (Llama 2) across all evaluated benchmarks, and the best released 34B\nmodel (Llama 1) in reasoning, mathematics, and code generation. Our model\nleverages grouped-query attention (GQA) for faster inference, coupled with sliding\nwindow attention (SWA) to effectively handle sequences of arbitrary length with a\nreduced inference cost. We also provide a model fine-tuned to follow instructions,\nMistral 7B - Instruct, that surpasses Llama 2 13B - chat model both on human and\nautomated benchmarks. Our models are released under the Apache 2.0 license.\nCode: https://github.com/mistralai/mistral-src\nWebpage: https://mistral.ai/news/announcing-mistral-7b/\n1 Introduction\nIn the rapidly evolving domain of Natural Language Processing (NLP), the race towards higher model\nperformance often necessitates an escalation in model size. However, this scaling tends to increase\ncomputational costs and inference latency, thereby raising barriers to deployment in practical,\nreal-world scenarios. In this context, the search for balanced models delivering both high-level\nperformance and efficiency becomes critically essential. Our model, Mistral 7B, demonstrates that\na carefully designed language model can deliver high performance while maintaining an efficient\ninference. Mistral 7B outperforms the previous best 13B model (Llama 2, [26]) across all tested\nbenchmarks, and surpasses the best 34B model (LLaMa 34B, [25]) in mathematics and code\ngeneration. Furthermore, Mistral 7B approaches the coding performance of Code-Llama 7B [20],\nwithout sacrificing performance on non-code related benchmarks.\nMistral 7B leverages grouped-query attention (GQA) [1], and sliding window attention (SWA) [6, 3].\nGQA significantly accelerates the inference speed, and also reduces the memory requirement during\ndecoding, allowing for higher batch sizes hence higher throughput, a crucial factor for real-time\napplications. In addition, SWA is designed to handle longer sequences more effectively at a reduced\ncomputational cost, thereby alleviating a common limitation in LLMs. These attention mechanisms\ncollectively contribute to the enhanced performance and efficiency of Mistral 7B.",
"lines": [
{
"text": "Mistral 7B",
"bounds": {
"top_left": {
"x": 685,
"y": 302
},
"top_right": {
"x": 905,
"y": 337
},
"bottom_right": {
"x": 898,
"y": 377
},
"bottom_left": {
"x": 679,
"y": 342
},
"width": 219.5,
"height": 40
},
"words": [
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 686,
"y": 303
},
"top_right": {
"x": 843,
"y": 328
},
"bottom_right": {
"x": 837,
"y": 368
},
"bottom_left": {
"x": 680,
"y": 343
},
"width": 157,
"height": 40
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 854,
"y": 330
},
"top_right": {
"x": 893,
"y": 336
},
"bottom_right": {
"x": 886,
"y": 376
},
"bottom_left": {
"x": 848,
"y": 369
},
"width": 38.5,
"height": 39.5
}
}
]
},
{
"text": "Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford,",
"bounds": {
"top_left": {
"x": 350,
"y": 411
},
"top_right": {
"x": 1189,
"y": 545
},
"bottom_right": {
"x": 1184,
"y": 576
},
"bottom_left": {
"x": 346,
"y": 440
},
"width": 838.5,
"height": 30
},
"words": [
{
"text": "Albert",
"bounds": {
"top_left": {
"x": 350,
"y": 411
},
"top_right": {
"x": 424,
"y": 423
},
"bottom_right": {
"x": 420,
"y": 452
},
"bottom_left": {
"x": 346,
"y": 440
},
"width": 74,
"height": 29
}
},
{
"text": "Q.",
"bounds": {
"top_left": {
"x": 430,
"y": 424
},
"top_right": {
"x": 459,
"y": 429
},
"bottom_right": {
"x": 455,
"y": 458
},
"bottom_left": {
"x": 426,
"y": 453
},
"width": 29,
"height": 29
}
},
{
"text": "Jiang,",
"bounds": {
"top_left": {
"x": 465,
"y": 430
},
"top_right": {
"x": 537,
"y": 442
},
"bottom_right": {
"x": 533,
"y": 471
},
"bottom_left": {
"x": 461,
"y": 459
},
"width": 72,
"height": 29
}
},
{
"text": "Alexandre",
"bounds": {
"top_left": {
"x": 543,
"y": 442
},
"top_right": {
"x": 662,
"y": 462
},
"bottom_right": {
"x": 658,
"y": 491
},
"bottom_left": {
"x": 539,
"y": 472
},
"width": 119,
"height": 29.5
}
},
{
"text": "Sablayrolles,",
"bounds": {
"top_left": {
"x": 668,
"y": 463
},
"top_right": {
"x": 816,
"y": 486
},
"bottom_right": {
"x": 811,
"y": 516
},
"bottom_left": {
"x": 663,
"y": 492
},
"width": 148,
"height": 29.5
}
},
{
"text": "Arthur",
"bounds": {
"top_left": {
"x": 822,
"y": 487
},
"top_right": {
"x": 902,
"y": 500
},
"bottom_right": {
"x": 897,
"y": 530
},
"bottom_left": {
"x": 817,
"y": 517
},
"width": 80,
"height": 30
}
},
{
"text": "Mensch,",
"bounds": {
"top_left": {
"x": 907,
"y": 501
},
"top_right": {
"x": 1007,
"y": 517
},
"bottom_right": {
"x": 1002,
"y": 547
},
"bottom_left": {
"x": 903,
"y": 531
},
"width": 99.5,
"height": 30
}
},
{
"text": "Chris",
"bounds": {
"top_left": {
"x": 1013,
"y": 517
},
"top_right": {
"x": 1077,
"y": 528
},
"bottom_right": {
"x": 1072,
"y": 558
},
"bottom_left": {
"x": 1008,
"y": 548
},
"width": 64,
"height": 30.5
}
},
{
"text": "Bamford,",
"bounds": {
"top_left": {
"x": 1083,
"y": 529
},
"top_right": {
"x": 1190,
"y": 545
},
"bottom_right": {
"x": 1185,
"y": 576
},
"bottom_left": {
"x": 1078,
"y": 559
},
"width": 107,
"height": 30.5
}
}
]
},
{
"text": "Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel,",
"bounds": {
"top_left": {
"x": 304,
"y": 435
},
"top_right": {
"x": 1226,
"y": 583
},
"bottom_right": {
"x": 1221,
"y": 613
},
"bottom_left": {
"x": 300,
"y": 464
},
"width": 921.5,
"height": 29.5
},
"words": [
{
"text": "Devendra",
"bounds": {
"top_left": {
"x": 305,
"y": 436
},
"top_right": {
"x": 417,
"y": 454
},
"bottom_right": {
"x": 412,
"y": 482
},
"bottom_left": {
"x": 300,
"y": 463
},
"width": 112,
"height": 27.5
}
},
{
"text": "Singh",
"bounds": {
"top_left": {
"x": 423,
"y": 455
},
"top_right": {
"x": 488,
"y": 465
},
"bottom_right": {
"x": 483,
"y": 494
},
"bottom_left": {
"x": 418,
"y": 483
},
"width": 65,
"height": 28.5
}
},
{
"text": "Chaplot,",
"bounds": {
"top_left": {
"x": 496,
"y": 466
},
"top_right": {
"x": 597,
"y": 483
},
"bottom_right": {
"x": 592,
"y": 512
},
"bottom_left": {
"x": 491,
"y": 495
},
"width": 101,
"height": 29
}
},
{
"text": "Diego",
"bounds": {
"top_left": {
"x": 602,
"y": 483
},
"top_right": {
"x": 668,
"y": 494
},
"bottom_right": {
"x": 663,
"y": 524
},
"bottom_left": {
"x": 598,
"y": 513
},
"width": 65.5,
"height": 30
}
},
{
"text": "de",
"bounds": {
"top_left": {
"x": 674,
"y": 495
},
"top_right": {
"x": 701,
"y": 499
},
"bottom_right": {
"x": 696,
"y": 529
},
"bottom_left": {
"x": 669,
"y": 525
},
"width": 27,
"height": 30
}
},
{
"text": "las",
"bounds": {
"top_left": {
"x": 706,
"y": 500
},
"top_right": {
"x": 741,
"y": 506
},
"bottom_right": {
"x": 737,
"y": 536
},
"bottom_left": {
"x": 702,
"y": 530
},
"width": 35,
"height": 30
}
},
{
"text": "Casas,",
"bounds": {
"top_left": {
"x": 747,
"y": 507
},
"top_right": {
"x": 820,
"y": 518
},
"bottom_right": {
"x": 816,
"y": 549
},
"bottom_left": {
"x": 743,
"y": 537
},
"width": 73,
"height": 30.5
}
},
{
"text": "Florian",
"bounds": {
"top_left": {
"x": 826,
"y": 519
},
"top_right": {
"x": 911,
"y": 533
},
"bottom_right": {
"x": 906,
"y": 563
},
"bottom_left": {
"x": 822,
"y": 550
},
"width": 84.5,
"height": 30.5
}
},
{
"text": "Bressand,",
"bounds": {
"top_left": {
"x": 917,
"y": 534
},
"top_right": {
"x": 1031,
"y": 552
},
"bottom_right": {
"x": 1027,
"y": 582
},
"bottom_left": {
"x": 913,
"y": 564
},
"width": 114,
"height": 30
}
},
{
"text": "Gianna",
"bounds": {
"top_left": {
"x": 1036,
"y": 553
},
"top_right": {
"x": 1121,
"y": 567
},
"bottom_right": {
"x": 1117,
"y": 597
},
"bottom_left": {
"x": 1032,
"y": 583
},
"width": 85,
"height": 30
}
},
{
"text": "Lengyel,",
"bounds": {
"top_left": {
"x": 1127,
"y": 567
},
"top_right": {
"x": 1226,
"y": 583
},
"bottom_right": {
"x": 1222,
"y": 613
},
"bottom_left": {
"x": 1123,
"y": 598
},
"width": 99,
"height": 30.5
}
}
]
},
{
"text": "Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux,",
"bounds": {
"top_left": {
"x": 295,
"y": 464
},
"top_right": {
"x": 1220,
"y": 613
},
"bottom_right": {
"x": 1215,
"y": 643
},
"bottom_left": {
"x": 291,
"y": 493
},
"width": 924.5,
"height": 29.5
},
"words": [
{
"text": "Guillaume",
"bounds": {
"top_left": {
"x": 297,
"y": 464
},
"top_right": {
"x": 418,
"y": 484
},
"bottom_right": {
"x": 414,
"y": 513
},
"bottom_left": {
"x": 293,
"y": 494
},
"width": 121,
"height": 29.5
}
},
{
"text": "Lample,",
"bounds": {
"top_left": {
"x": 424,
"y": 485
},
"top_right": {
"x": 520,
"y": 500
},
"bottom_right": {
"x": 516,
"y": 530
},
"bottom_left": {
"x": 420,
"y": 514
},
"width": 96,
"height": 29.5
}
},
{
"text": "Lucile",
"bounds": {
"top_left": {
"x": 526,
"y": 501
},
"top_right": {
"x": 599,
"y": 513
},
"bottom_right": {
"x": 595,
"y": 543
},
"bottom_left": {
"x": 522,
"y": 531
},
"width": 73,
"height": 30
}
},
{
"text": "Saulnier,",
"bounds": {
"top_left": {
"x": 605,
"y": 514
},
"top_right": {
"x": 707,
"y": 531
},
"bottom_right": {
"x": 702,
"y": 560
},
"bottom_left": {
"x": 601,
"y": 544
},
"width": 101.5,
"height": 29.5
}
},
{
"text": "Lélio",
"bounds": {
"top_left": {
"x": 712,
"y": 532
},
"top_right": {
"x": 772,
"y": 541
},
"bottom_right": {
"x": 767,
"y": 571
},
"bottom_left": {
"x": 708,
"y": 561
},
"width": 59.5,
"height": 29.5
}
},
{
"text": "Renard",
"bounds": {
"top_left": {
"x": 778,
"y": 542
},
"top_right": {
"x": 864,
"y": 556
},
"bottom_right": {
"x": 859,
"y": 586
},
"bottom_left": {
"x": 773,
"y": 572
},
"width": 86,
"height": 30
}
},
{
"text": "Lavaud,",
"bounds": {
"top_left": {
"x": 870,
"y": 557
},
"top_right": {
"x": 962,
"y": 572
},
"bottom_right": {
"x": 958,
"y": 602
},
"bottom_left": {
"x": 865,
"y": 587
},
"width": 92.5,
"height": 30
}
},
{
"text": "Marie-Anne",
"bounds": {
"top_left": {
"x": 968,
"y": 573
},
"top_right": {
"x": 1110,
"y": 596
},
"bottom_right": {
"x": 1105,
"y": 625
},
"bottom_left": {
"x": 963,
"y": 603
},
"width": 142,
"height": 29.5
}
},
{
"text": "Lachaux,",
"bounds": {
"top_left": {
"x": 1116,
"y": 597
},
"top_right": {
"x": 1220,
"y": 614
},
"bottom_right": {
"x": 1215,
"y": 643
},
"bottom_left": {
"x": 1111,
"y": 626
},
"width": 104,
"height": 29
}
}
]
},
{
"text": "Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,",
"bounds": {
"top_left": {
"x": 302,
"y": 495
},
"top_right": {
"x": 1210,
"y": 641
},
"bottom_right": {
"x": 1205,
"y": 671
},
"bottom_left": {
"x": 297,
"y": 525
},
"width": 908,
"height": 30
},
"words": [
{
"text": "Pierre",
"bounds": {
"top_left": {
"x": 305,
"y": 495
},
"top_right": {
"x": 376,
"y": 507
},
"bottom_right": {
"x": 371,
"y": 537
},
"bottom_left": {
"x": 300,
"y": 524
},
"width": 71,
"height": 29.5
}
},
{
"text": "Stock,",
"bounds": {
"top_left": {
"x": 382,
"y": 508
},
"top_right": {
"x": 455,
"y": 520
},
"bottom_right": {
"x": 450,
"y": 550
},
"bottom_left": {
"x": 377,
"y": 538
},
"width": 73,
"height": 30
}
},
{
"text": "Teven",
"bounds": {
"top_left": {
"x": 461,
"y": 521
},
"top_right": {
"x": 526,
"y": 532
},
"bottom_right": {
"x": 521,
"y": 562
},
"bottom_left": {
"x": 456,
"y": 551
},
"width": 65,
"height": 30
}
},
{
"text": "Le",
"bounds": {
"top_left": {
"x": 532,
"y": 533
},
"top_right": {
"x": 564,
"y": 538
},
"bottom_right": {
"x": 559,
"y": 568
},
"bottom_left": {
"x": 527,
"y": 563
},
"width": 32,
"height": 30
}
},
{
"text": "Scao,",
"bounds": {
"top_left": {
"x": 570,
"y": 539
},
"top_right": {
"x": 634,
"y": 549
},
"bottom_right": {
"x": 628,
"y": 580
},
"bottom_left": {
"x": 565,
"y": 569
},
"width": 63.5,
"height": 30.5
}
},
{
"text": "Thibaut",
"bounds": {
"top_left": {
"x": 639,
"y": 550
},
"top_right": {
"x": 729,
"y": 565
},
"bottom_right": {
"x": 723,
"y": 595
},
"bottom_left": {
"x": 634,
"y": 581
},
"width": 89.5,
"height": 30.5
}
},
{
"text": "Lavril,",
"bounds": {
"top_left": {
"x": 735,
"y": 566
},
"top_right": {
"x": 816,
"y": 579
},
"bottom_right": {
"x": 810,
"y": 609
},
"bottom_left": {
"x": 729,
"y": 596
},
"width": 81,
"height": 30
}
},
{
"text": "Thomas",
"bounds": {
"top_left": {
"x": 822,
"y": 580
},
"top_right": {
"x": 913,
"y": 594
},
"bottom_right": {
"x": 907,
"y": 625
},
"bottom_left": {
"x": 816,
"y": 610
},
"width": 91,
"height": 30.5
}
},
{
"text": "Wang,",
"bounds": {
"top_left": {
"x": 919,
"y": 595
},
"top_right": {
"x": 994,
"y": 607
},
"bottom_right": {
"x": 988,
"y": 638
},
"bottom_left": {
"x": 913,
"y": 626
},
"width": 75,
"height": 31
}
},
{
"text": "Timothée",
"bounds": {
"top_left": {
"x": 1000,
"y": 608
},
"top_right": {
"x": 1107,
"y": 625
},
"bottom_right": {
"x": 1101,
"y": 655
},
"bottom_left": {
"x": 994,
"y": 639
},
"width": 107,
"height": 30.5
}
},
{
"text": "Lacroix,",
"bounds": {
"top_left": {
"x": 1113,
"y": 626
},
"top_right": {
"x": 1210,
"y": 641
},
"bottom_right": {
"x": 1204,
"y": 671
},
"bottom_left": {
"x": 1107,
"y": 656
},
"width": 97,
"height": 30
}
}
]
},
{
"text": "William El Sayed",
"bounds": {
"top_left": {
"x": 649,
"y": 582
},
"top_right": {
"x": 851,
"y": 616
},
"bottom_right": {
"x": 847,
"y": 645
},
"bottom_left": {
"x": 645,
"y": 610
},
"width": 202,
"height": 28.5
},
"words": [
{
"text": "William",
"bounds": {
"top_left": {
"x": 653,
"y": 583
},
"top_right": {
"x": 735,
"y": 596
},
"bottom_right": {
"x": 732,
"y": 626
},
"bottom_left": {
"x": 649,
"y": 612
},
"width": 82.5,
"height": 29.5
}
},
{
"text": "El",
"bounds": {
"top_left": {
"x": 750,
"y": 599
},
"top_right": {
"x": 777,
"y": 604
},
"bottom_right": {
"x": 774,
"y": 632
},
"bottom_left": {
"x": 747,
"y": 628
},
"width": 27,
"height": 28.5
}
},
{
"text": "Sayed",
"bounds": {
"top_left": {
"x": 782,
"y": 605
},
"top_right": {
"x": 849,
"y": 617
},
"bottom_right": {
"x": 846,
"y": 645
},
"bottom_left": {
"x": 779,
"y": 633
},
"width": 67,
"height": 28
}
}
]
},
{
"text": "Mistral Al",
"bounds": {
"top_left": {
"x": 316,
"y": 713
},
"top_right": {
"x": 1129,
"y": 708
},
"bottom_right": {
"x": 1130,
"y": 844
},
"bottom_left": {
"x": 316,
"y": 848
},
"width": 813.5,
"height": 135.5
},
"words": [
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 321,
"y": 718
},
"top_right": {
"x": 916,
"y": 708
},
"bottom_right": {
"x": 910,
"y": 848
},
"bottom_left": {
"x": 316,
"y": 847
},
"width": 594.5,
"height": 134.5
}
},
{
"text": "Al",
"bounds": {
"top_left": {
"x": 943,
"y": 708
},
"top_right": {
"x": 1121,
"y": 708
},
"bottom_right": {
"x": 1115,
"y": 847
},
"bottom_left": {
"x": 937,
"y": 848
},
"width": 178,
"height": 139.5
}
}
]
},
{
"text": "Abstract",
"bounds": {
"top_left": {
"x": 634,
"y": 955
},
"top_right": {
"x": 754,
"y": 976
},
"bottom_right": {
"x": 749,
"y": 1004
},
"bottom_left": {
"x": 631,
"y": 984
},
"width": 119,
"height": 28.5
},
"words": [
{
"text": "Abstract",
"bounds": {
"top_left": {
"x": 637,
"y": 956
},
"top_right": {
"x": 754,
"y": 978
},
"bottom_right": {
"x": 749,
"y": 1004
},
"bottom_left": {
"x": 633,
"y": 985
},
"width": 116.5,
"height": 27.5
}
}
]
},
{
"text": "We introduce Mistral 7B, a 7-billion-parameter language model engineered for",
"bounds": {
"top_left": {
"x": 250,
"y": 965
},
"top_right": {
"x": 1108,
"y": 1103
},
"bottom_right": {
"x": 1105,
"y": 1132
},
"bottom_left": {
"x": 248,
"y": 994
},
"width": 857.5,
"height": 29
},
"words": [
{
"text": "We",
"bounds": {
"top_left": {
"x": 256,
"y": 967
},
"top_right": {
"x": 290,
"y": 972
},
"bottom_right": {
"x": 288,
"y": 1002
},
"bottom_left": {
"x": 253,
"y": 996
},
"width": 34.5,
"height": 29.5
}
},
{
"text": "introduce",
"bounds": {
"top_left": {
"x": 296,
"y": 973
},
"top_right": {
"x": 399,
"y": 991
},
"bottom_right": {
"x": 396,
"y": 1019
},
"bottom_left": {
"x": 293,
"y": 1003
},
"width": 103,
"height": 29
}
},
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 404,
"y": 992
},
"top_right": {
"x": 487,
"y": 1005
},
"bottom_right": {
"x": 485,
"y": 1034
},
"bottom_left": {
"x": 402,
"y": 1020
},
"width": 83,
"height": 28.5
}
},
{
"text": "7B,",
"bounds": {
"top_left": {
"x": 493,
"y": 1006
},
"top_right": {
"x": 529,
"y": 1012
},
"bottom_right": {
"x": 527,
"y": 1040
},
"bottom_left": {
"x": 490,
"y": 1035
},
"width": 36.5,
"height": 28.5
}
},
{
"text": "a",
"bounds": {
"top_left": {
"x": 534,
"y": 1013
},
"top_right": {
"x": 550,
"y": 1016
},
"bottom_right": {
"x": 548,
"y": 1044
},
"bottom_left": {
"x": 532,
"y": 1041
},
"width": 16,
"height": 28
}
},
{
"text": "7-billion-parameter",
"bounds": {
"top_left": {
"x": 556,
"y": 1017
},
"top_right": {
"x": 764,
"y": 1051
},
"bottom_right": {
"x": 762,
"y": 1078
},
"bottom_left": {
"x": 554,
"y": 1045
},
"width": 208,
"height": 27.5
}
},
{
"text": "language",
"bounds": {
"top_left": {
"x": 770,
"y": 1052
},
"top_right": {
"x": 868,
"y": 1067
},
"bottom_right": {
"x": 865,
"y": 1094
},
"bottom_left": {
"x": 767,
"y": 1079
},
"width": 98,
"height": 27
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 873,
"y": 1068
},
"top_right": {
"x": 942,
"y": 1079
},
"bottom_right": {
"x": 939,
"y": 1106
},
"bottom_left": {
"x": 871,
"y": 1095
},
"width": 68.5,
"height": 27
}
},
{
"text": "engineered",
"bounds": {
"top_left": {
"x": 947,
"y": 1080
},
"top_right": {
"x": 1067,
"y": 1098
},
"bottom_right": {
"x": 1064,
"y": 1125
},
"bottom_left": {
"x": 945,
"y": 1107
},
"width": 119.5,
"height": 27
}
},
{
"text": "for",
"bounds": {
"top_left": {
"x": 1072,
"y": 1099
},
"top_right": {
"x": 1107,
"y": 1105
},
"bottom_right": {
"x": 1104,
"y": 1132
},
"bottom_left": {
"x": 1070,
"y": 1126
},
"width": 34.5,
"height": 27
}
}
]
},
{
"text": "superior performance and efficiency. Mistral 7B outperforms the best open 13B",
"bounds": {
"top_left": {
"x": 252,
"y": 999
},
"top_right": {
"x": 1099,
"y": 1132
},
"bottom_right": {
"x": 1095,
"y": 1160
},
"bottom_left": {
"x": 248,
"y": 1026
},
"width": 847,
"height": 27.5
},
"words": [
{
"text": "superior",
"bounds": {
"top_left": {
"x": 253,
"y": 1000
},
"top_right": {
"x": 340,
"y": 1013
},
"bottom_right": {
"x": 336,
"y": 1041
},
"bottom_left": {
"x": 249,
"y": 1027
},
"width": 87,
"height": 27.5
}
},
{
"text": "performance",
"bounds": {
"top_left": {
"x": 345,
"y": 1014
},
"top_right": {
"x": 484,
"y": 1035
},
"bottom_right": {
"x": 480,
"y": 1063
},
"bottom_left": {
"x": 342,
"y": 1041
},
"width": 138.5,
"height": 27.5
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 490,
"y": 1036
},
"top_right": {
"x": 527,
"y": 1042
},
"bottom_right": {
"x": 523,
"y": 1069
},
"bottom_left": {
"x": 486,
"y": 1064
},
"width": 37,
"height": 27.5
}
},
{
"text": "efficiency.",
"bounds": {
"top_left": {
"x": 532,
"y": 1043
},
"top_right": {
"x": 645,
"y": 1060
},
"bottom_right": {
"x": 641,
"y": 1088
},
"bottom_left": {
"x": 529,
"y": 1070
},
"width": 112.5,
"height": 27.5
}
},
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 650,
"y": 1061
},
"top_right": {
"x": 730,
"y": 1074
},
"bottom_right": {
"x": 726,
"y": 1101
},
"bottom_left": {
"x": 646,
"y": 1089
},
"width": 80,
"height": 27.5
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 736,
"y": 1075
},
"top_right": {
"x": 762,
"y": 1079
},
"bottom_right": {
"x": 758,
"y": 1106
},
"bottom_left": {
"x": 732,
"y": 1102
},
"width": 26,
"height": 27
}
},
{
"text": "outperforms",
"bounds": {
"top_left": {
"x": 771,
"y": 1080
},
"top_right": {
"x": 901,
"y": 1101
},
"bottom_right": {
"x": 898,
"y": 1128
},
"bottom_left": {
"x": 767,
"y": 1108
},
"width": 130.5,
"height": 27.5
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 907,
"y": 1102
},
"top_right": {
"x": 940,
"y": 1108
},
"bottom_right": {
"x": 937,
"y": 1134
},
"bottom_left": {
"x": 903,
"y": 1129
},
"width": 33.5,
"height": 26.5
}
},
{
"text": "best",
"bounds": {
"top_left": {
"x": 946,
"y": 1108
},
"top_right": {
"x": 990,
"y": 1116
},
"bottom_right": {
"x": 987,
"y": 1142
},
"bottom_left": {
"x": 942,
"y": 1135
},
"width": 44.5,
"height": 26.5
}
},
{
"text": "open",
"bounds": {
"top_left": {
"x": 996,
"y": 1116
},
"top_right": {
"x": 1049,
"y": 1125
},
"bottom_right": {
"x": 1045,
"y": 1152
},
"bottom_left": {
"x": 992,
"y": 1143
},
"width": 53,
"height": 27
}
},
{
"text": "13B",
"bounds": {
"top_left": {
"x": 1058,
"y": 1127
},
"top_right": {
"x": 1097,
"y": 1133
},
"bottom_right": {
"x": 1093,
"y": 1160
},
"bottom_left": {
"x": 1054,
"y": 1153
},
"width": 39,
"height": 26.5
}
}
]
},
{
"text": "model (Llama 2) across all evaluated benchmarks, and the best released 34B",
"bounds": {
"top_left": {
"x": 248,
"y": 1026
},
"top_right": {
"x": 1094,
"y": 1157
},
"bottom_right": {
"x": 1091,
"y": 1186
},
"bottom_left": {
"x": 244,
"y": 1054
},
"width": 846.5,
"height": 28.5
},
"words": [
{
"text": "model",
"bounds": {
"top_left": {
"x": 248,
"y": 1026
},
"top_right": {
"x": 319,
"y": 1037
},
"bottom_right": {
"x": 315,
"y": 1066
},
"bottom_left": {
"x": 245,
"y": 1055
},
"width": 70.5,
"height": 29
}
},
{
"text": "(Llama",
"bounds": {
"top_left": {
"x": 325,
"y": 1038
},
"top_right": {
"x": 404,
"y": 1051
},
"bottom_right": {
"x": 400,
"y": 1079
},
"bottom_left": {
"x": 321,
"y": 1067
},
"width": 79,
"height": 28.5
}
},
{
"text": "2)",
"bounds": {
"top_left": {
"x": 413,
"y": 1052
},
"top_right": {
"x": 439,
"y": 1056
},
"bottom_right": {
"x": 435,
"y": 1085
},
"bottom_left": {
"x": 410,
"y": 1081
},
"width": 25.5,
"height": 29
}
},
{
"text": "across",
"bounds": {
"top_left": {
"x": 444,
"y": 1057
},
"top_right": {
"x": 514,
"y": 1068
},
"bottom_right": {
"x": 511,
"y": 1096
},
"bottom_left": {
"x": 441,
"y": 1085
},
"width": 70,
"height": 28
}
},
{
"text": "all",
"bounds": {
"top_left": {
"x": 520,
"y": 1069
},
"top_right": {
"x": 549,
"y": 1074
},
"bottom_right": {
"x": 546,
"y": 1102
},
"bottom_left": {
"x": 517,
"y": 1097
},
"width": 29,
"height": 28
}
},
{
"text": "evaluated",
"bounds": {
"top_left": {
"x": 555,
"y": 1074
},
"top_right": {
"x": 657,
"y": 1091
},
"bottom_right": {
"x": 654,
"y": 1118
},
"bottom_left": {
"x": 551,
"y": 1103
},
"width": 102.5,
"height": 28
}
},
{
"text": "benchmarks,",
"bounds": {
"top_left": {
"x": 665,
"y": 1092
},
"top_right": {
"x": 804,
"y": 1114
},
"bottom_right": {
"x": 801,
"y": 1141
},
"bottom_left": {
"x": 662,
"y": 1120
},
"width": 139,
"height": 27.5
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 810,
"y": 1115
},
"top_right": {
"x": 850,
"y": 1121
},
"bottom_right": {
"x": 847,
"y": 1148
},
"bottom_left": {
"x": 807,
"y": 1142
},
"width": 40,
"height": 27
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 857,
"y": 1122
},
"top_right": {
"x": 893,
"y": 1128
},
"bottom_right": {
"x": 890,
"y": 1155
},
"bottom_left": {
"x": 854,
"y": 1149
},
"width": 36,
"height": 27
}
},
{
"text": "best",
"bounds": {
"top_left": {
"x": 899,
"y": 1128
},
"top_right": {
"x": 946,
"y": 1136
},
"bottom_right": {
"x": 943,
"y": 1163
},
"bottom_left": {
"x": 896,
"y": 1156
},
"width": 47,
"height": 27.5
}
},
{
"text": "released",
"bounds": {
"top_left": {
"x": 951,
"y": 1137
},
"top_right": {
"x": 1040,
"y": 1150
},
"bottom_right": {
"x": 1037,
"y": 1177
},
"bottom_left": {
"x": 949,
"y": 1164
},
"width": 88.5,
"height": 27
}
},
{
"text": "34B",
"bounds": {
"top_left": {
"x": 1049,
"y": 1152
},
"top_right": {
"x": 1091,
"y": 1158
},
"bottom_right": {
"x": 1088,
"y": 1185
},
"bottom_left": {
"x": 1047,
"y": 1179
},
"width": 41.5,
"height": 27
}
}
]
},
{
"text": "model (Llama 1) in reasoning, mathematics, and code generation. Our model",
"bounds": {
"top_left": {
"x": 241,
"y": 1052
},
"top_right": {
"x": 1091,
"y": 1186
},
"bottom_right": {
"x": 1086,
"y": 1215
},
"bottom_left": {
"x": 236,
"y": 1081
},
"width": 850,
"height": 29
},
"words": [
{
"text": "model",
"bounds": {
"top_left": {
"x": 245,
"y": 1053
},
"top_right": {
"x": 315,
"y": 1065
},
"bottom_right": {
"x": 309,
"y": 1093
},
"bottom_left": {
"x": 238,
"y": 1082
},
"width": 70.5,
"height": 28.5
}
},
{
"text": "(Llama",
"bounds": {
"top_left": {
"x": 321,
"y": 1066
},
"top_right": {
"x": 400,
"y": 1079
},
"bottom_right": {
"x": 393,
"y": 1107
},
"bottom_left": {
"x": 314,
"y": 1094
},
"width": 79,
"height": 28
}
},
{
"text": "1)",
"bounds": {
"top_left": {
"x": 409,
"y": 1080
},
"top_right": {
"x": 431,
"y": 1084
},
"bottom_right": {
"x": 425,
"y": 1111
},
"bottom_left": {
"x": 403,
"y": 1108
},
"width": 22,
"height": 27.5
}
},
{
"text": "in",
"bounds": {
"top_left": {
"x": 436,
"y": 1085
},
"top_right": {
"x": 459,
"y": 1088
},
"bottom_right": {
"x": 453,
"y": 1116
},
"bottom_left": {
"x": 430,
"y": 1112
},
"width": 23,
"height": 27.5
}
},
{
"text": "reasoning,",
"bounds": {
"top_left": {
"x": 467,
"y": 1090
},
"top_right": {
"x": 581,
"y": 1108
},
"bottom_right": {
"x": 575,
"y": 1135
},
"bottom_left": {
"x": 461,
"y": 1117
},
"width": 114,
"height": 27
}
},
{
"text": "mathematics,",
"bounds": {
"top_left": {
"x": 587,
"y": 1109
},
"top_right": {
"x": 730,
"y": 1132
},
"bottom_right": {
"x": 723,
"y": 1158
},
"bottom_left": {
"x": 580,
"y": 1136
},
"width": 143,
"height": 26.5
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 735,
"y": 1133
},
"top_right": {
"x": 773,
"y": 1139
},
"bottom_right": {
"x": 766,
"y": 1165
},
"bottom_left": {
"x": 729,
"y": 1159
},
"width": 37.5,
"height": 26
}
},
{
"text": "code",
"bounds": {
"top_left": {
"x": 780,
"y": 1140
},
"top_right": {
"x": 834,
"y": 1148
},
"bottom_right": {
"x": 828,
"y": 1175
},
"bottom_left": {
"x": 774,
"y": 1166
},
"width": 54,
"height": 26.5
}
},
{
"text": "generation.",
"bounds": {
"top_left": {
"x": 842,
"y": 1149
},
"top_right": {
"x": 965,
"y": 1168
},
"bottom_right": {
"x": 958,
"y": 1195
},
"bottom_left": {
"x": 836,
"y": 1176
},
"width": 122.5,
"height": 27
}
},
{
"text": "Our",
"bounds": {
"top_left": {
"x": 974,
"y": 1170
},
"top_right": {
"x": 1018,
"y": 1177
},
"bottom_right": {
"x": 1011,
"y": 1203
},
"bottom_left": {
"x": 968,
"y": 1197
},
"width": 43.5,
"height": 26.5
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 1023,
"y": 1177
},
"top_right": {
"x": 1090,
"y": 1188
},
"bottom_right": {
"x": 1084,
"y": 1215
},
"bottom_left": {
"x": 1017,
"y": 1204
},
"width": 67,
"height": 27
}
}
]
},
{
"text": "leverages grouped-query attention (GQA) for faster inference, coupled with sliding",
"bounds": {
"top_left": {
"x": 240,
"y": 1084
},
"top_right": {
"x": 1086,
"y": 1215
},
"bottom_right": {
"x": 1082,
"y": 1243
},
"bottom_left": {
"x": 235,
"y": 1111
},
"width": 846.5,
"height": 27.5
},
"words": [
{
"text": "leverages",
"bounds": {
"top_left": {
"x": 241,
"y": 1084
},
"top_right": {
"x": 338,
"y": 1099
},
"bottom_right": {
"x": 333,
"y": 1127
},
"bottom_left": {
"x": 236,
"y": 1112
},
"width": 97,
"height": 28
}
},
{
"text": "grouped-query",
"bounds": {
"top_left": {
"x": 343,
"y": 1100
},
"top_right": {
"x": 494,
"y": 1123
},
"bottom_right": {
"x": 489,
"y": 1151
},
"bottom_left": {
"x": 338,
"y": 1128
},
"width": 151,
"height": 28
}
},
{
"text": "attention",
"bounds": {
"top_left": {
"x": 500,
"y": 1124
},
"top_right": {
"x": 588,
"y": 1138
},
"bottom_right": {
"x": 583,
"y": 1165
},
"bottom_left": {
"x": 495,
"y": 1152
},
"width": 88,
"height": 27.5
}
},
{
"text": "(GQA)",
"bounds": {
"top_left": {
"x": 594,
"y": 1139
},
"top_right": {
"x": 668,
"y": 1150
},
"bottom_right": {
"x": 662,
"y": 1177
},
"bottom_left": {
"x": 588,
"y": 1166
},
"width": 74,
"height": 27
}
},
{
"text": "for",
"bounds": {
"top_left": {
"x": 673,
"y": 1151
},
"top_right": {
"x": 702,
"y": 1156
},
"bottom_right": {
"x": 696,
"y": 1183
},
"bottom_left": {
"x": 667,
"y": 1178
},
"width": 29,
"height": 27
}
},
{
"text": "faster",
"bounds": {
"top_left": {
"x": 707,
"y": 1156
},
"top_right": {
"x": 765,
"y": 1166
},
"bottom_right": {
"x": 759,
"y": 1192
},
"bottom_left": {
"x": 702,
"y": 1183
},
"width": 57.5,
"height": 26.5
}
},
{
"text": "inference,",
"bounds": {
"top_left": {
"x": 770,
"y": 1166
},
"top_right": {
"x": 871,
"y": 1182
},
"bottom_right": {
"x": 865,
"y": 1209
},
"bottom_left": {
"x": 765,
"y": 1193
},
"width": 100.5,
"height": 27
}
},
{
"text": "coupled",
"bounds": {
"top_left": {
"x": 877,
"y": 1183
},
"top_right": {
"x": 959,
"y": 1196
},
"bottom_right": {
"x": 953,
"y": 1223
},
"bottom_left": {
"x": 871,
"y": 1210
},
"width": 82,
"height": 27
}
},
{
"text": "with",
"bounds": {
"top_left": {
"x": 965,
"y": 1197
},
"top_right": {
"x": 1009,
"y": 1204
},
"bottom_right": {
"x": 1003,
"y": 1231
},
"bottom_left": {
"x": 959,
"y": 1224
},
"width": 44,
"height": 27
}
},
{
"text": "sliding",
"bounds": {
"top_left": {
"x": 1017,
"y": 1206
},
"top_right": {
"x": 1085,
"y": 1217
},
"bottom_right": {
"x": 1079,
"y": 1243
},
"bottom_left": {
"x": 1011,
"y": 1232
},
"width": 68,
"height": 26
}
}
]
},
{
"text": "window attention (SWA) to effectively handle sequences of arbitrary length with a",
"bounds": {
"top_left": {
"x": 236,
"y": 1109
},
"top_right": {
"x": 1082,
"y": 1243
},
"bottom_right": {
"x": 1078,
"y": 1271
},
"bottom_left": {
"x": 232,
"y": 1137
},
"width": 846,
"height": 28
},
"words": [
{
"text": "window",
"bounds": {
"top_left": {
"x": 237,
"y": 1110
},
"top_right": {
"x": 314,
"y": 1122
},
"bottom_right": {
"x": 310,
"y": 1150
},
"bottom_left": {
"x": 233,
"y": 1138
},
"width": 77,
"height": 28
}
},
{
"text": "attention",
"bounds": {
"top_left": {
"x": 324,
"y": 1123
},
"top_right": {
"x": 415,
"y": 1138
},
"bottom_right": {
"x": 410,
"y": 1166
},
"bottom_left": {
"x": 320,
"y": 1151
},
"width": 90.5,
"height": 28
}
},
{
"text": "(SWA)",
"bounds": {
"top_left": {
"x": 420,
"y": 1138
},
"top_right": {
"x": 493,
"y": 1150
},
"bottom_right": {
"x": 488,
"y": 1178
},
"bottom_left": {
"x": 416,
"y": 1167
},
"width": 72.5,
"height": 28.5
}
},
{
"text": "to",
"bounds": {
"top_left": {
"x": 498,
"y": 1151
},
"top_right": {
"x": 520,
"y": 1154
},
"bottom_right": {
"x": 515,
"y": 1182
},
"bottom_left": {
"x": 494,
"y": 1179
},
"width": 21.5,
"height": 28
}
},
{
"text": "effectively",
"bounds": {
"top_left": {
"x": 526,
"y": 1155
},
"top_right": {
"x": 633,
"y": 1172
},
"bottom_right": {
"x": 628,
"y": 1200
},
"bottom_left": {
"x": 521,
"y": 1183
},
"width": 107,
"height": 28
}
},
{
"text": "handle",
"bounds": {
"top_left": {
"x": 638,
"y": 1173
},
"top_right": {
"x": 709,
"y": 1184
},
"bottom_right": {
"x": 704,
"y": 1212
},
"bottom_left": {
"x": 633,
"y": 1201
},
"width": 71,
"height": 28
}
},
{
"text": "sequences",
"bounds": {
"top_left": {
"x": 715,
"y": 1185
},
"top_right": {
"x": 818,
"y": 1202
},
"bottom_right": {
"x": 813,
"y": 1229
},
"bottom_left": {
"x": 709,
"y": 1213
},
"width": 103.5,
"height": 27.5
}
},
{
"text": "of",
"bounds": {
"top_left": {
"x": 824,
"y": 1203
},
"top_right": {
"x": 845,
"y": 1206
},
"bottom_right": {
"x": 840,
"y": 1234
},
"bottom_left": {
"x": 818,
"y": 1230
},
"width": 21.5,
"height": 27.5
}
},
{
"text": "arbitrary",
"bounds": {
"top_left": {
"x": 851,
"y": 1207
},
"top_right": {
"x": 940,
"y": 1221
},
"bottom_right": {
"x": 934,
"y": 1249
},
"bottom_left": {
"x": 845,
"y": 1234
},
"width": 89,
"height": 27.5
}
},
{
"text": "length",
"bounds": {
"top_left": {
"x": 945,
"y": 1222
},
"top_right": {
"x": 1010,
"y": 1233
},
"bottom_right": {
"x": 1004,
"y": 1260
},
"bottom_left": {
"x": 939,
"y": 1249
},
"width": 65,
"height": 27
}
},
{
"text": "with",
"bounds": {
"top_left": {
"x": 1016,
"y": 1234
},
"top_right": {
"x": 1063,
"y": 1242
},
"bottom_right": {
"x": 1057,
"y": 1268
},
"bottom_left": {
"x": 1010,
"y": 1261
},
"width": 47,
"height": 26.5
}
},
{
"text": "a",
"bounds": {
"top_left": {
"x": 1069,
"y": 1243
},
"top_right": {
"x": 1081,
"y": 1245
},
"bottom_right": {
"x": 1075,
"y": 1271
},
"bottom_left": {
"x": 1063,
"y": 1269
},
"width": 12,
"height": 26
}
}
]
},
{
"text": "reduced inference cost. We also provide a model fine-tuned to follow instructions,",
"bounds": {
"top_left": {
"x": 233,
"y": 1139
},
"top_right": {
"x": 1078,
"y": 1271
},
"bottom_right": {
"x": 1074,
"y": 1299
},
"bottom_left": {
"x": 228,
"y": 1165
},
"width": 845.5,
"height": 27
},
"words": [
{
"text": "reduced",
"bounds": {
"top_left": {
"x": 234,
"y": 1139
},
"top_right": {
"x": 315,
"y": 1151
},
"bottom_right": {
"x": 310,
"y": 1179
},
"bottom_left": {
"x": 229,
"y": 1166
},
"width": 81,
"height": 27.5
}
},
{
"text": "inference",
"bounds": {
"top_left": {
"x": 321,
"y": 1152
},
"top_right": {
"x": 418,
"y": 1167
},
"bottom_right": {
"x": 413,
"y": 1195
},
"bottom_left": {
"x": 315,
"y": 1179
},
"width": 97.5,
"height": 27.5
}
},
{
"text": "cost.",
"bounds": {
"top_left": {
"x": 423,
"y": 1168
},
"top_right": {
"x": 476,
"y": 1176
},
"bottom_right": {
"x": 471,
"y": 1204
},
"bottom_left": {
"x": 418,
"y": 1195
},
"width": 53,
"height": 27.5
}
},
{
"text": "We",
"bounds": {
"top_left": {
"x": 482,
"y": 1177
},
"top_right": {
"x": 517,
"y": 1182
},
"bottom_right": {
"x": 511,
"y": 1210
},
"bottom_left": {
"x": 476,
"y": 1205
},
"width": 35,
"height": 28
}
},
{
"text": "also",
"bounds": {
"top_left": {
"x": 522,
"y": 1183
},
"top_right": {
"x": 563,
"y": 1190
},
"bottom_right": {
"x": 558,
"y": 1217
},
"bottom_left": {
"x": 517,
"y": 1211
},
"width": 41,
"height": 27.5
}
},
{
"text": "provide",
"bounds": {
"top_left": {
"x": 568,
"y": 1191
},
"top_right": {
"x": 648,
"y": 1203
},
"bottom_right": {
"x": 643,
"y": 1231
},
"bottom_left": {
"x": 563,
"y": 1218
},
"width": 80,
"height": 27.5
}
},
{
"text": "a",
"bounds": {
"top_left": {
"x": 653,
"y": 1204
},
"top_right": {
"x": 666,
"y": 1206
},
"bottom_right": {
"x": 660,
"y": 1233
},
"bottom_left": {
"x": 648,
"y": 1231
},
"width": 12.5,
"height": 27
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 671,
"y": 1207
},
"top_right": {
"x": 737,
"y": 1217
},
"bottom_right": {
"x": 731,
"y": 1245
},
"bottom_left": {
"x": 666,
"y": 1234
},
"width": 65.5,
"height": 27.5
}
},
{
"text": "fine-tuned",
"bounds": {
"top_left": {
"x": 742,
"y": 1218
},
"top_right": {
"x": 846,
"y": 1235
},
"bottom_right": {
"x": 840,
"y": 1262
},
"bottom_left": {
"x": 736,
"y": 1245
},
"width": 104,
"height": 27
}
},
{
"text": "to",
"bounds": {
"top_left": {
"x": 852,
"y": 1236
},
"top_right": {
"x": 873,
"y": 1240
},
"bottom_right": {
"x": 867,
"y": 1266
},
"bottom_left": {
"x": 846,
"y": 1263
},
"width": 21,
"height": 26.5
}
},
{
"text": "follow",
"bounds": {
"top_left": {
"x": 878,
"y": 1240
},
"top_right": {
"x": 941,
"y": 1251
},
"bottom_right": {
"x": 935,
"y": 1277
},
"bottom_left": {
"x": 872,
"y": 1267
},
"width": 63,
"height": 26.5
}
},
{
"text": "instructions,",
"bounds": {
"top_left": {
"x": 951,
"y": 1252
},
"top_right": {
"x": 1078,
"y": 1274
},
"bottom_right": {
"x": 1072,
"y": 1299
},
"bottom_left": {
"x": 945,
"y": 1278
},
"width": 127,
"height": 25.5
}
}
]
},
{
"text": "Mistral 7B - Instruct, that surpasses Llama 2 13B - chat model both on human and",
"bounds": {
"top_left": {
"x": 226,
"y": 1164
},
"top_right": {
"x": 1073,
"y": 1299
},
"bottom_right": {
"x": 1068,
"y": 1327
},
"bottom_left": {
"x": 222,
"y": 1192
},
"width": 846.5,
"height": 28
},
"words": [
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 228,
"y": 1165
},
"top_right": {
"x": 305,
"y": 1177
},
"bottom_right": {
"x": 301,
"y": 1206
},
"bottom_left": {
"x": 224,
"y": 1194
},
"width": 77,
"height": 29
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 310,
"y": 1178
},
"top_right": {
"x": 335,
"y": 1182
},
"bottom_right": {
"x": 331,
"y": 1210
},
"bottom_left": {
"x": 306,
"y": 1206
},
"width": 25,
"height": 28
}
},
{
"text": "-",
"bounds": {
"top_left": {
"x": 341,
"y": 1183
},
"top_right": {
"x": 357,
"y": 1186
},
"bottom_right": {
"x": 353,
"y": 1214
},
"bottom_left": {
"x": 337,
"y": 1211
},
"width": 16,
"height": 28
}
},
{
"text": "Instruct,",
"bounds": {
"top_left": {
"x": 363,
"y": 1186
},
"top_right": {
"x": 449,
"y": 1200
},
"bottom_right": {
"x": 445,
"y": 1228
},
"bottom_left": {
"x": 359,
"y": 1215
},
"width": 86,
"height": 28.5
}
},
{
"text": "that",
"bounds": {
"top_left": {
"x": 454,
"y": 1201
},
"top_right": {
"x": 492,
"y": 1207
},
"bottom_right": {
"x": 489,
"y": 1235
},
"bottom_left": {
"x": 450,
"y": 1229
},
"width": 38.5,
"height": 28
}
},
{
"text": "surpasses",
"bounds": {
"top_left": {
"x": 498,
"y": 1208
},
"top_right": {
"x": 595,
"y": 1223
},
"bottom_right": {
"x": 591,
"y": 1251
},
"bottom_left": {
"x": 494,
"y": 1236
},
"width": 97,
"height": 28
}
},
{
"text": "Llama",
"bounds": {
"top_left": {
"x": 600,
"y": 1224
},
"top_right": {
"x": 666,
"y": 1235
},
"bottom_right": {
"x": 662,
"y": 1263
},
"bottom_left": {
"x": 596,
"y": 1252
},
"width": 66,
"height": 28
}
},
{
"text": "2",
"bounds": {
"top_left": {
"x": 671,
"y": 1236
},
"top_right": {
"x": 685,
"y": 1238
},
"bottom_right": {
"x": 681,
"y": 1266
},
"bottom_left": {
"x": 667,
"y": 1263
},
"width": 14,
"height": 27.5
}
},
{
"text": "13B",
"bounds": {
"top_left": {
"x": 691,
"y": 1239
},
"top_right": {
"x": 729,
"y": 1245
},
"bottom_right": {
"x": 725,
"y": 1273
},
"bottom_left": {
"x": 687,
"y": 1267
},
"width": 38,
"height": 28
}
},
{
"text": "-",
"bounds": {
"top_left": {
"x": 735,
"y": 1246
},
"top_right": {
"x": 751,
"y": 1248
},
"bottom_right": {
"x": 747,
"y": 1276
},
"bottom_left": {
"x": 731,
"y": 1274
},
"width": 16,
"height": 28
}
},
{
"text": "chat",
"bounds": {
"top_left": {
"x": 757,
"y": 1249
},
"top_right": {
"x": 799,
"y": 1256
},
"bottom_right": {
"x": 795,
"y": 1284
},
"bottom_left": {
"x": 753,
"y": 1277
},
"width": 42,
"height": 28
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 804,
"y": 1257
},
"top_right": {
"x": 870,
"y": 1267
},
"bottom_right": {
"x": 866,
"y": 1295
},
"bottom_left": {
"x": 800,
"y": 1285
},
"width": 66,
"height": 28
}
},
{
"text": "both",
"bounds": {
"top_left": {
"x": 875,
"y": 1268
},
"top_right": {
"x": 920,
"y": 1276
},
"bottom_right": {
"x": 917,
"y": 1303
},
"bottom_left": {
"x": 871,
"y": 1296
},
"width": 45.5,
"height": 27.5
}
},
{
"text": "on",
"bounds": {
"top_left": {
"x": 926,
"y": 1277
},
"top_right": {
"x": 952,
"y": 1281
},
"bottom_right": {
"x": 948,
"y": 1308
},
"bottom_left": {
"x": 922,
"y": 1304
},
"width": 26,
"height": 27
}
},
{
"text": "human",
"bounds": {
"top_left": {
"x": 957,
"y": 1282
},
"top_right": {
"x": 1028,
"y": 1293
},
"bottom_right": {
"x": 1024,
"y": 1320
},
"bottom_left": {
"x": 953,
"y": 1309
},
"width": 71,
"height": 27
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 1034,
"y": 1294
},
"top_right": {
"x": 1072,
"y": 1300
},
"bottom_right": {
"x": 1068,
"y": 1328
},
"bottom_left": {
"x": 1030,
"y": 1321
},
"width": 38,
"height": 27.5
}
}
]
},
{
"text": "automated benchmarks. Our models are released under the Apache 2.0 license.",
"bounds": {
"top_left": {
"x": 223,
"y": 1195
},
"top_right": {
"x": 1043,
"y": 1323
},
"bottom_right": {
"x": 1038,
"y": 1351
},
"bottom_left": {
"x": 219,
"y": 1220
},
"width": 819.5,
"height": 26.5
},
"words": [
{
"text": "automated",
"bounds": {
"top_left": {
"x": 225,
"y": 1196
},
"top_right": {
"x": 332,
"y": 1212
},
"bottom_right": {
"x": 328,
"y": 1239
},
"bottom_left": {
"x": 221,
"y": 1222
},
"width": 107,
"height": 26.5
}
},
{
"text": "benchmarks.",
"bounds": {
"top_left": {
"x": 338,
"y": 1212
},
"top_right": {
"x": 474,
"y": 1233
},
"bottom_right": {
"x": 469,
"y": 1261
},
"bottom_left": {
"x": 333,
"y": 1239
},
"width": 136,
"height": 27.5
}
},
{
"text": "Our",
"bounds": {
"top_left": {
"x": 479,
"y": 1234
},
"top_right": {
"x": 520,
"y": 1240
},
"bottom_right": {
"x": 515,
"y": 1268
},
"bottom_left": {
"x": 474,
"y": 1262
},
"width": 41,
"height": 28
}
},
{
"text": "models",
"bounds": {
"top_left": {
"x": 526,
"y": 1241
},
"top_right": {
"x": 603,
"y": 1253
},
"bottom_right": {
"x": 597,
"y": 1281
},
"bottom_left": {
"x": 520,
"y": 1269
},
"width": 77,
"height": 28
}
},
{
"text": "are",
"bounds": {
"top_left": {
"x": 608,
"y": 1254
},
"top_right": {
"x": 638,
"y": 1259
},
"bottom_right": {
"x": 633,
"y": 1287
},
"bottom_left": {
"x": 602,
"y": 1282
},
"width": 30.5,
"height": 28
}
},
{
"text": "released",
"bounds": {
"top_left": {
"x": 644,
"y": 1260
},
"top_right": {
"x": 729,
"y": 1273
},
"bottom_right": {
"x": 723,
"y": 1301
},
"bottom_left": {
"x": 638,
"y": 1288
},
"width": 85,
"height": 28
}
},
{
"text": "under",
"bounds": {
"top_left": {
"x": 735,
"y": 1274
},
"top_right": {
"x": 794,
"y": 1284
},
"bottom_right": {
"x": 788,
"y": 1311
},
"bottom_left": {
"x": 729,
"y": 1302
},
"width": 59,
"height": 27.5
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 799,
"y": 1285
},
"top_right": {
"x": 834,
"y": 1291
},
"bottom_right": {
"x": 828,
"y": 1318
},
"bottom_left": {
"x": 793,
"y": 1312
},
"width": 35,
"height": 27
}
},
{
"text": "Apache",
"bounds": {
"top_left": {
"x": 840,
"y": 1292
},
"top_right": {
"x": 920,
"y": 1305
},
"bottom_right": {
"x": 913,
"y": 1331
},
"bottom_left": {
"x": 834,
"y": 1319
},
"width": 79.5,
"height": 26.5
}
},
{
"text": "2.0",
"bounds": {
"top_left": {
"x": 926,
"y": 1306
},
"top_right": {
"x": 956,
"y": 1311
},
"bottom_right": {
"x": 949,
"y": 1337
},
"bottom_left": {
"x": 919,
"y": 1332
},
"width": 30,
"height": 26
}
},
{
"text": "license.",
"bounds": {
"top_left": {
"x": 962,
"y": 1312
},
"top_right": {
"x": 1042,
"y": 1326
},
"bottom_right": {
"x": 1035,
"y": 1351
},
"bottom_left": {
"x": 955,
"y": 1338
},
"width": 80,
"height": 25.5
}
}
]
},
{
"text": "Code: https://github.com/mistralai/mistral-src",
"bounds": {
"top_left": {
"x": 224,
"y": 1223
},
"top_right": {
"x": 843,
"y": 1319
},
"bottom_right": {
"x": 839,
"y": 1347
},
"bottom_left": {
"x": 220,
"y": 1249
},
"width": 619,
"height": 27
},
"words": [
{
"text": "Code:",
"bounds": {
"top_left": {
"x": 224,
"y": 1223
},
"top_right": {
"x": 288,
"y": 1233
},
"bottom_right": {
"x": 285,
"y": 1260
},
"bottom_left": {
"x": 220,
"y": 1250
},
"width": 64.5,
"height": 27
}
},
{
"text": "https://github.com/mistralai/mistral-src",
"bounds": {
"top_left": {
"x": 293,
"y": 1233
},
"top_right": {
"x": 839,
"y": 1322
},
"bottom_right": {
"x": 836,
"y": 1347
},
"bottom_left": {
"x": 290,
"y": 1261
},
"width": 546,
"height": 26.5
}
}
]
},
{
"text": "Webpage: https://mistral.ai/news/announcing-mistral-7b/",
"bounds": {
"top_left": {
"x": 215,
"y": 1250
},
"top_right": {
"x": 963,
"y": 1367
},
"bottom_right": {
"x": 959,
"y": 1395
},
"bottom_left": {
"x": 211,
"y": 1278
},
"width": 748,
"height": 28
},
"words": [
{
"text": "Webpage:",
"bounds": {
"top_left": {
"x": 217,
"y": 1251
},
"top_right": {
"x": 327,
"y": 1268
},
"bottom_right": {
"x": 324,
"y": 1296
},
"bottom_left": {
"x": 213,
"y": 1279
},
"width": 110.5,
"height": 28
}
},
{
"text": "https://mistral.ai/news/announcing-mistral-7b/",
"bounds": {
"top_left": {
"x": 333,
"y": 1269
},
"top_right": {
"x": 961,
"y": 1369
},
"bottom_right": {
"x": 956,
"y": 1395
},
"bottom_left": {
"x": 329,
"y": 1297
},
"width": 627.5,
"height": 27
}
}
]
},
{
"text": "1 Introduction",
"bounds": {
"top_left": {
"x": 112,
"y": 1335
},
"top_right": {
"x": 327,
"y": 1368
},
"bottom_right": {
"x": 323,
"y": 1395
},
"bottom_left": {
"x": 109,
"y": 1361
},
"width": 214.5,
"height": 26.5
},
"words": [
{
"text": "1",
"bounds": {
"top_left": {
"x": 113,
"y": 1335
},
"top_right": {
"x": 126,
"y": 1337
},
"bottom_right": {
"x": 122,
"y": 1364
},
"bottom_left": {
"x": 109,
"y": 1362
},
"width": 13,
"height": 27
}
},
{
"text": "Introduction",
"bounds": {
"top_left": {
"x": 156,
"y": 1342
},
"top_right": {
"x": 322,
"y": 1368
},
"bottom_right": {
"x": 317,
"y": 1395
},
"bottom_left": {
"x": 152,
"y": 1368
},
"width": 165.5,
"height": 26.5
}
}
]
},
{
"text": "In the rapidly evolving domain of Natural Language Processing (NLP), the race towards higher model",
"bounds": {
"top_left": {
"x": 100,
"y": 1399
},
"top_right": {
"x": 1122,
"y": 1562
},
"bottom_right": {
"x": 1118,
"y": 1589
},
"bottom_left": {
"x": 97,
"y": 1426
},
"width": 1021.5,
"height": 27
},
"words": [
{
"text": "In",
"bounds": {
"top_left": {
"x": 102,
"y": 1400
},
"top_right": {
"x": 121,
"y": 1403
},
"bottom_right": {
"x": 119,
"y": 1429
},
"bottom_left": {
"x": 99,
"y": 1426
},
"width": 19.5,
"height": 26
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 127,
"y": 1404
},
"top_right": {
"x": 159,
"y": 1409
},
"bottom_right": {
"x": 156,
"y": 1435
},
"bottom_left": {
"x": 124,
"y": 1430
},
"width": 32,
"height": 26
}
},
{
"text": "rapidly",
"bounds": {
"top_left": {
"x": 164,
"y": 1410
},
"top_right": {
"x": 235,
"y": 1421
},
"bottom_right": {
"x": 232,
"y": 1448
},
"bottom_left": {
"x": 161,
"y": 1436
},
"width": 71,
"height": 26.5
}
},
{
"text": "evolving",
"bounds": {
"top_left": {
"x": 241,
"y": 1422
},
"top_right": {
"x": 328,
"y": 1436
},
"bottom_right": {
"x": 325,
"y": 1463
},
"bottom_left": {
"x": 238,
"y": 1449
},
"width": 87,
"height": 27
}
},
{
"text": "domain",
"bounds": {
"top_left": {
"x": 333,
"y": 1437
},
"top_right": {
"x": 409,
"y": 1449
},
"bottom_right": {
"x": 406,
"y": 1476
},
"bottom_left": {
"x": 330,
"y": 1464
},
"width": 76,
"height": 27
}
},
{
"text": "of",
"bounds": {
"top_left": {
"x": 415,
"y": 1450
},
"top_right": {
"x": 437,
"y": 1453
},
"bottom_right": {
"x": 433,
"y": 1481
},
"bottom_left": {
"x": 412,
"y": 1477
},
"width": 21.5,
"height": 27.5
}
},
{
"text": "Natural",
"bounds": {
"top_left": {
"x": 442,
"y": 1454
},
"top_right": {
"x": 517,
"y": 1466
},
"bottom_right": {
"x": 513,
"y": 1494
},
"bottom_left": {
"x": 438,
"y": 1482
},
"width": 75,
"height": 28
}
},
{
"text": "Language",
"bounds": {
"top_left": {
"x": 522,
"y": 1467
},
"top_right": {
"x": 622,
"y": 1482
},
"bottom_right": {
"x": 617,
"y": 1511
},
"bottom_left": {
"x": 518,
"y": 1495
},
"width": 99.5,
"height": 28.5
}
},
{
"text": "Processing",
"bounds": {
"top_left": {
"x": 627,
"y": 1483
},
"top_right": {
"x": 736,
"y": 1501
},
"bottom_right": {
"x": 731,
"y": 1529
},
"bottom_left": {
"x": 623,
"y": 1511
},
"width": 108.5,
"height": 28
}
},
{
"text": "(NLP),",
"bounds": {
"top_left": {
"x": 741,
"y": 1502
},
"top_right": {
"x": 812,
"y": 1513
},
"bottom_right": {
"x": 807,
"y": 1541
},
"bottom_left": {
"x": 736,
"y": 1530
},
"width": 71,
"height": 28
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 818,
"y": 1514
},
"top_right": {
"x": 850,
"y": 1519
},
"bottom_right": {
"x": 844,
"y": 1547
},
"bottom_left": {
"x": 812,
"y": 1542
},
"width": 32,
"height": 28
}
},
{
"text": "race",
"bounds": {
"top_left": {
"x": 855,
"y": 1520
},
"top_right": {
"x": 896,
"y": 1527
},
"bottom_right": {
"x": 890,
"y": 1554
},
"bottom_left": {
"x": 850,
"y": 1548
},
"width": 40.5,
"height": 27.5
}
},
{
"text": "towards",
"bounds": {
"top_left": {
"x": 901,
"y": 1527
},
"top_right": {
"x": 979,
"y": 1540
},
"bottom_right": {
"x": 974,
"y": 1567
},
"bottom_left": {
"x": 896,
"y": 1555
},
"width": 78,
"height": 27.5
}
},
{
"text": "higher",
"bounds": {
"top_left": {
"x": 985,
"y": 1541
},
"top_right": {
"x": 1051,
"y": 1552
},
"bottom_right": {
"x": 1045,
"y": 1578
},
"bottom_left": {
"x": 979,
"y": 1568
},
"width": 66,
"height": 26.5
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 1056,
"y": 1553
},
"top_right": {
"x": 1122,
"y": 1563
},
"bottom_right": {
"x": 1116,
"y": 1589
},
"bottom_left": {
"x": 1050,
"y": 1579
},
"width": 66,
"height": 26
}
}
]
},
{
"text": "performance often necessitates an escalation in model size. However, this scaling tends to increase",
"bounds": {
"top_left": {
"x": 98,
"y": 1428
},
"top_right": {
"x": 1116,
"y": 1589
},
"bottom_right": {
"x": 1113,
"y": 1616
},
"bottom_left": {
"x": 94,
"y": 1453
},
"width": 1018.5,
"height": 26
},
"words": [
{
"text": "performance",
"bounds": {
"top_left": {
"x": 98,
"y": 1429
},
"top_right": {
"x": 228,
"y": 1449
},
"bottom_right": {
"x": 225,
"y": 1475
},
"bottom_left": {
"x": 94,
"y": 1454
},
"width": 130.5,
"height": 25.5
}
},
{
"text": "often",
"bounds": {
"top_left": {
"x": 234,
"y": 1450
},
"top_right": {
"x": 288,
"y": 1458
},
"bottom_right": {
"x": 285,
"y": 1484
},
"bottom_left": {
"x": 230,
"y": 1476
},
"width": 54.5,
"height": 26
}
},
{
"text": "necessitates",
"bounds": {
"top_left": {
"x": 294,
"y": 1459
},
"top_right": {
"x": 417,
"y": 1478
},
"bottom_right": {
"x": 414,
"y": 1505
},
"bottom_left": {
"x": 290,
"y": 1485
},
"width": 123.5,
"height": 26.5
}
},
{
"text": "an",
"bounds": {
"top_left": {
"x": 422,
"y": 1479
},
"top_right": {
"x": 446,
"y": 1483
},
"bottom_right": {
"x": 443,
"y": 1510
},
"bottom_left": {
"x": 419,
"y": 1506
},
"width": 24,
"height": 27
}
},
{
"text": "escalation",
"bounds": {
"top_left": {
"x": 453,
"y": 1484
},
"top_right": {
"x": 556,
"y": 1500
},
"bottom_right": {
"x": 553,
"y": 1527
},
"bottom_left": {
"x": 450,
"y": 1511
},
"width": 103,
"height": 27
}
},
{
"text": "in",
"bounds": {
"top_left": {
"x": 561,
"y": 1501
},
"top_right": {
"x": 581,
"y": 1504
},
"bottom_right": {
"x": 578,
"y": 1531
},
"bottom_left": {
"x": 558,
"y": 1528
},
"width": 20,
"height": 27
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 587,
"y": 1505
},
"top_right": {
"x": 654,
"y": 1516
},
"bottom_right": {
"x": 651,
"y": 1543
},
"bottom_left": {
"x": 584,
"y": 1532
},
"width": 67,
"height": 27
}
},
{
"text": "size.",
"bounds": {
"top_left": {
"x": 659,
"y": 1517
},
"top_right": {
"x": 709,
"y": 1525
},
"bottom_right": {
"x": 706,
"y": 1552
},
"bottom_left": {
"x": 656,
"y": 1544
},
"width": 50,
"height": 27
}
},
{
"text": "However,",
"bounds": {
"top_left": {
"x": 714,
"y": 1525
},
"top_right": {
"x": 813,
"y": 1542
},
"bottom_right": {
"x": 810,
"y": 1568
},
"bottom_left": {
"x": 711,
"y": 1553
},
"width": 99,
"height": 27
}
},
{
"text": "this",
"bounds": {
"top_left": {
"x": 819,
"y": 1542
},
"top_right": {
"x": 858,
"y": 1549
},
"bottom_right": {
"x": 855,
"y": 1576
},
"bottom_left": {
"x": 816,
"y": 1569
},
"width": 39,
"height": 27
}
},
{
"text": "scaling",
"bounds": {
"top_left": {
"x": 863,
"y": 1550
},
"top_right": {
"x": 935,
"y": 1561
},
"bottom_right": {
"x": 932,
"y": 1588
},
"bottom_left": {
"x": 860,
"y": 1576
},
"width": 72,
"height": 26.5
}
},
{
"text": "tends",
"bounds": {
"top_left": {
"x": 940,
"y": 1562
},
"top_right": {
"x": 995,
"y": 1571
},
"bottom_right": {
"x": 992,
"y": 1598
},
"bottom_left": {
"x": 937,
"y": 1589
},
"width": 55,
"height": 27
}
},
{
"text": "to",
"bounds": {
"top_left": {
"x": 1000,
"y": 1572
},
"top_right": {
"x": 1022,
"y": 1576
},
"bottom_right": {
"x": 1019,
"y": 1602
},
"bottom_left": {
"x": 998,
"y": 1598
},
"width": 21.5,
"height": 26
}
},
{
"text": "increase",
"bounds": {
"top_left": {
"x": 1028,
"y": 1577
},
"top_right": {
"x": 1115,
"y": 1591
},
"bottom_right": {
"x": 1112,
"y": 1617
},
"bottom_left": {
"x": 1025,
"y": 1603
},
"width": 87,
"height": 26
}
}
]
},
{
"text": "computational costs and inference latency, thereby raising barriers to deployment in practical,",
"bounds": {
"top_left": {
"x": 94,
"y": 1455
},
"top_right": {
"x": 1110,
"y": 1616
},
"bottom_right": {
"x": 1106,
"y": 1644
},
"bottom_left": {
"x": 90,
"y": 1481
},
"width": 1016,
"height": 27
},
"words": [
{
"text": "computational",
"bounds": {
"top_left": {
"x": 94,
"y": 1456
},
"top_right": {
"x": 248,
"y": 1480
},
"bottom_right": {
"x": 244,
"y": 1506
},
"bottom_left": {
"x": 90,
"y": 1481
},
"width": 154,
"height": 25.5
}
},
{
"text": "costs",
"bounds": {
"top_left": {
"x": 254,
"y": 1480
},
"top_right": {
"x": 310,
"y": 1489
},
"bottom_right": {
"x": 306,
"y": 1516
},
"bottom_left": {
"x": 250,
"y": 1507
},
"width": 56,
"height": 27
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 315,
"y": 1490
},
"top_right": {
"x": 354,
"y": 1496
},
"bottom_right": {
"x": 350,
"y": 1523
},
"bottom_left": {
"x": 311,
"y": 1517
},
"width": 39,
"height": 27
}
},
{
"text": "inference",
"bounds": {
"top_left": {
"x": 363,
"y": 1497
},
"top_right": {
"x": 463,
"y": 1513
},
"bottom_right": {
"x": 459,
"y": 1541
},
"bottom_left": {
"x": 359,
"y": 1525
},
"width": 100,
"height": 28
}
},
{
"text": "latency,",
"bounds": {
"top_left": {
"x": 470,
"y": 1514
},
"top_right": {
"x": 556,
"y": 1528
},
"bottom_right": {
"x": 552,
"y": 1556
},
"bottom_left": {
"x": 466,
"y": 1542
},
"width": 86,
"height": 28
}
},
{
"text": "thereby",
"bounds": {
"top_left": {
"x": 562,
"y": 1529
},
"top_right": {
"x": 640,
"y": 1541
},
"bottom_right": {
"x": 636,
"y": 1570
},
"bottom_left": {
"x": 557,
"y": 1557
},
"width": 78.5,
"height": 28.5
}
},
{
"text": "raising",
"bounds": {
"top_left": {
"x": 649,
"y": 1543
},
"top_right": {
"x": 723,
"y": 1555
},
"bottom_right": {
"x": 718,
"y": 1583
},
"bottom_left": {
"x": 645,
"y": 1571
},
"width": 73.5,
"height": 28
}
},
{
"text": "barriers",
"bounds": {
"top_left": {
"x": 730,
"y": 1556
},
"top_right": {
"x": 816,
"y": 1570
},
"bottom_right": {
"x": 811,
"y": 1598
},
"bottom_left": {
"x": 726,
"y": 1584
},
"width": 85.5,
"height": 28
}
},
{
"text": "to",
"bounds": {
"top_left": {
"x": 822,
"y": 1571
},
"top_right": {
"x": 842,
"y": 1574
},
"bottom_right": {
"x": 838,
"y": 1602
},
"bottom_left": {
"x": 817,
"y": 1599
},
"width": 20.5,
"height": 28
}
},
{
"text": "deployment",
"bounds": {
"top_left": {
"x": 853,
"y": 1576
},
"top_right": {
"x": 980,
"y": 1597
},
"bottom_right": {
"x": 975,
"y": 1623
},
"bottom_left": {
"x": 849,
"y": 1603
},
"width": 126.5,
"height": 26.5
}
},
{
"text": "in",
"bounds": {
"top_left": {
"x": 985,
"y": 1598
},
"top_right": {
"x": 1006,
"y": 1601
},
"bottom_right": {
"x": 1001,
"y": 1627
},
"bottom_left": {
"x": 980,
"y": 1624
},
"width": 21,
"height": 26
}
},
{
"text": "practical,",
"bounds": {
"top_left": {
"x": 1015,
"y": 1603
},
"top_right": {
"x": 1110,
"y": 1619
},
"bottom_right": {
"x": 1105,
"y": 1644
},
"bottom_left": {
"x": 1010,
"y": 1629
},
"width": 95,
"height": 25.5
}
}
]
},
{
"text": "real-world scenarios. In this context, the search for balanced models delivering both high-level",
"bounds": {
"top_left": {
"x": 89,
"y": 1481
},
"top_right": {
"x": 1108,
"y": 1644
},
"bottom_right": {
"x": 1104,
"y": 1671
},
"bottom_left": {
"x": 86,
"y": 1508
},
"width": 1018.5,
"height": 27
},
"words": [
{
"text": "real-world",
"bounds": {
"top_left": {
"x": 90,
"y": 1482
},
"top_right": {
"x": 199,
"y": 1499
},
"bottom_right": {
"x": 195,
"y": 1526
},
"bottom_left": {
"x": 86,
"y": 1509
},
"width": 109,
"height": 27
}
},
{
"text": "scenarios.",
"bounds": {
"top_left": {
"x": 208,
"y": 1501
},
"top_right": {
"x": 316,
"y": 1518
},
"bottom_right": {
"x": 312,
"y": 1545
},
"bottom_left": {
"x": 205,
"y": 1528
},
"width": 107.5,
"height": 27
}
},
{
"text": "In",
"bounds": {
"top_left": {
"x": 325,
"y": 1519
},
"top_right": {
"x": 348,
"y": 1523
},
"bottom_right": {
"x": 344,
"y": 1550
},
"bottom_left": {
"x": 321,
"y": 1546
},
"width": 23,
"height": 27
}
},
{
"text": "this",
"bounds": {
"top_left": {
"x": 355,
"y": 1524
},
"top_right": {
"x": 396,
"y": 1531
},
"bottom_right": {
"x": 392,
"y": 1558
},
"bottom_left": {
"x": 351,
"y": 1551
},
"width": 41,
"height": 27
}
},
{
"text": "context,",
"bounds": {
"top_left": {
"x": 401,
"y": 1532
},
"top_right": {
"x": 486,
"y": 1545
},
"bottom_right": {
"x": 482,
"y": 1572
},
"bottom_left": {
"x": 398,
"y": 1559
},
"width": 84.5,
"height": 27
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 492,
"y": 1546
},
"top_right": {
"x": 527,
"y": 1552
},
"bottom_right": {
"x": 523,
"y": 1579
},
"bottom_left": {
"x": 488,
"y": 1573
},
"width": 35,
"height": 27
}
},
{
"text": "search",
"bounds": {
"top_left": {
"x": 534,
"y": 1553
},
"top_right": {
"x": 599,
"y": 1563
},
"bottom_right": {
"x": 595,
"y": 1590
},
"bottom_left": {
"x": 530,
"y": 1580
},
"width": 65,
"height": 27
}
},
{
"text": "for",
"bounds": {
"top_left": {
"x": 607,
"y": 1564
},
"top_right": {
"x": 640,
"y": 1570
},
"bottom_right": {
"x": 636,
"y": 1597
},
"bottom_left": {
"x": 603,
"y": 1591
},
"width": 33,
"height": 27
}
},
{
"text": "balanced",
"bounds": {
"top_left": {
"x": 646,
"y": 1571
},
"top_right": {
"x": 741,
"y": 1586
},
"bottom_right": {
"x": 737,
"y": 1613
},
"bottom_left": {
"x": 642,
"y": 1598
},
"width": 95,
"height": 27
}
},
{
"text": "models",
"bounds": {
"top_left": {
"x": 746,
"y": 1587
},
"top_right": {
"x": 826,
"y": 1599
},
"bottom_right": {
"x": 822,
"y": 1627
},
"bottom_left": {
"x": 743,
"y": 1614
},
"width": 79.5,
"height": 27.5
}
},
{
"text": "delivering",
"bounds": {
"top_left": {
"x": 831,
"y": 1600
},
"top_right": {
"x": 937,
"y": 1617
},
"bottom_right": {
"x": 933,
"y": 1644
},
"bottom_left": {
"x": 828,
"y": 1627
},
"width": 105.5,
"height": 27
}
},
{
"text": "both",
"bounds": {
"top_left": {
"x": 943,
"y": 1618
},
"top_right": {
"x": 990,
"y": 1626
},
"bottom_right": {
"x": 987,
"y": 1653
},
"bottom_left": {
"x": 939,
"y": 1645
},
"width": 47.5,
"height": 27
}
},
{
"text": "high-level",
"bounds": {
"top_left": {
"x": 998,
"y": 1627
},
"top_right": {
"x": 1108,
"y": 1644
},
"bottom_right": {
"x": 1104,
"y": 1672
},
"bottom_left": {
"x": 994,
"y": 1654
},
"width": 110,
"height": 27.5
}
}
]
},
{
"text": "performance and efficiency becomes critically essential. Our model, Mistral 7B, demonstrates that",
"bounds": {
"top_left": {
"x": 87,
"y": 1510
},
"top_right": {
"x": 1102,
"y": 1670
},
"bottom_right": {
"x": 1098,
"y": 1697
},
"bottom_left": {
"x": 83,
"y": 1535
},
"width": 1015,
"height": 26
},
"words": [
{
"text": "performance",
"bounds": {
"top_left": {
"x": 87,
"y": 1510
},
"top_right": {
"x": 218,
"y": 1530
},
"bottom_right": {
"x": 214,
"y": 1557
},
"bottom_left": {
"x": 83,
"y": 1535
},
"width": 131,
"height": 26
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 224,
"y": 1531
},
"top_right": {
"x": 260,
"y": 1537
},
"bottom_right": {
"x": 256,
"y": 1563
},
"bottom_left": {
"x": 220,
"y": 1558
},
"width": 36,
"height": 26.5
}
},
{
"text": "efficiency",
"bounds": {
"top_left": {
"x": 266,
"y": 1538
},
"top_right": {
"x": 369,
"y": 1554
},
"bottom_right": {
"x": 364,
"y": 1581
},
"bottom_left": {
"x": 262,
"y": 1564
},
"width": 102.5,
"height": 26.5
}
},
{
"text": "becomes",
"bounds": {
"top_left": {
"x": 374,
"y": 1555
},
"top_right": {
"x": 465,
"y": 1569
},
"bottom_right": {
"x": 461,
"y": 1596
},
"bottom_left": {
"x": 370,
"y": 1582
},
"width": 91,
"height": 27
}
},
{
"text": "critically",
"bounds": {
"top_left": {
"x": 471,
"y": 1570
},
"top_right": {
"x": 562,
"y": 1584
},
"bottom_right": {
"x": 557,
"y": 1612
},
"bottom_left": {
"x": 466,
"y": 1597
},
"width": 91,
"height": 27.5
}
},
{
"text": "essential.",
"bounds": {
"top_left": {
"x": 567,
"y": 1585
},
"top_right": {
"x": 667,
"y": 1601
},
"bottom_right": {
"x": 662,
"y": 1629
},
"bottom_left": {
"x": 563,
"y": 1613
},
"width": 99.5,
"height": 28
}
},
{
"text": "Our",
"bounds": {
"top_left": {
"x": 672,
"y": 1602
},
"top_right": {
"x": 712,
"y": 1608
},
"bottom_right": {
"x": 708,
"y": 1636
},
"bottom_left": {
"x": 668,
"y": 1629
},
"width": 40,
"height": 27.5
}
},
{
"text": "model,",
"bounds": {
"top_left": {
"x": 718,
"y": 1609
},
"top_right": {
"x": 789,
"y": 1621
},
"bottom_right": {
"x": 785,
"y": 1648
},
"bottom_left": {
"x": 713,
"y": 1637
},
"width": 71.5,
"height": 27.5
}
},
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 795,
"y": 1622
},
"top_right": {
"x": 873,
"y": 1634
},
"bottom_right": {
"x": 869,
"y": 1661
},
"bottom_left": {
"x": 790,
"y": 1649
},
"width": 78.5,
"height": 27
}
},
{
"text": "7B,",
"bounds": {
"top_left": {
"x": 879,
"y": 1635
},
"top_right": {
"x": 914,
"y": 1641
},
"bottom_right": {
"x": 909,
"y": 1668
},
"bottom_left": {
"x": 874,
"y": 1662
},
"width": 35,
"height": 27
}
},
{
"text": "demonstrates",
"bounds": {
"top_left": {
"x": 919,
"y": 1642
},
"top_right": {
"x": 1055,
"y": 1664
},
"bottom_right": {
"x": 1050,
"y": 1690
},
"bottom_left": {
"x": 914,
"y": 1669
},
"width": 136,
"height": 26.5
}
},
{
"text": "that",
"bounds": {
"top_left": {
"x": 1061,
"y": 1665
},
"top_right": {
"x": 1103,
"y": 1672
},
"bottom_right": {
"x": 1098,
"y": 1697
},
"bottom_left": {
"x": 1056,
"y": 1691
},
"width": 42,
"height": 25.5
}
}
]
},
{
"text": "a carefully designed language model can deliver high performance while maintaining an efficient",
"bounds": {
"top_left": {
"x": 83,
"y": 1537
},
"top_right": {
"x": 1098,
"y": 1697
},
"bottom_right": {
"x": 1093,
"y": 1725
},
"bottom_left": {
"x": 79,
"y": 1563
},
"width": 1014.5,
"height": 27
},
"words": [
{
"text": "a",
"bounds": {
"top_left": {
"x": 84,
"y": 1537
},
"top_right": {
"x": 96,
"y": 1539
},
"bottom_right": {
"x": 92,
"y": 1566
},
"bottom_left": {
"x": 80,
"y": 1564
},
"width": 12,
"height": 27
}
},
{
"text": "carefully",
"bounds": {
"top_left": {
"x": 101,
"y": 1540
},
"top_right": {
"x": 195,
"y": 1554
},
"bottom_right": {
"x": 191,
"y": 1581
},
"bottom_left": {
"x": 98,
"y": 1567
},
"width": 93.5,
"height": 27
}
},
{
"text": "designed",
"bounds": {
"top_left": {
"x": 201,
"y": 1555
},
"top_right": {
"x": 293,
"y": 1569
},
"bottom_right": {
"x": 289,
"y": 1597
},
"bottom_left": {
"x": 197,
"y": 1582
},
"width": 92,
"height": 27.5
}
},
{
"text": "language",
"bounds": {
"top_left": {
"x": 299,
"y": 1570
},
"top_right": {
"x": 395,
"y": 1585
},
"bottom_right": {
"x": 391,
"y": 1613
},
"bottom_left": {
"x": 295,
"y": 1598
},
"width": 96,
"height": 28
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 400,
"y": 1586
},
"top_right": {
"x": 466,
"y": 1596
},
"bottom_right": {
"x": 462,
"y": 1624
},
"bottom_left": {
"x": 396,
"y": 1614
},
"width": 66,
"height": 28
}
},
{
"text": "can",
"bounds": {
"top_left": {
"x": 471,
"y": 1597
},
"top_right": {
"x": 508,
"y": 1603
},
"bottom_right": {
"x": 504,
"y": 1631
},
"bottom_left": {
"x": 467,
"y": 1625
},
"width": 37,
"height": 28
}
},
{
"text": "deliver",
"bounds": {
"top_left": {
"x": 516,
"y": 1604
},
"top_right": {
"x": 587,
"y": 1615
},
"bottom_right": {
"x": 582,
"y": 1643
},
"bottom_left": {
"x": 511,
"y": 1632
},
"width": 71,
"height": 28
}
},
{
"text": "high",
"bounds": {
"top_left": {
"x": 592,
"y": 1616
},
"top_right": {
"x": 640,
"y": 1624
},
"bottom_right": {
"x": 635,
"y": 1652
},
"bottom_left": {
"x": 588,
"y": 1644
},
"width": 47.5,
"height": 28
}
},
{
"text": "performance",
"bounds": {
"top_left": {
"x": 645,
"y": 1625
},
"top_right": {
"x": 779,
"y": 1647
},
"bottom_right": {
"x": 774,
"y": 1674
},
"bottom_left": {
"x": 641,
"y": 1653
},
"width": 133.5,
"height": 27.5
}
},
{
"text": "while",
"bounds": {
"top_left": {
"x": 784,
"y": 1648
},
"top_right": {
"x": 843,
"y": 1657
},
"bottom_right": {
"x": 838,
"y": 1684
},
"bottom_left": {
"x": 779,
"y": 1675
},
"width": 59,
"height": 27
}
},
{
"text": "maintaining",
"bounds": {
"top_left": {
"x": 848,
"y": 1658
},
"top_right": {
"x": 974,
"y": 1679
},
"bottom_right": {
"x": 969,
"y": 1705
},
"bottom_left": {
"x": 843,
"y": 1685
},
"width": 126,
"height": 26.5
}
},
{
"text": "an",
"bounds": {
"top_left": {
"x": 979,
"y": 1680
},
"top_right": {
"x": 1004,
"y": 1684
},
"bottom_right": {
"x": 999,
"y": 1710
},
"bottom_left": {
"x": 974,
"y": 1706
},
"width": 25,
"height": 26
}
},
{
"text": "efficient",
"bounds": {
"top_left": {
"x": 1009,
"y": 1685
},
"top_right": {
"x": 1098,
"y": 1700
},
"bottom_right": {
"x": 1093,
"y": 1726
},
"bottom_left": {
"x": 1005,
"y": 1711
},
"width": 88.5,
"height": 26
}
}
]
},
{
"text": "inference. Mistral 7B outperforms the previous best 13B model (Llama 2, [26]) across all tested",
"bounds": {
"top_left": {
"x": 83,
"y": 1564
},
"top_right": {
"x": 1094,
"y": 1724
},
"bottom_right": {
"x": 1089,
"y": 1752
},
"bottom_left": {
"x": 79,
"y": 1590
},
"width": 1010.5,
"height": 27
},
"words": [
{
"text": "inference.",
"bounds": {
"top_left": {
"x": 84,
"y": 1565
},
"top_right": {
"x": 189,
"y": 1581
},
"bottom_right": {
"x": 184,
"y": 1607
},
"bottom_left": {
"x": 79,
"y": 1591
},
"width": 105,
"height": 26
}
},
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 194,
"y": 1581
},
"top_right": {
"x": 273,
"y": 1594
},
"bottom_right": {
"x": 268,
"y": 1621
},
"bottom_left": {
"x": 189,
"y": 1608
},
"width": 79,
"height": 27
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 279,
"y": 1594
},
"top_right": {
"x": 304,
"y": 1598
},
"bottom_right": {
"x": 299,
"y": 1626
},
"bottom_left": {
"x": 274,
"y": 1622
},
"width": 25,
"height": 28
}
},
{
"text": "outperforms",
"bounds": {
"top_left": {
"x": 313,
"y": 1600
},
"top_right": {
"x": 443,
"y": 1620
},
"bottom_right": {
"x": 438,
"y": 1648
},
"bottom_left": {
"x": 308,
"y": 1627
},
"width": 130,
"height": 27.5
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 448,
"y": 1621
},
"top_right": {
"x": 481,
"y": 1626
},
"bottom_right": {
"x": 476,
"y": 1654
},
"bottom_left": {
"x": 443,
"y": 1649
},
"width": 33,
"height": 28
}
},
{
"text": "previous",
"bounds": {
"top_left": {
"x": 486,
"y": 1627
},
"top_right": {
"x": 576,
"y": 1641
},
"bottom_right": {
"x": 571,
"y": 1670
},
"bottom_left": {
"x": 481,
"y": 1655
},
"width": 90,
"height": 28.5
}
},
{
"text": "best",
"bounds": {
"top_left": {
"x": 582,
"y": 1642
},
"top_right": {
"x": 629,
"y": 1649
},
"bottom_right": {
"x": 624,
"y": 1678
},
"bottom_left": {
"x": 577,
"y": 1671
},
"width": 47,
"height": 29
}
},
{
"text": "13B",
"bounds": {
"top_left": {
"x": 634,
"y": 1650
},
"top_right": {
"x": 673,
"y": 1657
},
"bottom_right": {
"x": 668,
"y": 1685
},
"bottom_left": {
"x": 629,
"y": 1679
},
"width": 39,
"height": 28.5
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 683,
"y": 1658
},
"top_right": {
"x": 749,
"y": 1669
},
"bottom_right": {
"x": 744,
"y": 1697
},
"bottom_left": {
"x": 678,
"y": 1687
},
"width": 66,
"height": 28.5
}
},
{
"text": "(Llama",
"bounds": {
"top_left": {
"x": 755,
"y": 1670
},
"top_right": {
"x": 833,
"y": 1683
},
"bottom_right": {
"x": 828,
"y": 1711
},
"bottom_left": {
"x": 750,
"y": 1698
},
"width": 78,
"height": 28
}
},
{
"text": "2,",
"bounds": {
"top_left": {
"x": 839,
"y": 1684
},
"top_right": {
"x": 857,
"y": 1687
},
"bottom_right": {
"x": 852,
"y": 1715
},
"bottom_left": {
"x": 834,
"y": 1712
},
"width": 18,
"height": 28
}
},
{
"text": "[26])",
"bounds": {
"top_left": {
"x": 863,
"y": 1688
},
"top_right": {
"x": 918,
"y": 1697
},
"bottom_right": {
"x": 913,
"y": 1725
},
"bottom_left": {
"x": 858,
"y": 1716
},
"width": 55,
"height": 28
}
},
{
"text": "across",
"bounds": {
"top_left": {
"x": 924,
"y": 1698
},
"top_right": {
"x": 990,
"y": 1709
},
"bottom_right": {
"x": 985,
"y": 1736
},
"bottom_left": {
"x": 919,
"y": 1725
},
"width": 66,
"height": 27
}
},
{
"text": "all",
"bounds": {
"top_left": {
"x": 996,
"y": 1710
},
"top_right": {
"x": 1023,
"y": 1715
},
"bottom_right": {
"x": 1018,
"y": 1741
},
"bottom_left": {
"x": 991,
"y": 1737
},
"width": 27,
"height": 26.5
}
},
{
"text": "tested",
"bounds": {
"top_left": {
"x": 1028,
"y": 1716
},
"top_right": {
"x": 1091,
"y": 1727
},
"bottom_right": {
"x": 1086,
"y": 1752
},
"bottom_left": {
"x": 1023,
"y": 1742
},
"width": 63,
"height": 25.5
}
}
]
},
{
"text": "benchmarks, and surpasses the best 34B model (LLaMa 34B, [25]) in mathematics and code",
"bounds": {
"top_left": {
"x": 77,
"y": 1591
},
"top_right": {
"x": 1089,
"y": 1750
},
"bottom_right": {
"x": 1084,
"y": 1779
},
"bottom_left": {
"x": 72,
"y": 1616
},
"width": 1012,
"height": 27
},
"words": [
{
"text": "benchmarks,",
"bounds": {
"top_left": {
"x": 78,
"y": 1591
},
"top_right": {
"x": 213,
"y": 1612
},
"bottom_right": {
"x": 208,
"y": 1639
},
"bottom_left": {
"x": 73,
"y": 1617
},
"width": 135,
"height": 26.5
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 219,
"y": 1613
},
"top_right": {
"x": 259,
"y": 1619
},
"bottom_right": {
"x": 253,
"y": 1646
},
"bottom_left": {
"x": 214,
"y": 1640
},
"width": 39.5,
"height": 27
}
},
{
"text": "surpasses",
"bounds": {
"top_left": {
"x": 268,
"y": 1620
},
"top_right": {
"x": 369,
"y": 1636
},
"bottom_right": {
"x": 364,
"y": 1664
},
"bottom_left": {
"x": 263,
"y": 1648
},
"width": 101,
"height": 28
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 375,
"y": 1637
},
"top_right": {
"x": 409,
"y": 1642
},
"bottom_right": {
"x": 403,
"y": 1670
},
"bottom_left": {
"x": 369,
"y": 1665
},
"width": 34,
"height": 28
}
},
{
"text": "best",
"bounds": {
"top_left": {
"x": 417,
"y": 1643
},
"top_right": {
"x": 465,
"y": 1651
},
"bottom_right": {
"x": 459,
"y": 1679
},
"bottom_left": {
"x": 411,
"y": 1671
},
"width": 48,
"height": 28
}
},
{
"text": "34B",
"bounds": {
"top_left": {
"x": 471,
"y": 1652
},
"top_right": {
"x": 511,
"y": 1658
},
"bottom_right": {
"x": 505,
"y": 1686
},
"bottom_left": {
"x": 465,
"y": 1680
},
"width": 40,
"height": 28
}
},
{
"text": "model",
"bounds": {
"top_left": {
"x": 524,
"y": 1660
},
"top_right": {
"x": 593,
"y": 1671
},
"bottom_right": {
"x": 586,
"y": 1700
},
"bottom_left": {
"x": 518,
"y": 1689
},
"width": 68.5,
"height": 29
}
},
{
"text": "(LLaMa",
"bounds": {
"top_left": {
"x": 598,
"y": 1672
},
"top_right": {
"x": 688,
"y": 1687
},
"bottom_right": {
"x": 682,
"y": 1715
},
"bottom_left": {
"x": 592,
"y": 1700
},
"width": 90,
"height": 28
}
},
{
"text": "34B,",
"bounds": {
"top_left": {
"x": 696,
"y": 1688
},
"top_right": {
"x": 748,
"y": 1697
},
"bottom_right": {
"x": 742,
"y": 1724
},
"bottom_left": {
"x": 689,
"y": 1716
},
"width": 52.5,
"height": 27.5
}
},
{
"text": "[25])",
"bounds": {
"top_left": {
"x": 754,
"y": 1697
},
"top_right": {
"x": 812,
"y": 1707
},
"bottom_right": {
"x": 805,
"y": 1735
},
"bottom_left": {
"x": 747,
"y": 1725
},
"width": 58,
"height": 28
}
},
{
"text": "in",
"bounds": {
"top_left": {
"x": 817,
"y": 1708
},
"top_right": {
"x": 839,
"y": 1711
},
"bottom_right": {
"x": 832,
"y": 1739
},
"bottom_left": {
"x": 810,
"y": 1735
},
"width": 22,
"height": 27.5
}
},
{
"text": "mathematics",
"bounds": {
"top_left": {
"x": 848,
"y": 1713
},
"top_right": {
"x": 983,
"y": 1735
},
"bottom_right": {
"x": 976,
"y": 1762
},
"bottom_left": {
"x": 841,
"y": 1740
},
"width": 135,
"height": 27
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 989,
"y": 1736
},
"top_right": {
"x": 1029,
"y": 1743
},
"bottom_right": {
"x": 1021,
"y": 1769
},
"bottom_left": {
"x": 982,
"y": 1763
},
"width": 39.5,
"height": 26.5
}
},
{
"text": "code",
"bounds": {
"top_left": {
"x": 1036,
"y": 1744
},
"top_right": {
"x": 1089,
"y": 1753
},
"bottom_right": {
"x": 1081,
"y": 1779
},
"bottom_left": {
"x": 1029,
"y": 1771
},
"width": 52.5,
"height": 26.5
}
}
]
},
{
"text": "generation. Furthermore, Mistral 7B approaches the coding performance of Code-Llama 7B [20],",
"bounds": {
"top_left": {
"x": 73,
"y": 1617
},
"top_right": {
"x": 1085,
"y": 1778
},
"bottom_right": {
"x": 1080,
"y": 1808
},
"bottom_left": {
"x": 68,
"y": 1645
},
"width": 1012,
"height": 29
},
"words": [
{
"text": "generation.",
"bounds": {
"top_left": {
"x": 74,
"y": 1618
},
"top_right": {
"x": 190,
"y": 1636
},
"bottom_right": {
"x": 185,
"y": 1663
},
"bottom_left": {
"x": 69,
"y": 1645
},
"width": 116,
"height": 27
}
},
{
"text": "Furthermore,",
"bounds": {
"top_left": {
"x": 196,
"y": 1637
},
"top_right": {
"x": 333,
"y": 1658
},
"bottom_right": {
"x": 328,
"y": 1685
},
"bottom_left": {
"x": 191,
"y": 1664
},
"width": 137,
"height": 27
}
},
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 338,
"y": 1659
},
"top_right": {
"x": 418,
"y": 1672
},
"bottom_right": {
"x": 413,
"y": 1699
},
"bottom_left": {
"x": 333,
"y": 1686
},
"width": 80,
"height": 27
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 424,
"y": 1673
},
"top_right": {
"x": 449,
"y": 1677
},
"bottom_right": {
"x": 444,
"y": 1704
},
"bottom_left": {
"x": 418,
"y": 1700
},
"width": 25.5,
"height": 27
}
},
{
"text": "approaches",
"bounds": {
"top_left": {
"x": 457,
"y": 1678
},
"top_right": {
"x": 573,
"y": 1696
},
"bottom_right": {
"x": 568,
"y": 1724
},
"bottom_left": {
"x": 451,
"y": 1705
},
"width": 116.5,
"height": 27.5
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 579,
"y": 1697
},
"top_right": {
"x": 612,
"y": 1702
},
"bottom_right": {
"x": 607,
"y": 1730
},
"bottom_left": {
"x": 573,
"y": 1725
},
"width": 33.5,
"height": 28
}
},
{
"text": "coding",
"bounds": {
"top_left": {
"x": 618,
"y": 1703
},
"top_right": {
"x": 690,
"y": 1715
},
"bottom_right": {
"x": 684,
"y": 1743
},
"bottom_left": {
"x": 612,
"y": 1731
},
"width": 72,
"height": 28
}
},
{
"text": "performance",
"bounds": {
"top_left": {
"x": 696,
"y": 1716
},
"top_right": {
"x": 827,
"y": 1737
},
"bottom_right": {
"x": 821,
"y": 1765
},
"bottom_left": {
"x": 690,
"y": 1744
},
"width": 131,
"height": 28
}
},
{
"text": "of",
"bounds": {
"top_left": {
"x": 832,
"y": 1738
},
"top_right": {
"x": 856,
"y": 1742
},
"bottom_right": {
"x": 851,
"y": 1770
},
"bottom_left": {
"x": 826,
"y": 1766
},
"width": 24.5,
"height": 28
}
},
{
"text": "Code-Llama",
"bounds": {
"top_left": {
"x": 862,
"y": 1742
},
"top_right": {
"x": 993,
"y": 1764
},
"bottom_right": {
"x": 987,
"y": 1793
},
"bottom_left": {
"x": 856,
"y": 1771
},
"width": 131,
"height": 29
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 999,
"y": 1765
},
"top_right": {
"x": 1024,
"y": 1769
},
"bottom_right": {
"x": 1018,
"y": 1798
},
"bottom_left": {
"x": 993,
"y": 1794
},
"width": 25,
"height": 29
}
},
{
"text": "[20],",
"bounds": {
"top_left": {
"x": 1032,
"y": 1770
},
"top_right": {
"x": 1084,
"y": 1778
},
"bottom_right": {
"x": 1078,
"y": 1808
},
"bottom_left": {
"x": 1026,
"y": 1799
},
"width": 52,
"height": 29.5
}
}
]
},
{
"text": "without sacrificing performance on non-code related benchmarks.",
"bounds": {
"top_left": {
"x": 70,
"y": 1644
},
"top_right": {
"x": 733,
"y": 1750
},
"bottom_right": {
"x": 729,
"y": 1777
},
"bottom_left": {
"x": 66,
"y": 1671
},
"width": 663,
"height": 27
},
"words": [
{
"text": "without",
"bounds": {
"top_left": {
"x": 70,
"y": 1645
},
"top_right": {
"x": 147,
"y": 1657
},
"bottom_right": {
"x": 143,
"y": 1684
},
"bottom_left": {
"x": 66,
"y": 1672
},
"width": 77,
"height": 27
}
},
{
"text": "sacrificing",
"bounds": {
"top_left": {
"x": 152,
"y": 1658
},
"top_right": {
"x": 257,
"y": 1674
},
"bottom_right": {
"x": 253,
"y": 1702
},
"bottom_left": {
"x": 148,
"y": 1685
},
"width": 105,
"height": 27.5
}
},
{
"text": "performance",
"bounds": {
"top_left": {
"x": 262,
"y": 1675
},
"top_right": {
"x": 391,
"y": 1695
},
"bottom_right": {
"x": 387,
"y": 1723
},
"bottom_left": {
"x": 259,
"y": 1703
},
"width": 128.5,
"height": 28
}
},
{
"text": "on",
"bounds": {
"top_left": {
"x": 396,
"y": 1696
},
"top_right": {
"x": 423,
"y": 1701
},
"bottom_right": {
"x": 419,
"y": 1728
},
"bottom_left": {
"x": 392,
"y": 1724
},
"width": 27,
"height": 27.5
}
},
{
"text": "non-code",
"bounds": {
"top_left": {
"x": 428,
"y": 1701
},
"top_right": {
"x": 522,
"y": 1717
},
"bottom_right": {
"x": 519,
"y": 1744
},
"bottom_left": {
"x": 424,
"y": 1729
},
"width": 94.5,
"height": 27.5
}
},
{
"text": "related",
"bounds": {
"top_left": {
"x": 528,
"y": 1718
},
"top_right": {
"x": 599,
"y": 1729
},
"bottom_right": {
"x": 595,
"y": 1756
},
"bottom_left": {
"x": 524,
"y": 1745
},
"width": 71,
"height": 27
}
},
{
"text": "benchmarks.",
"bounds": {
"top_left": {
"x": 604,
"y": 1730
},
"top_right": {
"x": 732,
"y": 1751
},
"bottom_right": {
"x": 729,
"y": 1777
},
"bottom_left": {
"x": 601,
"y": 1757
},
"width": 128,
"height": 26.5
}
}
]
},
{
"text": "Mistral 7B leverages grouped-query attention (GQA) [1], and sliding window attention (SWA) [6, 3].",
"bounds": {
"top_left": {
"x": 61,
"y": 1683
},
"top_right": {
"x": 1074,
"y": 1845
},
"bottom_right": {
"x": 1070,
"y": 1873
},
"bottom_left": {
"x": 58,
"y": 1710
},
"width": 1012.5,
"height": 27.5
},
"words": [
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 63,
"y": 1683
},
"top_right": {
"x": 137,
"y": 1695
},
"bottom_right": {
"x": 134,
"y": 1723
},
"bottom_left": {
"x": 60,
"y": 1711
},
"width": 74,
"height": 28
}
},
{
"text": "7B",
"bounds": {
"top_left": {
"x": 142,
"y": 1696
},
"top_right": {
"x": 167,
"y": 1700
},
"bottom_right": {
"x": 165,
"y": 1728
},
"bottom_left": {
"x": 140,
"y": 1724
},
"width": 25,
"height": 28
}
},
{
"text": "leverages",
"bounds": {
"top_left": {
"x": 175,
"y": 1701
},
"top_right": {
"x": 269,
"y": 1716
},
"bottom_right": {
"x": 267,
"y": 1745
},
"bottom_left": {
"x": 173,
"y": 1729
},
"width": 94,
"height": 28.5
}
},
{
"text": "grouped-query",
"bounds": {
"top_left": {
"x": 275,
"y": 1717
},
"top_right": {
"x": 422,
"y": 1741
},
"bottom_right": {
"x": 420,
"y": 1769
},
"bottom_left": {
"x": 272,
"y": 1746
},
"width": 147.5,
"height": 28.5
}
},
{
"text": "attention",
"bounds": {
"top_left": {
"x": 428,
"y": 1742
},
"top_right": {
"x": 515,
"y": 1756
},
"bottom_right": {
"x": 512,
"y": 1784
},
"bottom_left": {
"x": 426,
"y": 1770
},
"width": 86.5,
"height": 28
}
},
{
"text": "(GQA)",
"bounds": {
"top_left": {
"x": 520,
"y": 1756
},
"top_right": {
"x": 590,
"y": 1768
},
"bottom_right": {
"x": 588,
"y": 1796
},
"bottom_left": {
"x": 518,
"y": 1785
},
"width": 70,
"height": 28.5
}
},
{
"text": "[1],",
"bounds": {
"top_left": {
"x": 596,
"y": 1769
},
"top_right": {
"x": 635,
"y": 1775
},
"bottom_right": {
"x": 632,
"y": 1804
},
"bottom_left": {
"x": 593,
"y": 1797
},
"width": 39,
"height": 28.5
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 640,
"y": 1776
},
"top_right": {
"x": 677,
"y": 1782
},
"bottom_right": {
"x": 674,
"y": 1810
},
"bottom_left": {
"x": 638,
"y": 1804
},
"width": 36.5,
"height": 28
}
},
{
"text": "sliding",
"bounds": {
"top_left": {
"x": 683,
"y": 1782
},
"top_right": {
"x": 751,
"y": 1793
},
"bottom_right": {
"x": 748,
"y": 1822
},
"bottom_left": {
"x": 680,
"y": 1811
},
"width": 68,
"height": 29
}
},
{
"text": "window",
"bounds": {
"top_left": {
"x": 756,
"y": 1794
},
"top_right": {
"x": 833,
"y": 1807
},
"bottom_right": {
"x": 831,
"y": 1835
},
"bottom_left": {
"x": 754,
"y": 1823
},
"width": 77,
"height": 28.5
}
},
{
"text": "attention",
"bounds": {
"top_left": {
"x": 843,
"y": 1808
},
"top_right": {
"x": 930,
"y": 1822
},
"bottom_right": {
"x": 927,
"y": 1851
},
"bottom_left": {
"x": 840,
"y": 1837
},
"width": 87,
"height": 29
}
},
{
"text": "(SWA)",
"bounds": {
"top_left": {
"x": 935,
"y": 1823
},
"top_right": {
"x": 1003,
"y": 1834
},
"bottom_right": {
"x": 1001,
"y": 1862
},
"bottom_left": {
"x": 933,
"y": 1852
},
"width": 68,
"height": 28.5
}
},
{
"text": "[6,",
"bounds": {
"top_left": {
"x": 1009,
"y": 1835
},
"top_right": {
"x": 1040,
"y": 1840
},
"bottom_right": {
"x": 1038,
"y": 1868
},
"bottom_left": {
"x": 1006,
"y": 1863
},
"width": 31.5,
"height": 28
}
},
{
"text": "3].",
"bounds": {
"top_left": {
"x": 1046,
"y": 1841
},
"top_right": {
"x": 1073,
"y": 1845
},
"bottom_right": {
"x": 1071,
"y": 1873
},
"bottom_left": {
"x": 1043,
"y": 1869
},
"width": 27.5,
"height": 28
}
}
]
},
{
"text": "GQA significantly accelerates the inference speed, and also reduces the memory requirement during",
"bounds": {
"top_left": {
"x": 59,
"y": 1710
},
"top_right": {
"x": 1063,
"y": 1872
},
"bottom_right": {
"x": 1058,
"y": 1899
},
"bottom_left": {
"x": 55,
"y": 1737
},
"width": 1003.5,
"height": 27
},
"words": [
{
"text": "GQA",
"bounds": {
"top_left": {
"x": 60,
"y": 1711
},
"top_right": {
"x": 113,
"y": 1719
},
"bottom_right": {
"x": 108,
"y": 1747
},
"bottom_left": {
"x": 55,
"y": 1738
},
"width": 53,
"height": 27.5
}
},
{
"text": "significantly",
"bounds": {
"top_left": {
"x": 120,
"y": 1720
},
"top_right": {
"x": 245,
"y": 1740
},
"bottom_right": {
"x": 240,
"y": 1768
},
"bottom_left": {
"x": 115,
"y": 1748
},
"width": 125,
"height": 28
}
},
{
"text": "accelerates",
"bounds": {
"top_left": {
"x": 251,
"y": 1741
},
"top_right": {
"x": 360,
"y": 1759
},
"bottom_right": {
"x": 355,
"y": 1786
},
"bottom_left": {
"x": 246,
"y": 1769
},
"width": 109,
"height": 27.5
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 365,
"y": 1760
},
"top_right": {
"x": 397,
"y": 1765
},
"bottom_right": {
"x": 392,
"y": 1792
},
"bottom_left": {
"x": 360,
"y": 1787
},
"width": 32,
"height": 27
}
},
{
"text": "inference",
"bounds": {
"top_left": {
"x": 403,
"y": 1766
},
"top_right": {
"x": 498,
"y": 1781
},
"bottom_right": {
"x": 492,
"y": 1808
},
"bottom_left": {
"x": 397,
"y": 1793
},
"width": 95,
"height": 27
}
},
{
"text": "speed,",
"bounds": {
"top_left": {
"x": 503,
"y": 1782
},
"top_right": {
"x": 567,
"y": 1792
},
"bottom_right": {
"x": 561,
"y": 1819
},
"bottom_left": {
"x": 498,
"y": 1809
},
"width": 63.5,
"height": 27
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 572,
"y": 1793
},
"top_right": {
"x": 609,
"y": 1799
},
"bottom_right": {
"x": 603,
"y": 1826
},
"bottom_left": {
"x": 566,
"y": 1820
},
"width": 37,
"height": 27
}
},
{
"text": "also",
"bounds": {
"top_left": {
"x": 614,
"y": 1800
},
"top_right": {
"x": 655,
"y": 1807
},
"bottom_right": {
"x": 649,
"y": 1834
},
"bottom_left": {
"x": 609,
"y": 1827
},
"width": 40.5,
"height": 27
}
},
{
"text": "reduces",
"bounds": {
"top_left": {
"x": 660,
"y": 1808
},
"top_right": {
"x": 738,
"y": 1820
},
"bottom_right": {
"x": 732,
"y": 1847
},
"bottom_left": {
"x": 654,
"y": 1834
},
"width": 78,
"height": 26.5
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 743,
"y": 1821
},
"top_right": {
"x": 775,
"y": 1826
},
"bottom_right": {
"x": 769,
"y": 1853
},
"bottom_left": {
"x": 737,
"y": 1848
},
"width": 32,
"height": 27
}
},
{
"text": "memory",
"bounds": {
"top_left": {
"x": 780,
"y": 1827
},
"top_right": {
"x": 865,
"y": 1841
},
"bottom_right": {
"x": 859,
"y": 1867
},
"bottom_left": {
"x": 774,
"y": 1854
},
"width": 85,
"height": 26.5
}
},
{
"text": "requirement",
"bounds": {
"top_left": {
"x": 870,
"y": 1842
},
"top_right": {
"x": 992,
"y": 1862
},
"bottom_right": {
"x": 986,
"y": 1887
},
"bottom_left": {
"x": 864,
"y": 1868
},
"width": 122,
"height": 25.5
}
},
{
"text": "during",
"bounds": {
"top_left": {
"x": 997,
"y": 1863
},
"top_right": {
"x": 1062,
"y": 1874
},
"bottom_right": {
"x": 1056,
"y": 1899
},
"bottom_left": {
"x": 991,
"y": 1888
},
"width": 65,
"height": 25
}
}
]
},
{
"text": "decoding, allowing for higher batch sizes hence higher throughput, a crucial factor for real-time",
"bounds": {
"top_left": {
"x": 55,
"y": 1736
},
"top_right": {
"x": 1061,
"y": 1898
},
"bottom_right": {
"x": 1056,
"y": 1925
},
"bottom_left": {
"x": 51,
"y": 1764
},
"width": 1005.5,
"height": 27.5
},
"words": [
{
"text": "decoding,",
"bounds": {
"top_left": {
"x": 56,
"y": 1736
},
"top_right": {
"x": 159,
"y": 1753
},
"bottom_right": {
"x": 156,
"y": 1782
},
"bottom_left": {
"x": 52,
"y": 1765
},
"width": 103.5,
"height": 29
}
},
{
"text": "allowing",
"bounds": {
"top_left": {
"x": 165,
"y": 1754
},
"top_right": {
"x": 255,
"y": 1769
},
"bottom_right": {
"x": 252,
"y": 1798
},
"bottom_left": {
"x": 161,
"y": 1783
},
"width": 90.5,
"height": 29
}
},
{
"text": "for",
"bounds": {
"top_left": {
"x": 261,
"y": 1770
},
"top_right": {
"x": 292,
"y": 1775
},
"bottom_right": {
"x": 288,
"y": 1804
},
"bottom_left": {
"x": 258,
"y": 1799
},
"width": 30.5,
"height": 29
}
},
{
"text": "higher",
"bounds": {
"top_left": {
"x": 298,
"y": 1776
},
"top_right": {
"x": 367,
"y": 1787
},
"bottom_right": {
"x": 363,
"y": 1816
},
"bottom_left": {
"x": 294,
"y": 1805
},
"width": 69,
"height": 29
}
},
{
"text": "batch",
"bounds": {
"top_left": {
"x": 372,
"y": 1788
},
"top_right": {
"x": 428,
"y": 1797
},
"bottom_right": {
"x": 425,
"y": 1825
},
"bottom_left": {
"x": 369,
"y": 1816
},
"width": 56,
"height": 28
}
},
{
"text": "sizes",
"bounds": {
"top_left": {
"x": 436,
"y": 1798
},
"top_right": {
"x": 487,
"y": 1806
},
"bottom_right": {
"x": 483,
"y": 1835
},
"bottom_left": {
"x": 432,
"y": 1827
},
"width": 51,
"height": 29
}
},
{
"text": "hence",
"bounds": {
"top_left": {
"x": 492,
"y": 1807
},
"top_right": {
"x": 554,
"y": 1817
},
"bottom_right": {
"x": 551,
"y": 1846
},
"bottom_left": {
"x": 489,
"y": 1836
},
"width": 62,
"height": 29
}
},
{
"text": "higher",
"bounds": {
"top_left": {
"x": 560,
"y": 1818
},
"top_right": {
"x": 629,
"y": 1829
},
"bottom_right": {
"x": 625,
"y": 1857
},
"bottom_left": {
"x": 556,
"y": 1846
},
"width": 69,
"height": 28
}
},
{
"text": "throughput,",
"bounds": {
"top_left": {
"x": 634,
"y": 1830
},
"top_right": {
"x": 756,
"y": 1850
},
"bottom_right": {
"x": 753,
"y": 1878
},
"bottom_left": {
"x": 631,
"y": 1858
},
"width": 122,
"height": 28
}
},
{
"text": "a",
"bounds": {
"top_left": {
"x": 762,
"y": 1851
},
"top_right": {
"x": 776,
"y": 1853
},
"bottom_right": {
"x": 772,
"y": 1881
},
"bottom_left": {
"x": 758,
"y": 1878
},
"width": 14,
"height": 27.5
}
},
{
"text": "crucial",
"bounds": {
"top_left": {
"x": 782,
"y": 1854
},
"top_right": {
"x": 853,
"y": 1866
},
"bottom_right": {
"x": 849,
"y": 1893
},
"bottom_left": {
"x": 778,
"y": 1882
},
"width": 71,
"height": 27.5
}
},
{
"text": "factor",
"bounds": {
"top_left": {
"x": 858,
"y": 1867
},
"top_right": {
"x": 920,
"y": 1877
},
"bottom_right": {
"x": 916,
"y": 1903
},
"bottom_left": {
"x": 855,
"y": 1894
},
"width": 61.5,
"height": 26.5
}
},
{
"text": "for",
"bounds": {
"top_left": {
"x": 926,
"y": 1878
},
"top_right": {
"x": 958,
"y": 1883
},
"bottom_right": {
"x": 955,
"y": 1909
},
"bottom_left": {
"x": 922,
"y": 1904
},
"width": 32.5,
"height": 26
}
},
{
"text": "real-time",
"bounds": {
"top_left": {
"x": 964,
"y": 1884
},
"top_right": {
"x": 1060,
"y": 1900
},
"bottom_right": {
"x": 1056,
"y": 1925
},
"bottom_left": {
"x": 960,
"y": 1910
},
"width": 96,
"height": 25.5
}
}
]
},
{
"text": "applications. In addition, SWA is designed to handle longer sequences more effectively at a reduced",
"bounds": {
"top_left": {
"x": 51,
"y": 1763
},
"top_right": {
"x": 1058,
"y": 1924
},
"bottom_right": {
"x": 1054,
"y": 1950
},
"bottom_left": {
"x": 48,
"y": 1791
},
"width": 1006.5,
"height": 27
},
"words": [
{
"text": "applications.",
"bounds": {
"top_left": {
"x": 54,
"y": 1763
},
"top_right": {
"x": 182,
"y": 1784
},
"bottom_right": {
"x": 178,
"y": 1812
},
"bottom_left": {
"x": 50,
"y": 1792
},
"width": 128,
"height": 28.5
}
},
{
"text": "In",
"bounds": {
"top_left": {
"x": 188,
"y": 1785
},
"top_right": {
"x": 209,
"y": 1788
},
"bottom_right": {
"x": 204,
"y": 1817
},
"bottom_left": {
"x": 183,
"y": 1813
},
"width": 21,
"height": 28.5
}
},
{
"text": "addition,",
"bounds": {
"top_left": {
"x": 215,
"y": 1789
},
"top_right": {
"x": 305,
"y": 1804
},
"bottom_right": {
"x": 300,
"y": 1832
},
"bottom_left": {
"x": 210,
"y": 1818
},
"width": 90,
"height": 28.5
}
},
{
"text": "SWA",
"bounds": {
"top_left": {
"x": 310,
"y": 1805
},
"top_right": {
"x": 361,
"y": 1813
},
"bottom_right": {
"x": 356,
"y": 1841
},
"bottom_left": {
"x": 306,
"y": 1833
},
"width": 50.5,
"height": 28
}
},
{
"text": "is",
"bounds": {
"top_left": {
"x": 367,
"y": 1814
},
"top_right": {
"x": 386,
"y": 1817
},
"bottom_right": {
"x": 381,
"y": 1845
},
"bottom_left": {
"x": 362,
"y": 1842
},
"width": 19,
"height": 28
}
},
{
"text": "designed",
"bounds": {
"top_left": {
"x": 392,
"y": 1818
},
"top_right": {
"x": 480,
"y": 1832
},
"bottom_right": {
"x": 475,
"y": 1860
},
"bottom_left": {
"x": 387,
"y": 1846
},
"width": 88,
"height": 28
}
},
{
"text": "to",
"bounds": {
"top_left": {
"x": 486,
"y": 1833
},
"top_right": {
"x": 506,
"y": 1836
},
"bottom_right": {
"x": 500,
"y": 1864
},
"bottom_left": {
"x": 480,
"y": 1861
},
"width": 20,
"height": 28
}
},
{
"text": "handle",
"bounds": {
"top_left": {
"x": 511,
"y": 1837
},
"top_right": {
"x": 580,
"y": 1848
},
"bottom_right": {
"x": 574,
"y": 1876
},
"bottom_left": {
"x": 506,
"y": 1865
},
"width": 68.5,
"height": 28
}
},
{
"text": "longer",
"bounds": {
"top_left": {
"x": 585,
"y": 1849
},
"top_right": {
"x": 650,
"y": 1859
},
"bottom_right": {
"x": 644,
"y": 1887
},
"bottom_left": {
"x": 580,
"y": 1877
},
"width": 64.5,
"height": 28
}
},
{
"text": "sequences",
"bounds": {
"top_left": {
"x": 656,
"y": 1860
},
"top_right": {
"x": 759,
"y": 1877
},
"bottom_right": {
"x": 752,
"y": 1904
},
"bottom_left": {
"x": 650,
"y": 1888
},
"width": 102.5,
"height": 27.5
}
},
{
"text": "more",
"bounds": {
"top_left": {
"x": 764,
"y": 1878
},
"top_right": {
"x": 815,
"y": 1886
},
"bottom_right": {
"x": 808,
"y": 1913
},
"bottom_left": {
"x": 758,
"y": 1905
},
"width": 50.5,
"height": 27
}
},
{
"text": "effectively",
"bounds": {
"top_left": {
"x": 820,
"y": 1887
},
"top_right": {
"x": 929,
"y": 1905
},
"bottom_right": {
"x": 922,
"y": 1930
},
"bottom_left": {
"x": 814,
"y": 1914
},
"width": 108.5,
"height": 26
}
},
{
"text": "at",
"bounds": {
"top_left": {
"x": 934,
"y": 1906
},
"top_right": {
"x": 952,
"y": 1909
},
"bottom_right": {
"x": 945,
"y": 1934
},
"bottom_left": {
"x": 927,
"y": 1931
},
"width": 18,
"height": 25
}
},
{
"text": "a",
"bounds": {
"top_left": {
"x": 958,
"y": 1910
},
"top_right": {
"x": 968,
"y": 1911
},
"bottom_right": {
"x": 961,
"y": 1937
},
"bottom_left": {
"x": 950,
"y": 1935
},
"width": 10.5,
"height": 25.5
}
},
{
"text": "reduced",
"bounds": {
"top_left": {
"x": 974,
"y": 1912
},
"top_right": {
"x": 1056,
"y": 1926
},
"bottom_right": {
"x": 1049,
"y": 1950
},
"bottom_left": {
"x": 967,
"y": 1937
},
"width": 82,
"height": 24.5
}
}
]
},
{
"text": "computational cost, thereby alleviating a common limitation in LLMs. These attention mechanisms",
"bounds": {
"top_left": {
"x": 48,
"y": 1791
},
"top_right": {
"x": 1049,
"y": 1949
},
"bottom_right": {
"x": 1044,
"y": 1975
},
"bottom_left": {
"x": 44,
"y": 1817
},
"width": 1000.5,
"height": 26
},
"words": [
{
"text": "computational",
"bounds": {
"top_left": {
"x": 49,
"y": 1791
},
"top_right": {
"x": 194,
"y": 1814
},
"bottom_right": {
"x": 189,
"y": 1841
},
"bottom_left": {
"x": 44,
"y": 1818
},
"width": 145,
"height": 27
}
},
{
"text": "cost,",
"bounds": {
"top_left": {
"x": 199,
"y": 1815
},
"top_right": {
"x": 246,
"y": 1822
},
"bottom_right": {
"x": 241,
"y": 1849
},
"bottom_left": {
"x": 194,
"y": 1842
},
"width": 47,
"height": 27
}
},
{
"text": "thereby",
"bounds": {
"top_left": {
"x": 251,
"y": 1823
},
"top_right": {
"x": 328,
"y": 1835
},
"bottom_right": {
"x": 323,
"y": 1862
},
"bottom_left": {
"x": 246,
"y": 1850
},
"width": 77,
"height": 27
}
},
{
"text": "alleviating",
"bounds": {
"top_left": {
"x": 333,
"y": 1836
},
"top_right": {
"x": 439,
"y": 1852
},
"bottom_right": {
"x": 434,
"y": 1880
},
"bottom_left": {
"x": 328,
"y": 1863
},
"width": 106,
"height": 27.5
}
},
{
"text": "a",
"bounds": {
"top_left": {
"x": 445,
"y": 1853
},
"top_right": {
"x": 457,
"y": 1855
},
"bottom_right": {
"x": 452,
"y": 1883
},
"bottom_left": {
"x": 440,
"y": 1881
},
"width": 12,
"height": 28
}
},
{
"text": "common",
"bounds": {
"top_left": {
"x": 462,
"y": 1856
},
"top_right": {
"x": 550,
"y": 1870
},
"bottom_right": {
"x": 545,
"y": 1898
},
"bottom_left": {
"x": 457,
"y": 1884
},
"width": 88,
"height": 28
}
},
{
"text": "limitation",
"bounds": {
"top_left": {
"x": 556,
"y": 1871
},
"top_right": {
"x": 655,
"y": 1887
},
"bottom_right": {
"x": 650,
"y": 1914
},
"bottom_left": {
"x": 551,
"y": 1899
},
"width": 99,
"height": 27.5
}
},
{
"text": "in",
"bounds": {
"top_left": {
"x": 660,
"y": 1888
},
"top_right": {
"x": 681,
"y": 1891
},
"bottom_right": {
"x": 676,
"y": 1918
},
"bottom_left": {
"x": 655,
"y": 1915
},
"width": 21,
"height": 27
}
},
{
"text": "LLMs.",
"bounds": {
"top_left": {
"x": 687,
"y": 1892
},
"top_right": {
"x": 761,
"y": 1904
},
"bottom_right": {
"x": 756,
"y": 1931
},
"bottom_left": {
"x": 682,
"y": 1919
},
"width": 74,
"height": 27
}
},
{
"text": "These",
"bounds": {
"top_left": {
"x": 767,
"y": 1905
},
"top_right": {
"x": 824,
"y": 1915
},
"bottom_right": {
"x": 819,
"y": 1940
},
"bottom_left": {
"x": 762,
"y": 1932
},
"width": 57,
"height": 26
}
},
{
"text": "attention",
"bounds": {
"top_left": {
"x": 829,
"y": 1915
},
"top_right": {
"x": 917,
"y": 1930
},
"bottom_right": {
"x": 912,
"y": 1955
},
"bottom_left": {
"x": 824,
"y": 1941
},
"width": 88,
"height": 25.5
}
},
{
"text": "mechanisms",
"bounds": {
"top_left": {
"x": 923,
"y": 1931
},
"top_right": {
"x": 1048,
"y": 1952
},
"bottom_right": {
"x": 1043,
"y": 1975
},
"bottom_left": {
"x": 918,
"y": 1956
},
"width": 125,
"height": 24
}
}
]
},
{
"text": "collectively contribute to the enhanced performance and efficiency of Mistral 7B.",
"bounds": {
"top_left": {
"x": 44,
"y": 1816
},
"top_right": {
"x": 862,
"y": 1946
},
"bottom_right": {
"x": 857,
"y": 1973
},
"bottom_left": {
"x": 42,
"y": 1843
},
"width": 816.5,
"height": 27
},
"words": [
{
"text": "collectively",
"bounds": {
"top_left": {
"x": 46,
"y": 1817
},
"top_right": {
"x": 161,
"y": 1835
},
"bottom_right": {
"x": 158,
"y": 1862
},
"bottom_left": {
"x": 44,
"y": 1843
},
"width": 114.5,
"height": 26.5
}
},
{
"text": "contribute",
"bounds": {
"top_left": {
"x": 166,
"y": 1836
},
"top_right": {
"x": 269,
"y": 1852
},
"bottom_right": {
"x": 266,
"y": 1880
},
"bottom_left": {
"x": 164,
"y": 1863
},
"width": 102.5,
"height": 27.5
}
},
{
"text": "to",
"bounds": {
"top_left": {
"x": 274,
"y": 1853
},
"top_right": {
"x": 295,
"y": 1856
},
"bottom_right": {
"x": 292,
"y": 1884
},
"bottom_left": {
"x": 271,
"y": 1881
},
"width": 21,
"height": 28
}
},
{
"text": "the",
"bounds": {
"top_left": {
"x": 301,
"y": 1857
},
"top_right": {
"x": 332,
"y": 1862
},
"bottom_right": {
"x": 329,
"y": 1890
},
"bottom_left": {
"x": 298,
"y": 1885
},
"width": 31,
"height": 28
}
},
{
"text": "enhanced",
"bounds": {
"top_left": {
"x": 338,
"y": 1863
},
"top_right": {
"x": 433,
"y": 1878
},
"bottom_right": {
"x": 430,
"y": 1907
},
"bottom_left": {
"x": 335,
"y": 1891
},
"width": 95,
"height": 28.5
}
},
{
"text": "performance",
"bounds": {
"top_left": {
"x": 438,
"y": 1879
},
"top_right": {
"x": 567,
"y": 1900
},
"bottom_right": {
"x": 564,
"y": 1928
},
"bottom_left": {
"x": 435,
"y": 1907
},
"width": 129,
"height": 28
}
},
{
"text": "and",
"bounds": {
"top_left": {
"x": 572,
"y": 1901
},
"top_right": {
"x": 609,
"y": 1907
},
"bottom_right": {
"x": 606,
"y": 1934
},
"bottom_left": {
"x": 569,
"y": 1928
},
"width": 37,
"height": 27
}
},
{
"text": "efficiency",
"bounds": {
"top_left": {
"x": 615,
"y": 1908
},
"top_right": {
"x": 714,
"y": 1923
},
"bottom_right": {
"x": 710,
"y": 1950
},
"bottom_left": {
"x": 611,
"y": 1935
},
"width": 99,
"height": 27
}
},
{
"text": "of",
"bounds": {
"top_left": {
"x": 719,
"y": 1924
},
"top_right": {
"x": 740,
"y": 1928
},
"bottom_right": {
"x": 737,
"y": 1954
},
"bottom_left": {
"x": 715,
"y": 1951
},
"width": 21.5,
"height": 26.5
}
},
{
"text": "Mistral",
"bounds": {
"top_left": {
"x": 745,
"y": 1929
},
"top_right": {
"x": 821,
"y": 1941
},
"bottom_right": {
"x": 818,
"y": 1966
},
"bottom_left": {
"x": 742,
"y": 1955
},
"width": 76,
"height": 25.5
}
},
{
"text": "7B.",
"bounds": {
"top_left": {
"x": 827,
"y": 1942
},
"top_right": {
"x": 862,
"y": 1947
},
"bottom_right": {
"x": 858,
"y": 1972
},
"bottom_left": {
"x": 823,
"y": 1967
},
"width": 35,
"height": 25
}
}
]
}
]
}
]
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment