Datasets:

bokane
/

hucker-bakeoff2-paddle16-hard

kind stringclasses 7 values	note stringclasses 8 values	markdown stringclasses 8 values	inference_info stringclasses 1 value
three_column_index_tiny_font	Pulleyblank Outline p187: 3-column vocabulary index, tiny font, Chinese Vocabulary Items	178 Outline of Classical Chinese Grammar qi… hu 其 … 平, wū hū 恐乎. yè hū 也乎, yú shì hū 於是乎 hū zài 乎哉, exclamatory question 146, 262, 535 hū 胡 interrogative pronoun ‘why?’ 10, 91, 95, 96, 107, 142; in a proper name 78 hū 互 137 huò 或 ‘some, someone, something, perhaps’ 92, 130, 134, 135, 136, 79, 80, 156, 478, 479, 4...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
two_column_index	Pulleyblank Outline p197: General Index, two-column	General Index scan 196 Only the main occurrences of grammatical terms are indexed. ablaut 11 active 23, 27, 42, 43, 44, 122 adjective 9, 12, 13, 14, 23, 24-26, 27, 28, 36, 39, 42-44, 55, 62, 99, 100, 101 adnominal 71, 72, 97, 122, 125, 127, 130 adverb 9, 14, 58, 99, 100, 101, 115, 120, 121, 126, 127, 130, 157, 160, 1...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
body_prose_wg_diacritics_cjk	Pulleyblank Outline p60: running prose with inline CJK + Wade-Giles diacritics + numbered examples	50 Outline of Classical Chinese Grammar The phrase suǒ yī 所以 ‘that by which’ must always be given its full value in Classical Chinese. It does not have the meaning ‘therefore’ which it has acquired in the modern language. The expression shì yī 是以 ‘because of that, therefore,’ used as a sentence connective, in contras...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
monospace_dotmatrix_font	Carlitz Rhetoric p30: dot-matrix typewriter body prose, unusual font	24 The Rhetoric of \(\underline{\text{Chin p'ing mei}}\) the transformation, \(\underline{\text{hua}}\), by the force of example, imagined as a wind, \(\underline{\text{feng}}\)). A desire to transform the morals of the age is expressed in prologues to certain plays early in the dynasty, but these are precisely the pl...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
sparse_chapter_calibration	Carlitz Rhetoric p120: sparse chapter-opening page	6 Drama and Song: The Conventions Undermined The author of Chin p'ing mei must have loved drama and song, which he knew so well and used so fluently, but certain things suggest that he found them limited in direction, in characterization, and in the presentation of moral issues. This, after all, may be why he chose ...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
dense_bilingual_dict_smallfont	Ming-Qing dialect dictionary p1: dense two-column bilingual (Chinese headword + citations) small font	A 阿 ☑〈副〉用在动词前，表示疑问，对现在或将来的事发问。作用相当于“吗”。也可用在短语内。□回转头来看见子卖草纸个后生，就叫：“卖草纸个，你阿有萧山，阿有富阳？”（山歌 8 卷）大家向前讨一卦，看道阿能勾到底太平。（山歌 9 卷）记里是书房，弗知阿拉屋里，等我走进去看。（才人福 9 出）耐想俚格人，房子末勿看，铜钱末呢不，耐看俚格人阿靠得住靠勿住。（官场现形记 10 回）故歇借铜钱实头非凡仔难，陆里去开口嘎？耐栈房钱阿欠仔几化？（海天鸿雪记 8 回）☻〈副〉基本意义和⑩同，对过去的事发问，曾否，有没有。今一般用“啊”。□淑人重复短，劈面遇见苏冠香，笑嘻嘻问淑人道：“倪大人到仔陆里去，五少爷阿看见？”（海上花列传 54 回）大人耐...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
dense_bilingual_dict_smallfont	Ming-Qing dialect dictionary p4: dense two-column bilingual small font	4 A 阿养人。(原注：人就薄于孝而厚于慈)(沪谚)②<名>对已婚女性的尊称，有时用于自称。□一到，这些闲管邻舍，一个道：“阿娘，便再住几日，回来作甚？”(清夜钟2回)竹夫人听得气膨膨，出口就骂老惜春：“你是冬来我是夏，缘何牵扰阿娘身？”(山歌8卷)个非阿娘所好，弗如寻几个和尚，与渠笃驾倒好。(笑林广记12卷) 阿奴 <名>年轻女性的自称。我。□有福个情哥弗知吃子阿奴个多少团脐蟹，我个亲夫弗知吃子小阿奴奴多少鳗。(山歌1卷)若还要我归家里，除非是县主登门请阿奴。(擂金凤8回)幼时欢爱之呼曰阿奴，《姑苏志》云：呼小儿为琴儿。琴，子孙也。以虞韵入麻，此方音也。今俗正作奴音。(土风录17卷) 阿奴奴 <名>同“阿奴”。□浮萍草翻身落子水...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]
bold_bilingual_dict_twocolumn	Hucker official-titles dictionary p218: two-column EN/CN terminology	feng-ch’ang ssu 1952–1969 212 1952 fēng-ch’áng ssu 奉常寺 SUI-CH’ING: unofficial reference to the Court of Imperial Sacrifices (tai-ch’ang ssu); from 662 to 671, the official name of the Court. RR: cour des sacrifices impériaux. P27. 1953 fēng-ch’áng tā-fū 奉常大夫 CH’ING: unofficial reference to the Vice Minister (shaoch’i...	[{"model_id": "PaddlePaddle/PaddleOCR-VL-1.6", "model_name": "PaddleOCR-VL-1.6", "model_size": "0.9B", "task_mode": "ocr", "column_name": "markdown", "timestamp": "2026-08-13T22:28:54.245764", "temperature": 0.0, "max_tokens": 4096, "smart_resize": true, "backend": "vllm"}]

Document Processing using PaddleOCR-VL-1.6 (OCR mode)

This dataset contains OCR results from images in bokane/hucker-ocr-bakeoff-hard using PaddleOCR-VL-1.6, an ultra-compact 0.9B OCR model (96.33% SOTA on OmniDocBench v1.6).

Processing Details

Source Dataset: bokane/hucker-ocr-bakeoff-hard
Model: PaddlePaddle/PaddleOCR-VL-1.6
Task Mode: ocr - General text extraction to markdown format
Number of Samples: 8
Processing Time: 2.6 min
Processing Date: 2026-08-13 22:28 UTC

Configuration

Image Column: image
Output Column: markdown
Dataset Split: train
Batch Size: 16
Smart Resize: Enabled
Max Model Length: 8,192 tokens
Max Output Tokens: 4,096
Temperature: 0.0
GPU Memory Utilization: 80.0%

Model Information

PaddleOCR-VL-1.6 is a state-of-the-art, resource-efficient model tailored for document parsing:

🎯 SOTA - 96.33% on OmniDocBench v1.6
🧩 Ultra-compact - Only 0.9B parameters
📝 OCR mode - General text extraction
📊 Table mode - HTML table recognition
📐 Formula mode - LaTeX mathematical notation
📈 Chart mode - Structured chart analysis
🔍 Spotting mode - Text spotting with localization
🔖 Seal mode - Seal/stamp recognition
🌍 Multilingual - Support for multiple languages
🔧 ERNIE-4.5 based - Different architecture from Qwen models

Task Modes

OCR: Extract text content to markdown format
Table Recognition: Extract tables to HTML format
Formula Recognition: Extract mathematical formulas to LaTeX
Chart Recognition: Analyze and describe charts/diagrams
Spotting: Text spotting with localization
Seal Recognition: Seal and stamp recognition

Dataset Structure

The dataset contains all original columns plus:

markdown: The extracted content based on task mode
inference_info: JSON list tracking all OCR models applied to this dataset

Usage

from datasets import load_dataset
import json

# Load the dataset
dataset = load_dataset("{output_dataset_id}", split="train")

# Access the extracted content
for example in dataset:
    print(example["markdown"])
    break

# View all OCR models applied to this dataset
inference_info = json.loads(dataset[0]["inference_info"])
for info in inference_info:
    print(f"Task: {info['task_mode']} - Model: {info['model_id']}")

Reproduction

This dataset was generated using the uv-scripts/ocr PaddleOCR-VL-1.6 script. On HF Jobs, run with the pre-built vLLM image (image-mode) so flashinfer kernels are reused:

hf jobs uv run \
    --image vllm/vllm-openai:latest --flavor a100-large \
    --python /usr/bin/python3 -e PYTHONPATH=/usr/local/lib/python3.12/dist-packages \
    -s HF_TOKEN \
    https://huggingface.co/datasets/uv-scripts/ocr/raw/main/paddleocr-vl-1.6.py \
    bokane/hucker-ocr-bakeoff-hard \
    <output-dataset> \
    --task-mode ocr \
    --image-column image \
    --batch-size 16 \
    --max-model-len 8192 \
    --max-tokens 4096 \
    --gpu-memory-utilization 0.8

Performance

Model Size: 0.9B parameters (smallest among top-tier OCR models)
Processing Speed: ~0.05 images/second
Architecture: NaViT visual encoder + ERNIE-4.5-0.3B language model

Generated with 🤖 UV Scripts

Downloads last month: 24

Number of rows:

Total file size:

6.23 MB