Dataset Preview
_id string | id string | author string | cardData string | disabled bool | gated string | lastModified timestamp[ns] | likes int64 | trendingScore float64 | private bool | sha string | description string | downloads int64 | downloadsAllTime int64 | tags string | createdAt timestamp[ns] | paperswithcode_id string | citation string | embedding list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
69524c8ad001e56220ced9bc | Alibaba-Apsara/Superior-Reasoning-SFT-gpt-oss-120b | Alibaba-Apsara | {"license": "cc-by-4.0", "task_categories": ["text-generation"], "language": ["en"], "tags": ["code", "math", "scientific-qa", "instruction-following", "reasoning", "thinking", "gpt-oss-120b", "distill"], "size_categories": ["435K"], "configs": [{"config_name": "stage1", "data_files": "Superior-Reasoning-SFT-gpt-oss-12... | false | False | 2026-01-15T06:39:55 | 236 | 115 | false | e9d54e2a3f376fd5c62cafd3c4c99b304cdda698 |
Superior-Reasoning-SFT-gpt-oss-120b
π Overview
The Superior-Reasoning-SFT-gpt-oss-120b dataset is a high-quality, open-source collection containing 435K samples designed to democratize the training of high-performance Long Chain-of-Thought (Long-CoT) models. Unlike standard dis... | 12,983 | 12,983 | ['task_categories:text-generation' 'language:en' 'license:cc-by-4.0'
'size_categories:100K<n<1M' 'format:json' 'modality:text'
'library:datasets' 'library:pandas' 'library:polars'
'library:mlcroissant' 'arxiv:2601.09088' 'arxiv:2512.20908' 'region:us'
'code' 'math' 'scientific-qa' 'instruction-following' 'reasoning... | 2025-12-29T09:40:26 | null | null | [
0.3089316785335541,
0.5958349704742432,
0.5629348754882812,
0.25809532403945923,
0.7968100309371948,
0.4372105300426483,
0.7240143418312073,
0.4595176875591278,
0.9374338984489441,
0.5523331165313721,
0.7086126804351807,
0.2317904531955719,
0.33883732557296753,
0.7165320515632629,
0.1670... |
69676b65aeecdadc87f8da8e | facebook/action100m-preview | facebook | {"license": "fair-noncommercial-research-license", "language": ["en"], "tags": ["video", "action"], "size_categories": ["10M<n<100M"]} | false | False | 2026-01-14T14:24:13 | 100 | 91 | false | c9404b5c9772d6883a2f062945273f171b585275 |
Action100M: A Large-scale Video Action Dataset
Our data can be loaded from the π€ huggingface repo at facebook/action100m-preview where we released 10% of the full Action100M for preview. For examples of loading from local parquet files (from cloned repo) and visualization, see our GitHub repo.
from datasets... | 2,908 | 2,908 | ['language:en' 'license:fair-noncommercial-research-license'
'size_categories:100K<n<1M' 'format:parquet' 'modality:text'
'modality:video' 'library:datasets' 'library:dask' 'library:polars'
'library:mlcroissant' 'region:us' 'video' 'action'] | 2026-01-14T10:09:41 | null | null | [
0.42302507162094116,
0.835867702960968,
0.5068385601043701,
0.1327640563249588,
0.986899197101593,
0.8992253541946411,
0.3560081720352173,
0.32750505208969116,
0.8084222674369812,
0.19109703600406647,
0.34933358430862427,
0.5686920881271362,
0.519487202167511,
0.6172237992286682,
0.33704... |
696b2406e6c69ff4f49745f4 | sojuL/RubricHub_v1 | sojuL | {"license": "apache-2.0", "language": ["zh", "en"], "tags": ["medical", "science", "wirting", "isntruction", "chat", "general"], "pretty_name": "RubricHub", "size_categories": ["100K<n<1M"], "task_categories": ["text-generation", "reinforcement-learning", "question-answering"]} | false | False | 2026-01-20T07:16:51 | 81 | 81 | false | bec50742963ed3672391fecbcc4b60067b9fa8bc |
RubricHub_v1
RubricHub is a large-scale (approximately 110K), multi-domain dataset that provides high-quality rubric-based supervision for open-ended generation tasks. It is constructed via an automated coarse-to-fine rubric generation framework, which integrates principle-guided synthesis, multi-model aggre... | 390 | 390 | ['task_categories:text-generation'
'task_categories:reinforcement-learning'
'task_categories:question-answering' 'language:zh' 'language:en'
'license:apache-2.0' 'size_categories:100K<n<1M' 'format:parquet'
'modality:text' 'library:datasets' 'library:dask' 'library:polars'
'library:mlcroissant' 'arxiv:2601.08430' ... | 2026-01-17T05:54:14 | null | null | [
0.49041885137557983,
0.8460575342178345,
0.9790360331535339,
0.08758077770471573,
0.5944743752479553,
0.8000954389572144,
0.3159536123275757,
0.8329492211341858,
0.33446627855300903,
0.8034713268280029,
0.38892868161201477,
0.330281525850296,
0.33613070845603943,
0.1632225066423416,
0.23... |
6969078587ce326016ddda46 | lightonai/LightOnOCR-mix-0126 | lightonai | {"dataset_info": {"features": [{"name": "key", "dtype": "string"}, {"name": "page_idx", "dtype": "int64"}, {"name": "content", "dtype": "string"}, {"name": "metadata", "struct": [{"name": "element_counts", "struct": [{"name": "formulas", "dtype": "int64"}, {"name": "images", "dtype": "int64"}, {"name": "tables", "dtype... | false | False | 2026-01-23T08:39:35 | 60 | 60 | false | 09e11af7f0aacde1553b4d164049831e5bb7adb7 |
LightOnOCR-mix-0126
LightOnOCR-mix-0126 is a large-scale OCR training dataset built via distillation: a strong visionβlanguage model is prompted to produce naturally ordered full-page transcriptions (Markdown with LaTeX math spans and HTML tables) from rendered document pages. The dataset is designed as supe... | 831 | 831 | ['task_categories:text-to-image' 'task_categories:object-detection'
'language:en' 'language:fr' 'language:de' 'language:es' 'language:it'
'language:ja' 'language:ru' 'language:pl' 'language:nl' 'language:zh'
'language:pt' 'language:bg' 'language:tr' 'language:ur' 'language:hi'
'language:th' 'language:ar' 'language:... | 2026-01-15T15:28:05 | null | null | [
0.823688805103302,
0.8406579494476318,
0.483701229095459,
0.9319063425064087,
0.8324227333068848,
0.09852192550897598,
0.8004595637321472,
0.7389633655548096,
0.8095628619194031,
0.43992146849632263,
0.3524768352508545,
0.11228302121162415,
0.8136829137802124,
0.13404196500778198,
0.6520... |
69607cc44b1761f4d0cf0403 | MiniMaxAI/OctoCodingBench | MiniMaxAI | {"license": "mit", "task_categories": ["text-generation"], "language": ["en"], "tags": ["code", "agent", "benchmark", "evaluation"], "pretty_name": "OctoCodingBench", "size_categories": ["n<1K"]} | false | False | 2026-01-13T13:02:26 | 245 | 55 | false | 1555ecb6650a4448c1f7f714ce82d53f140b3414 |
OctoCodingBench: Instruction-Following Benchmark for Coding Agents
English | δΈζ
π Overview
OctoCodingBench benchmarks scaffold-aware instruction following in repository-grounded agentic coding.
Why OctoCodingBench?
Existing benchmarks (SWE-bench, etc.) focus on task completion β wheth... | 13,077 | 13,077 | ['task_categories:text-generation' 'language:en' 'license:mit'
'size_categories:n<1K' 'format:json' 'modality:text' 'library:datasets'
'library:pandas' 'library:polars' 'library:mlcroissant' 'region:us'
'code' 'agent' 'benchmark' 'evaluation'] | 2026-01-09T03:57:56 | null | null | [
0.8765692710876465,
0.5790888071060181,
0.9762074947357178,
0.9661572575569153,
0.7798381447792053,
0.6735142469406128,
0.9520696401596069,
0.005745115224272013,
0.19977733492851257,
0.021700406447052956,
0.8745269775390625,
0.7878456115722656,
0.034170836210250854,
0.3435129225254059,
0... |
68ba0ffd343a84103b603c45 | Pageshift-Entertainment/LongPage | Pageshift-Entertainment | {"pretty_name": "LongPage", "dataset_name": "LongPage", "library_name": "datasets", "language": ["en"], "license": ["cc-by-4.0", "other"], "task_categories": ["text-generation"], "task_ids": ["language-modeling", "text2text-generation"], "size_categories": ["n<1K"], "source_datasets": ["original"], "annotations_creator... | false | False | 2026-01-20T14:01:26 | 102 | 51 | false | 27d907b6a9f92682110e68ef91f001b4812698d6 |
Overview ππ
The first comprehensive dataset for training AI models to write complete novels with sophisticated reasoning.
π§ Hierarchical Reasoning Architecture β Multi-layered planning traces including character archetypes, story arcs, world rules, and scene breakdowns. A complete cognitive roadmap for l... | 1,999 | 13,284 | ['task_categories:text-generation' 'task_ids:language-modeling'
'task_ids:text2text-generation' 'annotations_creators:machine-generated'
'language_creators:found' 'multilinguality:monolingual'
'source_datasets:original' 'language:en' 'license:cc-by-4.0'
'license:other' 'size_categories:1K<n<10K' 'format:parquet'
'... | 2025-09-04T22:17:33 | null | null | [
0.25633519887924194,
0.18907758593559265,
0.17922718822956085,
0.10687944293022156,
0.6752791404724121,
0.10808505117893219,
0.3827035129070282,
0.5174180865287781,
0.44070863723754883,
0.6763702034950256,
0.6158460974693298,
0.3872328996658325,
0.1837841272354126,
0.4385623037815094,
0.... |
695fb1b373628fa861fe84cf | HuggingFaceFW/finetranslations | HuggingFaceFW | {"license": "odc-by", "task_categories": ["text-generation", "translation"], "pretty_name": "FineTranslations", "size_categories": ["n>1T"], "language": ["abk", "abq", "abs", "acm", "adh", "adi", "ady", "aeb", "afr", "agx", "aii", "aim", "ain", "ajz", "akb", "aln", "als", "alt", "amh", "anp", "aoz", "apc", "apt", "arb"... | false | False | 2026-01-09T16:45:58 | 248 | 42 | false | af3f4ca895450216d4771cdbf3e3b95c5bacaa2a |
π¬ FineTranslations
The world's knowledge in 1+1T tokens of parallel text
What is it?
This dataset contains over 1 trillion tokens of parallel text in English and 500+ languages. It was obtained by translating data from π₯ FineWeb2 into English using Gemma3 27B.
We relied on datatrove's ... | 38,530 | 38,530 | ['task_categories:text-generation' 'task_categories:translation'
'language:abk' 'language:abq' 'language:abs' 'language:acm'
'language:adh' 'language:adi' 'language:ady' 'language:aeb'
'language:afr' 'language:agx' 'language:aii' 'language:aim'
'language:ain' 'language:ajz' 'language:akb' 'language:aln'
'language:... | 2026-01-08T13:31:31 | null | null | [
0.5460655689239502,
0.4864030182361603,
0.5000516176223755,
0.9900186657905579,
0.1692168265581131,
0.7447656393051147,
0.32584118843078613,
0.7762579321861267,
0.7680622339248657,
0.4788389205932617,
0.6054204106330872,
0.6475067734718323,
0.5458201766014099,
0.6172909736633301,
0.92108... |
695df55a4e351abe5277cca5 | UniParser/OmniScience | UniParser | {"license": "cc-by-nc-sa-4.0", "task_categories": ["image-to-text"], "extra_gated_heading": "Request Access to This Dataset", "extra_gated_description": "Please complete the required fields below to request access. Access will be automatically granted upon submission.", "extra_gated_fields": {"Full Name": {"type": "tex... | false | auto | 2026-01-22T02:55:43 | 73 | 41 | false | 9c9fdac9ea87b36e3889330463cd4aee2e81ce95 |
OmniScience: A Large-scale Dataset for Scientific Image Understanding
π 2026-01-21: The OmniScience dataset ranked Top 8 on Hugging Face Datasets Trending (Top 1 on Image Caption Filed). π 2026-01-17: The OmniScience dataset surpassed 5,000 downloads within 5 days of its release. π 2026-01-12: Official r... | 7,703 | 7,710 | ['task_categories:image-to-text' 'license:cc-by-nc-sa-4.0'
'size_categories:1M<n<10M' 'format:parquet' 'format:optimized-parquet'
'modality:image' 'modality:text' 'library:datasets' 'library:dask'
'library:polars' 'library:mlcroissant' 'arxiv:2512.15098' 'region:us'] | 2026-01-07T05:55:38 | null | null | [
0.7002298831939697,
0.6913377046585083,
0.28462982177734375,
0.07918864488601685,
0.4246640205383301,
0.22101019322872162,
0.31037789583206177,
0.22279313206672668,
0.6234893202781677,
0.33018070459365845,
0.2684820592403412,
0.2061517983675003,
0.5017613172531128,
0.3566916584968567,
0.... |
68bb43410b54503c335cb3d8 | HuggingFaceFW/finepdfs | HuggingFaceFW | "{\"license\": \"odc-by\", \"task_categories\": [\"text-generation\"], \"pretty_name\": \"\\ud83d\\u(...TRUNCATED) | false | False | 2026-01-09T10:37:26 | 790 | 39 | false | 89f5411afb089ee310a09df61e7a58a1bf6d081c | "\n\nLiberating 3T of the finest tokens from PDFs\n\n\n\t\n\t\t\n\t\tWhat is this?\n\t\n\nAs we run (...TRUNCATED) | 23,756 | 250,875 | "['task_categories:text-generation' 'language:aai' 'language:aak' ...\n 'arxiv:2506.18421' 'arxiv:21(...TRUNCATED) | 2025-09-05T20:08:33 | null | null | [0.0881318747997284,0.8725446462631226,0.9592418074607849,0.767154335975647,0.7906309962272644,0.931(...TRUNCATED) |
69314c12930718bfbd732f22 | LEMAS-Project/LEMAS-Dataset-train | LEMAS-Project | "{\"license\": \"cc-by-nc-4.0\", \"language\": [\"it\", \"pt\", \"es\", \"fr\", \"de\", \"vi\", \"id(...TRUNCATED) | false | False | 2026-01-09T03:33:49 | 71 | 29 | false | e8bc66643f59bb55097203529022ff809de69c5d | "\n\t\n\t\t\n\t\tOverview\n\t\n\nThis dataset is part of LEMAS-Project (lemas-project.github.io/LEMA(...TRUNCATED) | 16,603 | 17,784 | "['task_categories:text-to-speech'\n 'task_categories:automatic-speech-recognition' 'language:it'\n (...TRUNCATED) | 2025-12-04T08:53:38 | null | null | [0.6445622444152832,0.7504190802574158,0.5153056979179382,0.656548261642456,0.1625341773033142,0.197(...TRUNCATED) |
End of preview.
Hub Stats (Lance format)
This dataset contains Hugging Face Hub statistics in Lance format, converted from the original cfahlgren1/hub-stats dataset.
Files
models.lance- Statistics for all models on the Hub (~2.5M rows)datasets.lance- Statistics for all datasets on the Hubspaces.lance- Statistics for all spaces on the Hub
Usage
import lance
# Load a dataset remotely
ds = lance.dataset("hf://datasets/julien-c/hub-stats-lance/datasets.lance")
# Convert to pandas
df = ds.to_table().to_pandas()
# Or query with SQL-like filters
table = ds.to_table(filter="downloads > 1000")
Example: Query datasets by author
import lance
ds = lance.dataset("hf://datasets/julien-c/hub-stats-lance/datasets.lance")
results = ds.to_table(filter="author = 'microsoft'").to_pandas()
# Sort by downloads
top = results.sort_values("downloads", ascending=False).head(10)
print(top[["id", "likes", "downloads"]])
Output:
id likes downloads
microsoft/ms_marco 221 11120
microsoft/orca-math-word-problems-200k 468 6499
microsoft/bing_coronavirus_query_set 0 6002
microsoft/wiki_qa 69 5737
microsoft/rStar-Coder 225 3492
microsoft/Updesh_beta 8 3223
microsoft/Dayhoff 7 2922
microsoft/meta_woz 6 2801
microsoft/cats_vs_dogs 61 1883
microsoft/IMAGE_UNDERSTANDING 6 1833
Example: Vector similarity search
import lance
import numpy as np
ds = lance.dataset("hf://datasets/julien-c/hub-stats-lance/datasets.lance")
# Get an embedding to use as query (e.g., from microsoft/ms_marco)
query_row = ds.to_table(filter="id = 'microsoft/ms_marco'").to_pandas()
query_embedding = np.array(query_row["embedding"].iloc[0])
# Find 10 nearest neighbors
results = ds.to_table(
nearest={"column": "embedding", "q": query_embedding, "k": 10}
).to_pandas()
print(results[["id", "likes", "downloads", "_distance"]])
Output:
id likes downloads _distance
microsoft/ms_marco 221 11120 2.23
jiwonii97/atalk_as3 0 0 10.61
AI-Art-Collab/ae5 0 1 10.85
wgwgwgwgw/dbbdbbd 0 9 10.90
1FDSFS/56803 0 8 10.94
Why Lance?
Lance is a modern columnar data format optimized for ML workflows:
- Fast random access and filtering
- Efficient for large datasets
- Native support for vector search
- Zero-copy integration with PyArrow/Pandas
- Downloads last month
- 140