datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
b-sides-hf-yearly-model-indexes
B-Sides Yearly Model Indexes
Semantic search index over 428,440 text generation models on the Hugging Face Hub,
covering 2023 through 2026. This is the data behind the
B-Sides V2 search Space.
Every model gets one 256 dimensional vector. Search is a plain dot product.
What got embedded
This is the part that makes it different from a model card scrape. Most Hub metadata
sets stop at the README. This one goes into the repository and reads the files:
Source… See the full description on the dataset page: https://huggingface.co/datasets/juiceb0xc0de/b-sides-hf-yearly-model-indexes.embodied-model-index
AirGPT Embodied Model Index
This dataset is the versioned catalog used by AirGPT Runtime and the future AirGPT Hub.
It indexes upstream model repositories; it does not redistribute their weights.
Every entry pins an upstream revision and records the runtime adapter, hardware guidance,
license identifier, and AirGPT verification status. Upstream model licenses continue to
apply to model weights and code.
Files:
index.json: current catalog consumed by airgpt hub.… See the full description on the dataset page: https://huggingface.co/datasets/airgpt/embodied-model-index.china-ai-model-index
China AI Model Dataset
A source-linked snapshot of 69 Chinese AI model releases from 21 labs.
Every record carries the URLs it was built from and the date it was last verified.
Maintained by The Latent East — independent English-language
reporting on China's AI industry, decoded from Chinese-language primary sources.
Canonical page: https://latenteast.com/data
Methodology: https://latenteast.com/data#methodology
Live JSON: https://latenteast.com/data/models.json
Live CSV:… See the full description on the dataset page: https://huggingface.co/datasets/JasonSG8848/china-ai-model-index.index-card-blank-detector-spec
Archival index-card blank/content pre-filter
🧪 Draft — hf-model-spec-0.1 · Personal working draft, not an HF convention. Feedback via Community tab.
Purpose
A tiny CPU-runnable binary image classifier that labels an archival index card as blank or content — a cheap pre-filter that skips empty cards before expensive VLM metadata extraction in card-catalogue digitisation pipelines. Companion to the NLS card detector (which crops cards from page scans).
Data… See the full description on the dataset page: https://huggingface.co/datasets/model-specs/index-card-blank-detector-spec.