datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
small-magpie
Smaller Magpie
A collection of smaller Magpie datasets compared to agentlans/magpie.
For argilla/magpie-ultra-v0.1, only instructions rated as good or excellent were selected.
output_quality corresponds to the original dataset’s score_difference, which is the gap between instruct model and base model responses as evaluated by a reward model.
Please see the original dataset for details.
Source
Rows
argilla/magpie-ultra-v0.1
43923
Mxode/Magpie-Pro-10K-GPT4o-mini10000
small-mind-post-training-data
small-mind-companion — post-training data
Every corpus used to post-train a ~2B vision-language model
(google/gemma-4-E2B-it) for long-horizon personalised companion dialogue, in the order it was
used: LoRA SFT → LoRA DPO → on-policy distillation.
Part of the OneBee Datasets
collection.
Contents
Path
Rows
Schema
sft/v0/{train,val}.jsonl
202 / 23
messages
sft/v1/{train,val}.jsonl
2232 / 248
messages
dpo/v0/{train,val}.jsonl
200 / 23
prompt, chosen… See the full description on the dataset page: https://huggingface.co/datasets/arjhinety/small-mind-post-training-data.small-mind-pmb-v0
PMB v0 — Personalised Memory Benchmark
An evaluation benchmark for long-horizon personalised memory in small language models. It asks
whether a model can recall what a specific user told it across many sessions, and — the part most
memory benchmarks skip — whether it can decline to answer when the memory does not contain the
answer.
Built for small-mind-companion, a study of how much of the long-horizon memory gap a
~2B multimodal model can close without scaling parameters.
Part… See the full description on the dataset page: https://huggingface.co/datasets/arjhinety/small-mind-pmb-v0.small-mind-probe-sets
small-mind-companion — probe sets
Two small, unrun probe sets from small-mind-companion. Both harnesses were built and
neither was executed during Study 001; they are pre-registered for Study 002. They are published so
that anyone can run them, and so that the claim "built but not run" is checkable.
Part of the OneBee Datasets
collection.
h22_judgment/ — abliteration and judgment quality (24 probes)
H22: removing a model's general refusal direction increases… See the full description on the dataset page: https://huggingface.co/datasets/arjhinety/small-mind-probe-sets.synthetic-aat-materials
Synthetic AAT Materials Dataset
Dataset Description
This dataset contains 1000 synthetic examples of cultural heritage object descriptions paired with their materials as they would appear in the Getty Art & Architecture Thesaurus (AAT). The data is formatted for training conversational AI models, particularly Qwen3, to identify and extract materials from cultural heritage object descriptions.
Dataset Structure
Each example contains:
messages: Conversation… See the full description on the dataset page: https://huggingface.co/datasets/small-models-for-glam/synthetic-aat-materials.small-model-schema-gym
Small Model Schema Gym Dataset
Deterministically generated chat examples for first-pass JSON compliance on
the project Dream brief and Safety plan contracts.
Files
train.jsonl: 500 training examples.
validation.jsonl: 200 held-out examples.
manifest.json: counts, seed, provenance, and overlap check.
Each row contains:
{
"id": "stable example identifier",
"messages": [
{"role": "system", "content": "..."},
{"role": "user", "content": "..."}… See the full description on the dataset page: https://huggingface.co/datasets/KwabsHug/small-model-schema-gym.small_mc4A colossal, cleaned version of Common Crawl's web crawl corpus.
Based on Common Crawl dataset: "https://commoncrawl.org".
This is the processed version of Google's mC4 dataset by AllenAI.