datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
SimpleRL-Zoo-Datagpt-edit-simplersimplerl-minerva-mathsimplerl-OlympiadBenchocto-small-simpler-cube-stack-rollout-bank-50
Octo-Small SIMPLER cube-stack rollout bank
This bank contains exactly 50 deterministic Octo-Small
rollouts for StackGreenCubeOnYellowCubeBakedTexInScene-v1: 3
successes and 47 failures. Every episode has a complete
61-frame H.264 video, a seven-frame contact sheet, losslessly stored per-step
telemetry, derived phase/geometry metrics, an external-LLM diagnosis, exact
evidence values, and a future action-patch hypothesis for failures.
Dataset:… See the full description on the dataset page: https://huggingface.co/datasets/lsnu/octo-small-simpler-cube-stack-rollout-bank-50.simpleRW-lite
Dataset Card for "simpleRW-lite"
interleaved simple wikipedia LM + refinedweb-3m
DatasetDict({
train: Dataset({
features: ['text'],
num_rows: 452484
})
validation: Dataset({
features: ['text'],
num_rows: 11908
})
test: Dataset({
features: ['text'],
num_rows: 11908
})
})
train
Descriptive Stats Using Pandas:
count 452484.000000
mean 430.923633
std 1391.959655
min 0.000000
25%… See the full description on the dataset page: https://huggingface.co/datasets/pszemraj/simpleRW-lite.simpler-sim-oracle-action-interventions
SIM Oracle Working Interventions v1
This is the simulator-oracle edition of the action-patching pipeline. It
uses privileged simulator metrics and is not deployment eligible. It does not
use metric estimators. The repository also describes the non-oracle,
metric-estimator path; these videos must not be represented as evidence that
those estimators recovered the same signals.
The dataset contains 132 of 132 exposed paired
episodes (complete). Every row
has a full baseline MP4… See the full description on the dataset page: https://huggingface.co/datasets/lsnu/simpler-sim-oracle-action-interventions.simple_r1simple_recipesThis is a simple recipes dataset, obtained by formatting/cleaning this one, that I think it was just made by scrapping the food.com website.
Here's the cleanup script I used to obtain it.
from datasets import load_dataset
def clean_recipe(recipe):
recipe = recipe.replace(" , ", ", ")
recipe = recipe.replace('"', "'")
recipe = recipe.replace("\\'", "'")
recipe = recipe.strip("\\']")
recipe = recipe.strip("['")
splitted = recipe.split("\', \'")
recipe =… See the full description on the dataset page: https://huggingface.co/datasets/tengomucho/simple_recipes.echidna-simplerag-massive
echidna-simplerag-massive
Echidna — large SimpleRAG assistant dataset.
Contents
simplerag_massive.jsonl (120 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for the Echidna RAG assistant (Michael Anthony Falabella).
snowfox-simplerag-phase5
snowfox-simplerag-phase5
SnowFox — SimpleRAG phase 5 (messages).
Contents
simplerag_snowfox_phase5_messages.jsonl (185 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for SnowFox / SimpleRAG (Michael Anthony Falabella).
imgedit-simplerechidna-simplerag-massive-combined
echidna-simplerag-massive-combined
Echidna — combined SimpleRAG massive dataset.
Contents
simplerag_massive_combined.jsonl (152 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for the Echidna RAG assistant (Michael Anthony Falabella).
Simple-RL-Zoo-Data-Deduplicatedechidna-round3-simplerag
echidna-round3-simplerag
Echidna — round 3 SimpleRAG-specific extraction examples.
Contents
round3_simplerag.jsonl (18 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for the Echidna RAG assistant (Michael Anthony Falabella).
snowfox-simplerag-phase4
snowfox-simplerag-phase4
SnowFox — SimpleRAG phase 4 (messages).
Contents
simplerag_snowfox_phase4_messages.jsonl (185 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for SnowFox / SimpleRAG (Michael Anthony Falabella).
SimpleRL-Zoosimple-rlhf-rmsimple-rlhf-rlsimplerlsimple_rawSimplerStories
📘📕 SimplerStories 📙📗
SimplerStories is a slight extension of the SimpleStories/SimpleStories dataset. It adds a simplified column, which has a version of each story in more simplified, less flowery language appropriate for a 4-5 year old. This change was done via gemini-2.0-flash.
The rest of the card is for the original SimpleStories dataset.
SimpleStories is dataset of >2 million model-generated short stories. It was made to train small, interpretable language models on it.… See the full description on the dataset page: https://huggingface.co/datasets/mtaran/SimplerStories.simplerl-critique-offlinesimplerl_qwen_level1to4import re
import datasets
ds = datasets.load_dataset("hkust-nlp/SimpleRL-Zoo-Data",
data_files="simplelr_qwen_level1to4/train.parquet",
split="train")
def extract_user_content(content):
pattern = r'<\|im_start\|>user\n(.*?)<\|im_end\|>'
match = re.search(pattern, content, re.DOTALL)
if match:
return match.group(1).strip()
return None
def map_func(example):
content =… See the full description on the dataset page: https://huggingface.co/datasets/rasdani/simplerl_qwen_level1to4.simple_rl_level3to5simplerl-math8ksimple-robot-manipulation
Simple Robot Manipulation Dataset
Task
Pick-and-place manipulation with a 6-DOF robotic arm.
Sensors
RGB camera (30 FPS)
Joint encoders
Actions
Cartesian delta position (x, y, z)
Delta orientation (roll, pitch, yaw)
Binary gripper control
Use Cases
Imitation learning
Reinforcement learning
Vision-Language-Action models
License
MIT
SimpleRL-Zoo-Data-random-5Ksimplified_simplerl_qwen_level1to4SimpleRL-Zoo-Cleaned