datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
data-gouv-datasets-catalog
📢 Sondage 2026 : Utilisation des datasets publiques de MediaTech
Vous utilisez ce dataset ou d’autres datasets de notre collection MediaTech ? Votre avis compte !
Aidez-nous à améliorer nos datasets publiques en répondant à ce sondage rapide (5 min) : 👉 https://grist.numerique.gouv.fr/o/albert/forms/gF4hLaq9VvUog6c5aVDuMw/11
Merci pour votre contribution ! 🙌
🇫🇷 Data.gouv.fr Datasets Catalog
This dataset contains a processed and embedded version of the… See the full description on the dataset page: https://huggingface.co/datasets/AgentPublic/data-gouv-datasets-catalog.data-agent
📈 Data Agent
Data-analysis tasks as a plain, load-and-go dataset — no runtime, no framework required. Each
row is one self-contained task: a real tabular dataset, a question about it, and a
deterministically-checkable gold answer. Load it, prompt any model however you like, and grade the
result with the bundled grader.
Where it comes from
Built from the jupyter-agent dataset
— real data-science notebooks over Kaggle datasets. Every question–answer pair was… See the full description on the dataset page: https://huggingface.co/datasets/FineEnvs/data-agent.terminal-agent-sft-data-v2Multi-Agent_Reinforcement_Learning_Trading_System_Data
📊 Multi-Agent RL Trading System - Dataset
This dataset contains historical OHLCV (Open, High, Low, Close, Volume) data for AAPL, MSFT, and GOOGL, pre-processed for Reinforcement Learning based trading systems.
📁 Dataset Content
The dataset consists of CSV files downloaded via yfinance:
AAPL.csv: Apple Inc. daily data (Jan 2018 - Dec 2024).
MSFT.csv: Microsoft Corp. daily data (Jan 2018 - Dec 2024).
GOOGL.csv: Alphabet Inc. daily data (Jan 2018 - Dec 2024).
📝… See the full description on the dataset page: https://huggingface.co/datasets/AdityaaXD/Multi-Agent_Reinforcement_Learning_Trading_System_Data.data-agent-daytona-repro
Daytona data-agent reproduction
Private frozen runtime, 1,000 training tasks, 250 evaluation tasks and baseline evidence.
See the HF launch scripts in the archived hf/ directory. No credentials are included.
Citation
@misc{fineenvs,
author = {Kolavi, Adithya S},
title = {FineEnvs: Open Source RL Environments for LLM Agents},
year = {2026},
url = {https://github.com/adithya-s-k/FineEnvs}
}
Multi-Agent_Reinforcement_Learning_Trading_System_Data
📊 Multi-Agent RL Trading System - Dataset
This dataset contains historical OHLCV (Open, High, Low, Close, Volume) data for AAPL, MSFT, and GOOGL, pre-processed for Reinforcement Learning based trading systems.
📁 Dataset Content
The dataset consists of CSV files downloaded via yfinance:
AAPL.csv: Apple Inc. daily data (Jan 2018 - Dec 2024).
MSFT.csv: Microsoft Corp. daily data (Jan 2018 - Dec 2024).
GOOGL.csv: Alphabet Inc. daily data (Jan 2018 - Dec 2024).… See the full description on the dataset page: https://huggingface.co/datasets/sanjaydoss/Multi-Agent_Reinforcement_Learning_Trading_System_Data.data-agent-sft
🛠️ Data Agent — SFT
4,677 worked examples of an agent doing data science the right way. Each row is a complete,
verified-correct trajectory: read the question, poke at the data with a shell tool, reason,
compute, and write the answer. Every one of these solved its task and passed a deterministic grader
— so you're fine-tuning on demonstrations that are known to be correct, not just plausible.
Drop-in ready for TRL: conversational messages + tools.
Where it comes… See the full description on the dataset page: https://huggingface.co/datasets/FineEnvs/data-agent-sft.tool-reasoning-sft-TOOLS-toolace-sft-tool-use-agent-data-cleaned-rectified
ToolACE - Tool-Use Agent Data Cleaned & Rectified
👥 Follow the Author
Aman Priyanshu
Overview
This dataset is a cleaned and restructured version of the Team-ACE/ToolACE dataset. ToolACE is a high-quality conversational tool-use dataset containing 11,300+ examples of natural language interactions requiring function calling across diverse domains. This version converts the original OpenAI function-call format into a standardized multi-turn tool-use… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/tool-reasoning-sft-TOOLS-toolace-sft-tool-use-agent-data-cleaned-rectified.agent-env-scaling-data
Agent Environment Scaling Reviewed Data
This Dataset repository stores reviewed data artifacts separately from the implementation
repository. index.json identifies the current accepted snapshot. The tables/ directory exposes
small JSONL views for the Dataset Viewer; snapshots/ preserves portable content-addressed Stores,
readable previews, manifests, and exact validation boundaries.
Current snapshot
warehouse-tongyi-formal-20260806 is the newest complete result… See the full description on the dataset page: https://huggingface.co/datasets/graycatHCO3/agent-env-scaling-data.tool-reasoning-sft-RESEARCH-dr-tulu-sft-deep-research-agent-data-cleaned-rectified
Deep Research - Tulu SFT Data Cleaned Rectified
👥 Follow the Author
Supriti Vijay
Overview
This dataset is a cleaned and restructured version of the DR-TULU SFT dataset released by AllenAI's RL Research team. The original DR-TULU dataset represents significant work in creating high-quality training data for reasoning-enhanced language models with tool use capabilities. This version addresses structural issues in the original release while preserving… See the full description on the dataset page: https://huggingface.co/datasets/SupritiVijay/tool-reasoning-sft-RESEARCH-dr-tulu-sft-deep-research-agent-data-cleaned-rectified.data_agent
data_agent
Plain, Harbor-free version of the data-analysis agent tasks — usable directly via load_dataset.
Splits: train 5000, test 250, eval 144. Deterministic grading, no LLM judge.
Columns
task_id, source_row_id — ids
question — the question to answer
answer — gold answer; reward_mode (numeric/exact_short/exact_bool/list/list_csv/flexible), atol/rtol — how to grade
difficulty_level (1-5), difficulty_tier (easy/medium/hard)
kaggle_dataset — source Kaggle… See the full description on the dataset page: https://huggingface.co/datasets/AdithyaSK/data_agent.ecommerce-ai-data-analyst-agent-benchmark
E-commerce AI Data Analyst Agent Benchmark
A synthetic e-commerce dataset for evaluating AI data analyst agents on
realistic, multi-step business analysis, data-quality investigation, and
analytical reasoning.
This dataset is part of the
E-commerce AI Data Analyst Agent Benchmark.
Dataset summary
This dataset supports evaluation of AI data analyst agents on realistic,
multi-step e-commerce analysis.
It contains:
customers.csv
products.csv
orders.csv
returns.csv… See the full description on the dataset page: https://huggingface.co/datasets/Omcrec/ecommerce-ai-data-analyst-agent-benchmark.data_agent_harbor_train_sft
data_agent_harbor_train_sft
4677 verified agent trajectories (SFT) for the data_agent_harbor_train environments. TRL-ready tool-calling format: messages + tools columns.
Each row is a reward=1 rollout — instruction -> bash tool calls (shell commands) -> final answer — graded deterministically (no LLM judge). Single bash tool throughout.
Columns
messages: OpenAI/TRL chat format (system, user, assistant+tool_calls, tool, ...). tool_calls[].function.arguments are… See the full description on the dataset page: https://huggingface.co/datasets/AdithyaSK/data_agent_harbor_train_sft.sft_data_vsi_wo_video_hint
agent-traces-data-pipeline-debugging
Agent Traces: data-pipeline-debugging
Synthetic multi-agent workflow traces with LLM-enriched content for the data-pipeline-debugging domain.
Part of the juliensimon/open-agent-traces collection — 10 datasets covering diverse domains and workflow patterns.
What is this dataset?
This dataset contains 2,033 events across 50 workflow runs, each representing a complete multi-agent execution trace. Every trace includes:
Agent reasoning — chain-of-thought for each… See the full description on the dataset page: https://huggingface.co/datasets/tesraghavan/agent-traces-data-pipeline-debugging.sft_data_longvila_wo_video_hint
agent-traces-data-pipeline-debugging
Agent Traces: data-pipeline-debugging
Synthetic multi-agent workflow traces with LLM-enriched content for the data-pipeline-debugging domain.
Part of the juliensimon/open-agent-traces collection — 10 datasets covering diverse domains and workflow patterns.
What is this dataset?
This dataset contains 2,033 events across 50 workflow runs, each representing a complete multi-agent execution trace. Every trace includes:
Agent reasoning — chain-of-thought for each agent step… See the full description on the dataset page: https://huggingface.co/datasets/juliensimon/agent-traces-data-pipeline-debugging.Agent-Trajectory-Data-Sample
Agent-Trajectory-Dataset
Description
This dataset covers office-based scenarios such as in-depth searches, data analysis, and industry research, encompassing complete multi-turn reasoning trajectories and tool-calling chains. It is designed to support the analysis of agent planning capabilities, research into tool selection strategies, and quality assessment, providing a structured benchmark for agent training and evaluation.
For more details, please refer to the… See the full description on the dataset page: https://huggingface.co/datasets/Nexdata-AI/Agent-Trajectory-Data-Sample.dataagent-dabench-altimate-noctx-sft-410
DataAgent DAB Altimate NoCtx SFT 410
This dataset contains 410 validated SFT trajectories generated by an Altimate-centered DataAgent pipeline. The paired RL release is forseasons/dataagent-dabench-altimate-noctx-rl-500.
Contents
train_sft.jsonl: final SFT rows, one selected trajectory per task.
train_sft.parquet: parquet version of the same final rows.
summary.json: merge/filter summary.
dataset_package_manifest.json: row counts, statistics, checksums.
audit/:… See the full description on the dataset page: https://huggingface.co/datasets/forseasons/dataagent-dabench-altimate-noctx-sft-410.data_agent_rl
AdithyaSK/data_agent_rl
Source-of-truth eval/train split for the data-agent RL pipeline.
Derived from jupyter-agent/jupyter-agent-dataset by:
Filtering to executor_type == "e2b" (29,555 rows survive; executor_type == "llm" rows have mismatched Kaggle metadata and are dropped).
Per-row classification of the gold answer into a default reward grading mode (see reward_mode_initial).
Stratified sampling by (reward_mode_initial × package_tier) with a max-K-per-Kaggle cap on the eval… See the full description on the dataset page: https://huggingface.co/datasets/AdithyaSK/data_agent_rl.qwen3.5-2b-data-agent-subset100-eval
Qwen3.5-2B on data_agent_rl_environment_train_subset_100
pass@1: 64/100 = 64% · model Qwen/Qwen3.5-2B · temperature 0.7 · 1 rollout/task · max 10 code turns
Task environments ran as Modal sandboxes (one container per task, Kaggle slice pulled from the HF bucket into /home/user/input). Rewards come from each task's own tests/grader.py with the LLM-judge tier disabled, so grading is deterministic: exact string match, else numeric match within 1e-3.
Where the… See the full description on the dataset page: https://huggingface.co/datasets/Nyandwi/qwen3.5-2b-data-agent-subset100-eval.agentradar-dataPersonal-Finance-DataTest_Data_Ex
Test_Data_Ex
This dataset was generated using gibbonbot.
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot.
To get started in robotics, get your own gibbonbot starter pack..
test_data_t
test_data_t
This dataset was generated using gibbonbot.
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot.
To get started in robotics, get your own gibbonbot starter pack..
coding-agent-synth-dataTest_Data_A
Test_Data_A
This dataset was generated using gibbonbot.
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot.
To get started in robotics, get your own gibbonbot starter pack..
qwen3-4b-data-agent-subset100-eval
Qwen/Qwen3-4B on data_agent_rl_environment_train_subset_100
pass@1: 81/100 = 81% · model Qwen/Qwen3-4B · temperature 0.7 · 1 rollout/task · max 10 code turns
Task environments ran as Modal sandboxes (one container per task, Kaggle slice pulled from the HF bucket into /home/user/input). Rewards come from each task's own tests/grader.py with the LLM-judge tier disabled, so grading is deterministic: exact string match, else numeric match within 1e-3.
Where the… See the full description on the dataset page: https://huggingface.co/datasets/Nyandwi/qwen3-4b-data-agent-subset100-eval.agentcall-dataqwen3.5-9b-data-agent-subset100-eval
Qwen3.5-2B on data_agent_rl_environment_train_subset_100
pass@1: 89/100 = 89% · model Qwen/Qwen3.5-9B · temperature 0.7 · 1 rollout/task · max 10 code turns
Task environments ran as Modal sandboxes (one container per task, Kaggle slice pulled from the HF bucket into /home/user/input). Rewards come from each task's own tests/grader.py with the LLM-judge tier disabled, so grading is deterministic: exact string match, else numeric match within 1e-3.
Where the… See the full description on the dataset page: https://huggingface.co/datasets/Nyandwi/qwen3.5-9b-data-agent-subset100-eval.