datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ToolACE
ToolACE
ToolACE is an automatic agentic pipeline designed to generate Accurate, Complex, and divErse tool-learning data.
ToolACE leverages a novel self-evolution synthesis process to curate a comprehensive API pool of 26,507 diverse APIs.
Dialogs are further generated through the interplay among multiple agents, guided by a formalized thinking process.
To ensure data accuracy, we implement a dual-layer verification system combining rule-based and model-based checks.
More details… See the full description on the dataset page: https://huggingface.co/datasets/lockon/ToolACE.ToolACE
ToolACE
ToolACE is an automatic agentic pipeline designed to generate Accurate, Complex, and divErse tool-learning data.
ToolACE leverages a novel self-evolution synthesis process to curate a comprehensive API pool of 26,507 diverse APIs.
Dialogs are further generated through the interplay among multiple agents, guided by a formalized thinking process.
To ensure data accuracy, we implement a dual-layer verification system combining rule-based and model-based checks.
More details… See the full description on the dataset page: https://huggingface.co/datasets/Team-ACE/ToolACE.toolace-parsed
[PARSED] ToolACE
The data in this dataset is a subset of the original Team-ACE/ToolACE
Subset name
multi-turn
parallel
multiple definition
Last turn type
number of dataset
toolace
yes
yes
yes
complex
11k
This is a re-parsing formatting dataset for the ToolACE official dataset.
Load the dataset
from datasets import load_dataset
ds = load_dataset("minpeter/toolace-parsed")
print(ds)
# DatasetDict({
# train: Dataset({
# features:… See the full description on the dataset page: https://huggingface.co/datasets/minpeter/toolace-parsed.toolace_hermes_tool_useToolACE-Qwen-cleaned
ToolACE for Qwen
Created by: Seungwoo Ryu
Introduction
This dataset is an adaptation of the ToolACE dataset, modified to be directly compatible with Qwen models for tool-calling fine-tuning.
The original dataset was not in a format that could be immediately used for tool-calling training, so we have transformed it accordingly.
This makes it more accessible for training Qwen-based models with function-calling capabilities.
This dataset is applicable to all… See the full description on the dataset page: https://huggingface.co/datasets/tryumanshow/ToolACE-Qwen-cleaned.tool_ace_cleanedtool-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.dfm11-toolace-native-tool-use-repaired
dfm11-toolace-native-tool-use-repaired
ToolACE conversations with declared-name parsing and complete parallel result binding.
This is a DFM11 replacement for schneiderkamplab/dfm10-toolace-native-tool-use. All rows pass exhaustive structural validation. See metadata/manifest.json.
ToolACEtoolace-unified-hallucinations
ToolACE Unified Hallucination Dataset
This repository contains a unified ToolACE-derived dataset for tool-calling hallucination research.
Files
data/train-00000-of-00001.parquet: leakage-safe grouped training split;
data/test-00000-of-00001.parquet: leakage-safe grouped test split.
The split was rebuilt at the normalized dialogue_id level,the same ToolACE dialogue can't appear in different splits.
Schema
column
description
system
system… See the full description on the dataset page: https://huggingface.co/datasets/VirVen/toolace-unified-hallucinations.toolace-tool-calling-hallucination-ragtruth
ToolACE-derived Tool-Calling Hallucination Dataset
This dataset was created for the course assignment Hallucination Detection in Tool Calling.
It is synthetic by design: starting from ToolACE-style tool-calling dialogues, we automatically inject three required hallucination types:
tool_contradiction
overgeneration
missing_tool
Each example follows a RAGTruth-like format:
query: user query
context: tool output
output: final model answer
hallucination_labels: span-level… See the full description on the dataset page: https://huggingface.co/datasets/marrita/toolace-tool-calling-hallucination-ragtruth.ToolACE
ToolACE (tool-calling conversations)
A cleaned-up mirror of the ToolACE function-calling corpus. Every line of
data/train.jsonl is one JSON object describing a full conversation.
Entry format
messages: list of turns.
The first message is always a system message with boilerplate about how
to use tools (the tool definitions themselves are NOT embedded there).
Assistant messages that invoke tools carry tool_calls: a list of
{id, type: "function", function: {name… See the full description on the dataset page: https://huggingface.co/datasets/dusersad12/ToolACE.toolace_tool_call_processing-v4toolace_sequential_tool_use_reasoningTrinity-ToolAce-SFT-splitToolACEToolACE-hallucinated
ToolACE-Hallucinated
Span-level hallucination detection for tool-calling dialogues.
This dataset pairs faithful, tool-grounded assistant answers with controlled, LLM-injected
corruptions of those same answers. Each corrupted answer carries a character-level span label
marking exactly the hallucinated text. It is built for training and evaluating detectors that must
localize where an agent's reply diverges from the tool output it was supposed to be grounded in —
not just whether a… See the full description on the dataset page: https://huggingface.co/datasets/ssslakter/ToolACE-hallucinated.toolace-poolToolACE-Llama-cleaned
ToolACE for LLaMA
Created by: Seungwoo Ryu
Introduction
This dataset is an adaptation of the ToolACE dataset, modified to be directly compatible with LLaMA models for tool-calling fine-tuning.
The original dataset was not in a format that could be immediately used for tool-calling training, so I have transformed it accordingly.
This makes it more accessible for training LLaMA-based models with function-calling capabilities.
This dataset is applicable to all… See the full description on the dataset page: https://huggingface.co/datasets/tryumanshow/ToolACE-Llama-cleaned.toolace_hermes_sequential_tool_useToolACE-sharegpt
Dataset Card for APIGen-MT-5k-sharegpt
This dataset is the sharegpt format of the original Team-ACE/ToolACE dataset. It is primarily designed for fine-tuning large language models (LLMs) for function calling and multi-turn conversations.
Dataset Description
The original Team-ACE/ToolACE dataset contains conversations between users and a language model, focusing on API usage and tool invocation scenarios. We have converted this dataset into the ShareGPT format, which is a… See the full description on the dataset page: https://huggingface.co/datasets/Beryex/ToolACE-sharegpt.toolace-ms
ToolACE (Malay)
A Malay (Bahasa Malaysia) edition of Team-ACE/ToolACE — a large-scale synthetic function-calling corpus with 11,300 multi-turn conversations.
Original format preserved: system (function-composition instruction + tool JSON) and conversations (user/assistant/tool turns, assistant calls written as [API_Name(param=value)]).
In this edition:
System instructions, user queries, and assistant prose are in Malay
Tool and parameter descriptions are in Malay
API call… See the full description on the dataset page: https://huggingface.co/datasets/khursani8/toolace-ms.ToolACE-maskFunctionName-HammerMasking work was performed according to the paper "hammer: robust function-calling for on-device language models via function masking"
基于https://huggingface.co/datasets/tryumanshow/ToolACE-Qwen-cleaned
和 https://huggingface.co/datasets/Team-ACE/ToolACE
Dataset Card for ToolACE Masked Function Name Dataset
数据结构
数据集包含以下字段：
tools: Value(dtype='string', id=None)
conversations: Value(dtype='string', id=None)
system: Value(dtype='string', id=None)
使用示例… See the full description on the dataset page: https://huggingface.co/datasets/FredericFan/ToolACE-maskFunctionName-Hammer.toolace_tool_call_processing-v2ToolACE-Qwen-cleaned-thaimake dataset from https://huggingface.co/datasets/Team-ACE/ToolACE
ToolACE-Japanese
ToolACE
Team-ACE/ToolACEを日本語に翻訳したデータセットです。
翻訳機にはLiquidAI/LFM2-350M-ENJP-MTを使用しました。
This is a Team-ACE/ToolACE dataset translated into Japanese.
The translation engine used was LiquidAI/LFM2-350M-ENJP-MT.
Translation quality
日本語母語話者から見ると不自然な表現が多々あるので、このデータセット単体でモデルをファインチューニングすることはお勧めしません。
Since this dataset contains numerous expressions that appear unnatural to native Japanese speakers, we do not recommend fine-tuning models using this dataset alone.
tool_ace_withclassestoolace_ragtruth_synthetic
ToolACE RAGTruth style synthetic hallucination dataset
This repository contains a synthetic span annotated hallucination detection dataset for tool calling responses.
The data was derived from Team ACE ToolACE conversations. Only grounded instances with a user query, tool call, tool output, and final assistant answer were used.
Motivation
The dataset targets hallucination detection in tool augmented generation, where a final assistant response should be grounded in the… See the full description on the dataset page: https://huggingface.co/datasets/Ali-Bhai/toolace_ragtruth_synthetic.Dans-Toolmaxx-Functions-ToolACEToolACE-Qwen-cleaned