tuandunghcmut/tencent-cloud-router-data

Router Data with Agent Descriptions

An enriched version of the Router/router_data dataset, augmented with agent description metadata (in JSON format) sourced from the TCAndon-Router project by Tencent.

Dataset Description

This dataset consolidates four well-known intent detection benchmarks — CLINC150, HWU64, MINDS14, and SGD — and enriches each sample with the corresponding agent description used by the TCAndon-Router multi-agent routing system.

Columns

Column	Type	Description
query	string	The user query / utterance
intent	string	The ground-truth intent label
dataset	string	Source dataset name (clinc150, hwu64, minds14, sgd)
agent_description	string (JSON)	JSON object containing the agent's name, description, and id as used by TCAndon-Router. null for OOS samples
num_agents	int32	Total number of agents in that dataset's config (e.g., 150 for CLINC150)
is_oos	bool	Whether this sample is an out-of-scope (OOS) query with no matching agent

Agent Description Format

The agent_description column contains a JSON string with the following structure:

{
  "name": "banking:balance",
  "description": "Agent for banking:balance is capable of retrieving and summarizing account balances and funds availability across various banking institutions...",
  "id": "banking:balance"
}

Dataset Statistics

Subset	Train	Test	Num Agents
CLINC150	~170K	~22K	150
HWU64	~25K	~3K	64
MINDS14	~10K	~1K	14
SGD	~42K	~5K	39
Total	~247K	~30K	267

Usage

from datasets import load_dataset
import json

ds = load_dataset("tuandunghcmut/router_data")

# Access a sample
sample = ds["test"][0]
print(sample["query"])
print(sample["intent"])

# Parse agent description
agent_desc = json.loads(sample["agent_description"])
print(agent_desc["name"])
print(agent_desc["description"])

Filter by dataset

import pandas as pd

df = ds["test"].to_pandas()
clinc150 = df[df["dataset"] == "clinc150"]
hwu64 = df[df["dataset"] == "hwu64"]

Source Datasets

CLINC150

• Paper: Larson et al. (2019). An Evaluation Dataset for Intent Classification and Out-of-Scope Prediction. EMNLP 2019.
• Link: https://github.com/clinc/oos-eval
• Description: 150 intent classes across 10 domains, designed for evaluating intent classification and out-of-scope detection.

HWU64

• Paper: Liu et al. (2019). Benchmarking Natural Language Understanding Services for Building Conversational Agents. IWSDS 2019.
• Link: https://github.com/xliuhw/NLU-Evaluation-Data
• Description: 64 intent classes across 21 domains collected from a personal robot assistant use case.

MINDS14

• Paper: Gerz et al. (2021). Multilingual and Cross-Lingual Intent Detection from Spoken Data. EMNLP 2021.
• Link: https://huggingface.co/datasets/PolyAI/minds14
• Description: 14 intent classes in the e-banking domain, covering 14 languages for multilingual intent detection evaluation.

SGD (Schema-Guided Dialogue)

• Paper: Rastogi et al. (2020). Towards Scalable Multi-Domain Conversational Agents: The Schema-Guided Dialogue Dataset. AAAI 2020.
• Link: https://github.com/google-research-datasets/dstc8-schema-guided-dialogue
• Description: Multi-domain, task-oriented dialogue dataset with schema-guided annotations spanning 20+ services.

Agent Descriptions

• Source: Tencent/TCAndon-Router XML configuration files
• Generated by: LLM-based agent description generation pipeline (see generate_agent_desc.py in the original repository)

Citation

If you use this dataset, please cite the original TCAndon-Router paper and the source datasets:

TCAndon-Router

@article{zhao2026TCAndonRouter,
  title={TCAndonRouter: Adaptive Reasoning Router for Multi-Agent Collaboration},
  author={Jiuzhou Zhao, Chunrong Chen, Chenqi Qiao, Lebin Zheng, Minqi Han, Yanchi Liu, Yongzhou Xu, Xiaochuan Xu, Min Zhang},
  journal={arXiv preprint:2601.04544},
  year={2026}
}

CLINC150

@inproceedings{larson2019evaluation,
  title={An Evaluation Dataset for Intent Classification and Out-of-Scope Prediction},
  author={Larson, Stefan and Mahendran, Anish and Patel, Joseph J and Carlson, Jonathan and Leach, Kevin and others},
  booktitle={Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  year={2019}
}

HWU64

@inproceedings{liu2019benchmarking,
  title={Benchmarking Natural Language Understanding Services for Building Conversational Agents},
  author={Liu, Xingkun and Eshghi, Arash and Swietojanski, Pawel and Rieser, Verena},
  booktitle={Proceedings of the 10th International Workshop on Spoken Dialogue Systems Technology (IWSDS)},
  year={2019}
}

MINDS14

@inproceedings{gerz2021multilingual,
  title={Multilingual and Cross-Lingual Intent Detection from Spoken Data},
  author={Gerz, Daniela and Su, Pei-Hao and Kusztos, Razvan and Mondal, Avishek and Lis, Michal and Singhal, Eshan and Mrk{\v{s}}i{\'c}, Nikola and Wen, Tsung-Hsien and Vuli{\'c}, Ivan},
  booktitle={Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  year={2021}
}

SGD

@inproceedings{rastogi2020towards,
  title={Towards Scalable Multi-Domain Conversational Agents: The Schema-Guided Dialogue Dataset},
  author={Rastogi, Abhinav and Zang, Xiaoxue and Sber, Srinivas and Guez, Raghav and Hakkani-T{\"u}r, Dilek},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2020}
}

License

This dataset is released under the Apache 2.0 License, consistent with the original TCAndon-Router repository.