- [2026.01] π TraveLLaMA is accepted to AAAI 2026 as an Oral presentation!
- [2025.12] π Paper released on arXiv.
- [2025.12] π Project page is online.
- Release paper on arXiv
- Release project page
- Release TravelQA dataset
- Release evaluation code
- Release training code
- Release model checkpoints
- Release demo
TraveLLaMA is a specialized multimodal language model for comprehensive travel assistance. Our key contributions:
-
π TravelQA Dataset: A large-scale dataset with 265K QA pairs:
- 160K text-based QA from authentic travel sources
- 100K vision-language QA featuring maps and location imagery
- 5K expert-annotated Chain-of-Thought reasoning examples
- Covering 35+ cities worldwide
-
π§ Travel-CoT Reasoning: A structured reasoning framework that:
- Decomposes travel queries into spatial, temporal, and practical dimensions
- Improves accuracy by 10.8% with interpretable reasoning paths
-
π€ Interactive Agent System:
- ReAct-based agent with real-time service integration
- SUS score of 82.5 (Excellent) in user studies with 500 participants
| Split | Text QA | Vision-Language QA | CoT Examples | Total |
|---|---|---|---|---|
| Train | 140K | 85K | 4K | 229K |
| Val | 10K | 7.5K | 0.5K | 18K |
| Test | 10K | 7.5K | 0.5K | 18K |
| Total | 160K | 100K | 5K | 265K |
Download the dataset:
# Coming soon
# huggingface-cli download TraveLLaMA/TravelQA --local-dir ./dataIf you find this work useful, please cite our paper:
@inproceedings{chu2026travellama,
title = {TraveLLaMA: A Multimodal Travel Assistant with Large-Scale Dataset and Structured Reasoning},
author = {Chu, Meng and Chen, Yukang and Gui, Haokun and Yu, Shaozuo and Wang, Yi and Jia, Jiaya},
booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026}
}We thank the following projects for their valuable contributions:
This project is released under the Apache 2.0 License.
If you have any questions, please feel free to open an issue or contact truemanv5666@gmail.com.
