HIVE Supplementary Material + MuSE Benchmark

Associated with our ICLR'25 publication entitled "From an LLM swarm to a PDDL-empowered HIVE: planning self-executed instructions in a multi-modal jungle", we prepare an archive containing additional materials which may be useful.

File Hierarchy

This archive contains:

• README.md this file;
• MuSE_Benchmark/ which provides the raw data for the MuSE benchmark we designed, composed of
  • README.md
  • queries.json
  • data/audios/
  • data/images/
• C-KG_excerpt/ to visualise in a Web-browser the sub-graph of our Capability-KG corresponding to the experiments we present in the article so to tackle the MuSE benchmark.
  • src/ containing various JavaScript data and library files
  • capability_kg.html the GUI to be opened
• PDDL-domain-files/ corresponding to the 10 domains involved in the MuSE Benchmark
• NLP_Taxonomy/ which contains our comprehensive taxonomy of model actions as json and png
• Screencast.mp4 which presents a complete run-through of a query for all the systems we tested in the submission, i.e., HuggingGPT, ControlLLM and Hive
• Report.pdf the associated report displayed in Screencast.mp4

Citation

@misc{vyas2025hive,
     title={{From An LLM Swarm To A PDDL-Empowered HIVE: Planning
      Self-Executed Instructions In A Multi-Modal Jungle}},
     author={Kaustubh Vyas and Damien Graux and Yijun Yang and
      Sébastien Montella and Chenxin Diao and Wendi Zhou and Pavlos
      Vougiouklis and Ruofei Lai and Yang Ren and Keshuang Li and Jeff
      Z. Pan},
     year={2025},
     booktitle={The Thirteenth International Conference on Learning
      Representations, {ICLR} 2025, Singapore, April 24-28, 2025},
     publisher={OpenReview.net},
     url={https://arxiv.org/abs/2412.12839},
}