@inproceedings{sandfuchs2025llms,
  author    = {Sandfuchs, Stephan and Melchert, Maximilian and Frochte, J\"{o}rg},
  title     = {{LLMs for Text-Based Exploration and Navigation under Partial Observability}},
  booktitle = {International Conference on Advancements in Automation, Robotics and Sensing (ICAARS 2025)},
  series    = {Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering (LNICST)},
  publisher = {Springer},
  year      = {2025},
  note      = {Accepted, to be published in 2026},
  doi       = {10.48550/arXiv.2604.09604},
  url       = {https://arxiv.org/abs/2604.09604},
  abstract  = {Exploration and goal-directed navigation in unknown layouts are central to inspection, logistics, and search-and-rescue. We ask whether large language models (LLMs) can function as text-only controllers under partial observability -- without code execution, tools, or program synthesis. We introduce a reproducible benchmark with oracle localisation in fixed ASCII gridworlds: each step reveals only a local 5x5 window around the agent and the model must select one of UP/RIGHT/DOWN/LEFT. Nine contemporary LLMs ranging from open/proprietary, dense / Mixture of Experts and instruction- vs. reasoning-tuned are evaluated on two tasks across three layouts of increasing difficulty: Exploration (maximising revealed cells) and Navigation (reach the goal on the shortest path). Reasoning-tuned models reliably complete navigation across all layouts, yet remain less efficient than oracle paths. Few-shot demonstrations in the prompt chiefly help these reasoning-tuned models by reducing invalid moves and shortening paths, while classic dense instruction models remain inconsistent. We observe characteristic action priors (UP/RIGHT) that can induce looping under partial observability. Overall, training regimen and test-time deliberation predict control ability better than raw parameter count. These findings suggest lightweight hybridisation with classical online planners as a practical route to deployable partial map systems.},
  keywords  = {robotics, llm}
}