SafePlan: Leveraging Formal Logic and Chain-of-Thought Reasoning for Enhanced Safety in LLM-based Robotic Task Planning
Ike Obi, Vishnunandan L.N. Venkatesh, Weizheng Wang, Ruiqi Wang, Dayoon Suh, Temitope I. Amosa, Wonse Jo, Byung-Cheol Min · 2025
Read SafePlan: Leveraging Formal Logic and Chain-of-Thought Reasoning for Enhanced Safety in LLM-based Robotic Task Planning by Ike Obi, Vishnunandan L.N. Venkatesh, Weizheng Wang, Ruiqi Wang, Dayoon Suh, Temitope I. Amosa, Wonse Jo, Byung-Cheol Min (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-06892.txt.
Description
Ike Obi, Vishnunandan Venkatesh, and six colleagues wrap language-model robot planners in a logic checker so unsafe commands get caught before a robot moves, in a paper published in two thousand twenty-five. Language models now take task commands, draft plans, form robot teams, and hand out assignments, but a malicious or careless prompt can turn that fluency dangerous.
Their framework, SafePlan, combines formal logic with chain-of-thought reasoning. Model outputs, from plans to team allocations, are examined against logical safety constraints. Bad outputs are refined or rejected before execution. The framework has several components that share the checking work. The point is defense in depth: fluency proposes, logic disposes. As robots take instructions in open language, the attack surface grows with the vocabulary. SafePlan treats verification as part of the planning loop rather than an external audit.
Catalog record
| CREATOR | Ike Obi, Vishnunandan L.N. Venkatesh, Weizheng Wang, Ruiqi Wang, Dayoon Suh, Temitope I. Amosa, Wonse Jo, Byung-Cheol Min |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-06892