Verifiable Natural Language to Linear Temporal Logic Translation: A Benchmark Dataset and Evaluation Suite
William English [email protected], Chase Walker [email protected], Dominic Simon [email protected], Sumit Jha [email protected], Rickard Ewetz [email protected] · 2025
Read Verifiable Natural Language to Linear Temporal Logic Translation: A Benchmark Dataset and Evaluation Suite by William English [email protected], Chase Walker [email protected], Dominic Simon [email protected], Sumit Jha [email protected], Rickard Ewetz [email protected] (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-2507-00877.txt.
Description
William English and colleagues published a computer science paper in 2025 on translating plain language into formal logic. Engineers often state timing rules in ordinary English. Computers need those rules in a precise temporal logic instead. Existing tests made translation systems look nearly perfect.
The authors argue those tests miss a key skill. A system must also tie each basic claim to a real scenario. They built a benchmark with four different state spaces. It holds thousands of paired plain and formal statements. Sample traces let users check whether formulas actually hold. Many current systems split the job into lifting and grounding steps. The benchmark scores the whole path end to end.
Catalog record
| CREATOR | William English [email protected], Chase Walker [email protected], Dominic Simon [email protected], Sumit Jha [email protected], Rickard Ewetz [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-00877