A Solver-in-the-Loop Framework for Improving LLMs on Answer Set Programming for Logic Puzzle Solving
Timo Pierre Schrader, Lukas Lange, Tobias Kaminski, Simon Razniewski, Annemarie Friedrich · 2025
Read A Solver-in-the-Loop Framework for Improving LLMs on Answer Set Programming for Logic Puzzle Solving by Timo Pierre Schrader, Lukas Lange, Tobias Kaminski, Simon Razniewski, Annemarie Friedrich (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-2512-17093.txt.
Description
Timo Pierre Schrader and coauthors teach language models to write Answer Set Programming code for logic puzzles, with a solver in the loop. ASP is strong for combinatorial search, but models saw few ASP examples in training, so their code often misses the mark.
Their method samples program continuations from the model for puzzle statements given in plain words. Since partial ASP code narrows the solution space step by step, solver feedback sorts samples into chosen and rejected ones. Fine-tuning on that curated split, plus solver-guided best-of-N search, lifted results in two prompting setups on two data sets, a recipe for other niche languages.
Catalog record
| CREATOR | Timo Pierre Schrader, Lukas Lange, Tobias Kaminski, Simon Razniewski, Annemarie Friedrich |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-17093