SymCode: A Neurosymbolic Approach to Mathematical Reasoning via Verifiable Code Generation
Sina Bagheri Nezhad, Yao Li, Ameeta Agrawal, Portland State University, Portland, USA · 2025
Read SymCode: A Neurosymbolic Approach to Mathematical Reasoning via Verifiable Code Generation by Sina Bagheri Nezhad, Yao Li, Ameeta Agrawal, Portland State University, Portland, USA (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-2510-25975.txt.
Description
Sina Bagheri Nezhad, Yao Li, and Ameeta Agrawal present a framework that makes language models solve math by writing checkable code instead of prose. Prose reasoning, even with step by step prompts, can hide arithmetic slips and unchecked logic. Their system, SymCode, reframes each problem as code generation for the SymPy library, a Python tool for symbolic mathematics, so a deterministic engine carries out the steps.
Bagheri Nezhad and the team test SymCode on hard benchmarks that include five hundred curated problems and olympiad style questions. Accuracy rose by up to thirteen point six percentage points over baselines, with fewer tokens used. Failures also changed character, from hidden logical errors to visible program errors that are easier to debug.
Catalog record
| CREATOR | Sina Bagheri Nezhad, Yao Li, Ameeta Agrawal, Portland State University, Portland, USA |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-25975