Towards a Deterministic Math Solver for Clinical Language Models
Quick summary
arXiv:2609.10728v1 Announce Type: new Abstract: Large language models are unreliable at arithmetic, which is a problem for clinical calculators where a single numerical error changes the recommendation. The standard response is to hardcode each calculator as a validated function, one at a time. We test an alternative: the model does not calculate. Instead, it writes case-specific Python that a restricted local executor runs as a deterministic solver, and the model's task reduces to deciding how to use it. We evaluate this Program-Solve interface on MedCalc-Bench Verified (1,100 cases, 55 calcu
Key takeaways
- arXiv:2609.10728v1 Announce Type: new Abstract: Large language models are unreliable at arithmetic, which is a problem for clinical calculators where a single numerical error changes the recommendation.
- The standard response is to hardcode each calculator as a validated function, one at a time.
- We test an alternative: the model does not calculate.
Why it matters
“Towards a Deterministic Math Solver for Clinical Language Models” should be evaluated beyond branding and benchmark scores. Its practical importance will emerge in task accuracy, latency, unit cost, safety and integration with real workflows.
Member comments