Prompt Architecture Determines Reasoning Quality: A Variable Isolation Study on the Car Wash Problem
Heejin Jo · 2026
Read Prompt Architecture Determines Reasoning Quality: A Variable Isolation Study on the Car Wash Problem by Heejin Jo (2026) 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-2602-21814.txt.
Description
Large language models always fail the car wash problem, a viral thinking benchmark needing hidden constraint guesses. Heejin Jo presents a variable split study, 20 trials per setup across 6 setups, 120 trials total, testing which prompt design layers in a live system enable correct answers. Using Claude Sonnet with fixed settings, she finds the STAR method alone raises scores from 0 percent to 85 percent, a strong result.
Adding user profile context via vector store lookup adds a further 10 points, while RAG context adds 5 more, reaching 100 percent correct in the full-stack setup. The results suggest ordered thinking scaffolds, notably forced goal notes before answers, matter far more than context adds for hidden constraint tasks.
Catalog record
| CREATOR | Heejin Jo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-21814