# Co-Located Tests, Better AI Code: How Test Syntax Structure Affects Foundation Model Code Generation

arXiv: https://arxiv.org/abs/2604.19826 · HTML: https://mlsanity.com/paper/2604.19826/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 50 references that are in the map.
- First posted on arXiv 20 Apr 2026.

## Builds on (most cited first, 23 of 23)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code](https://mlsanity.com/paper/2403.07974/) (2024)
- [Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation](https://mlsanity.com/paper/2305.01210/) (2023)
- [RWKV: Reinventing RNNs for the Transformer Era](https://mlsanity.com/paper/2305.13048/) (2023)
- [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://mlsanity.com/paper/2304.14997/) (2023)
- [StarCoder 2 and The Stack v2: The Next Generation](https://mlsanity.com/paper/2402.19173/) (2024)
- [Asleep at the Keyboard? Assessing the Security of GitHub Copilot's Code Contributions](https://mlsanity.com/paper/2108.09293/) (2021)
- [CodeT: Code Generation with Generated Tests](https://mlsanity.com/paper/2207.10397/) (2022)
- [The Impact of AI on Developer Productivity: Evidence from GitHub Copilot](https://mlsanity.com/paper/2302.06590/) (2023)
- [Analyzing the Structure of Attention in a Transformer Language Model](https://mlsanity.com/paper/1906.04284/) (2019)
- [An Empirical Evaluation of Using Large Language Models for Automated Unit Test Generation](https://mlsanity.com/paper/2302.06527/) (2023)
- [Eagle and Finch: RWKV with Matrix-Valued States and Dynamic Recurrence](https://mlsanity.com/paper/2404.05892/) (2024)
- [DocPrompting: Generating Code by Retrieving the Docs](https://mlsanity.com/paper/2207.05987/) (2022)
- [CodeGemma: Open Code Models Based on Gemma](https://mlsanity.com/paper/2406.11409/) (2024)
- [An Empirical Study of the Non-determinism of ChatGPT in Code Generation](https://mlsanity.com/paper/2308.02828/) (2023)
- [MultiPL-E: A Scalable and Extensible Approach to Benchmarking Neural Code Generation](https://mlsanity.com/paper/2208.08227/) (2022)
- [No More Manual Tests? Evaluating and Improving ChatGPT for Unit Test Generation](https://mlsanity.com/paper/2305.04207/) (2023)
- [Bugs in Large Language Models Generated Code: An Empirical Study](https://mlsanity.com/paper/2403.08937/) (2024)
- [What's Wrong with Your Code Generated by Large Language Models? An Extensive Study](https://mlsanity.com/paper/2407.06153/) (2024)
- [What Do They Capture? -- A Structural Analysis of Pre-Trained Language Models for Source Code](https://mlsanity.com/paper/2202.06840/) (2022)
- [What do pre-trained code models know about code?](https://mlsanity.com/paper/2108.11308/) (2021)

## Abstract (arXiv)

AI coding assistants increasingly generate code alongside tests. How developers structure test code, whether inline with the implementation or in separate blocks, has traditionally been a matter of testing philosophy. We investigate whether this choice affects AI code generation quality. We conduct a large-scale empirical study (830+ generated files, 12 models, 3 providers) using SEGA, a three-dimensional evaluation framework measuring Determinism, Preservation, and Correctness. Comparing inline test syntax (Python doctests) against separated test syntax (Rust #[test] blocks) on a d-ary heap implementation, we find that: (1) inline tests yield near-perfect preservation (100%) and correctness (92-100%) across all models; (2) separated tests expose stark model-tier gaps (0-100% correctness) and independence between preservation and correctness; (3) model behavior evolves across generations, and notably one model breaks the test suppression pattern of its three predecessors; (4) mechanistic analysis on 7 open-source architectures (6 transformers and a gated-linear Recurrent Neural Network (RNN)) reveals inline test markers receive 2.8-4.4× stronger attention in 5/7 models, with causal validation via knockout and steering experiments on the 4 code-specialized transformers and RWKV-6; the co-location mechanism extends to a non-transformer architecture, suggesting the design recommendation is robust to future architectural shifts. In the Foundation Model era, test syntax structure is a software design concern: co-locating tests with implementation code produces measurably better AI-generated code. This arxiv long version includes appendices that further qualify the effect as bounded by both model capability and programming language.