Beyond the Prompt: Assessing Domain Knowledge Strategies for High-Dimensional LLM Optimization in Software Engineering DOI: XXXXXXX.XXXXXXX Conference: 23rd International Mining Software Repositories Conference; April 13–14, 2026; Rio de Janeiro, Brazil ISBN: 978-1-4503-XXXX-X/2018/06
Srinath Srinivasan, Tim Menzies · 2026
Read Beyond the Prompt: Assessing Domain Knowledge Strategies for High-Dimensional LLM Optimization in Software Engineering DOI: XXXXXXX.XXXXXXX Conference: 23rd International Mining Software Repositories Conference; April 13–14, 2026; Rio de Janeiro, Brazil ISBN: 978-1-4503-XXXX-X/2018/06 by Srinath Srinivasan, Tim Menzies (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-2602-02752.txt.
Description
Srinath Srinivasan and 1 co-author compare four ways of feeding domain knowledge so language models can warm-start high-dimensional software engineering tuning, posted to arXiv in 2026. Models do well on tuning problems with few features but lose to Bayesian methods when the dimension count grows. Whether structured knowledge can bridge that gap was unclear.
Four architectures are tested on data sets stratified by dimensionality. Human-in-the-loop prompting uses delayed expert feedback. Adaptive multi-stage prompting finds and validates constraints sequentially. Dimension-aware refinement tunes in progressively expanding feature subspaces. A hybrid approach mixes statistical scouting with retrieval-boosted prompting. Performance is measured by Chebyshev distance to known optima and ranked with Scott-Knott clustering against a known rival for model-generated warm starts. The study pits human-supplied knowledge against machine-generated knowledge head-on, asking which source helps a model start closer to the optimum.
Catalog record
| CREATOR | Srinath Srinivasan, Tim Menzies |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02752