From Brittle to Robust: Improving LLM Annotations for SE Optimization Journal: Empirical Software Engineering

Lohith Senthilkumar, Tim Menzies · 2026

Read From Brittle to Robust: Improving LLM Annotations for SE Optimization Journal: Empirical Software Engineering by Lohith Senthilkumar, 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-2603-22474.txt.

Description

Senthilkumar and Menzies ask large language models to label data for software engineering optimization. Such labels are slow and costly to get from scarce human experts. The authors find model labels fail on high-dimensional, multi-objective tasks when used one at a time. Their fix, SynthCore, polls several isolated model sessions and merges their views into an ensemble of few-shot learners. On 49 tasks the ensemble beat leading search methods. All labels came from models, with no human input.

Catalog record

CREATORLohith Senthilkumar, Tim Menzies
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64583
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.22474

Browse the full library · About the Tower of Babel