Prompt perturbation and fraction facilitation sometimes strengthen large language model scores

Large Language Models (LLMs) can be tasked with scoring texts according to pre-defined criteria, but there is no recognised optimal prompting strategy yet. This article focuses on scoring articles for research quality on a four-point scale, testing how user prompt design can enhance this ability. Based primarily on 1.7 million Gemma3 27b queries for 2780 health and life science journal article titles and abstracts with 58 similar prompts, the results show that improvements can be obtained by (a) testing semantically equivalent prompt variations, (b) averaging scores from semantically equivalent prompts, (c) specifying that fractional scores are allowed, and possibly also (d) not drawing attention to the input being partial. Whilst (a) and (d) suggest that models can be sensitive to how a task is phrased, (b) and (c) suggest that strategies to leverage more of the model's knowledge are helpful, including by perturbing prompts and facilitating fractions. Perhaps counterintuitively, encouraging incorrect answers (fractions for this task) usually releases useful information about the model's certainty about its answers. Mixing semantically equivalent prompts also reduces the chance of getting no score for an input. Additional testing showed that the best prompts vary between LLMs: almost the opposite for ChatGPT-4.1 mini, weakly aligned for Llama4 Scout and Magistral, and little difference for Qwen3 32b and DeepSeekR1 32b. Overall, whilst there is no single best prompt, a good strategy for all models was to average the scores from a range of different semantically equivalent or similar prompts, and eliciting fractions improved most models' results.