Inductive Bias Extraction and Matching for LLM Prompts

ACL ARR 2025 May Submission5562 Authors

Abstract: All LLMs are sensitive to small changes in prompt wording, and a portion of this can be ascribed to the inductive bias that is present in the LLM. By using an LLM's output as a portion of its prompt, we can easily create satisfactory wording for prompts. This has the effect of creating a prompt that matches the inductive bias in model. Empirically, we show that using this Inductive Bias Extraction and Matching strategy improves LLM Likert ratings used for classification by up to 19% and LLM Likert ratings used for ranking by up to 27%.

Paper Type: Long

Research Area: Language Modeling

Research Area Keywords: applications,prompting

Contribution Types: Model analysis & interpretability, NLP engineering experiment

Languages Studied: English

Submission Number: 5562