Vernacular? I Barely Know Her: Challenges with Style Control and Stereotyping
This addresses style control challenges in LLMs for educational use, highlighting stereotyping issues, but it is incremental as it builds on existing research.
The study evaluated five state-of-the-art LLMs on style control tasks for educational applications, finding significant inconsistencies with reading levels up to 8th grade for first-grader tasks and performance improvements from 0.02 to 0.26, but also frequent generation of culturally insensitive content.
Large Language Models (LLMs) are increasingly being used in educational and learning applications. Research has demonstrated that controlling for style, to fit the needs of the learner, fosters increased understanding, promotes inclusion, and helps with knowledge distillation. To understand the capabilities and limitations of contemporary LLMs in style control, we evaluated five state-of-the-art models: GPT-3.5, GPT-4, GPT-4o, Llama-3, and Mistral-instruct- 7B across two style control tasks. We observed significant inconsistencies in the first task, with model performances averaging between 5th and 8th grade reading levels for tasks intended for first-graders, and standard deviations up to 27.6. For our second task, we observed a statistically significant improvement in performance from 0.02 to 0.26. However, we find that even without stereotypes in reference texts, LLMs often generated culturally insensitive content during their tasks. We provide a thorough analysis and discussion of the results.