Iterative prompting to revise overly complex responses raised readability alignment, with most regenerated responses falling at or below target grade levels¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? A follow-up revision prompt improved most regenerated responses to at or below target grade levels, ranging from 55.10% (GPT-5-mini) to 95.92% (Gemini-2.5-Flash). → Nguyen 2026

Evidence¶

Nguyen 2026¶

Nguyen, H., & Cao, J. (2026). Exploring the Capacity of Large Language Models to Simulate Students' Scientific Thinking: Insights for Responsive Teaching. Journal of Science Education and Technology. https://doi.org/10.1007/s10956-026-10333-5

q2 · i? · design · r2

Multi-prompt pipeline evaluation: responses exceeding readability thresholds were re-prompted to "Revise this response to match the language expected for {grade_level}." Readability alignment reached 100% for high school grades but varied widely for elementary grades (6.67%–80%).

"Overall, most of the regenerated responses improved and fell at or below the target grade levels, Gemini-2.5-Flash (95.92%), Llama-4 (91.84%), Mistral-7B (91.84%), GPT-4o (89.80%), Claude-Sonnet-4 (83.67%), and GPT-5-mini (55.10%)."

Discussion¶

Related Claims¶

• 91% of LLM-simulated ideas fell within grade-level knowledge scope, but within-scope proportions were significantly lower for elementary and middle school than high school — related
• Only 69% of LLM-simulated ideas met the target grade's language level, with elementary grades faring worst and Llama-4 performing best — related
• Revising texts to improve readability-formula scores often produces nonexistent or insignificant comprehension gains — related