Few-shot chain-of-thought prompting outperforms zero-shot prompting for detecting teacher questions¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-28 (automated, 1 of 1 entry) · not reviewed by a person

Subclaims¶

q2 i? GPT-3.5-turbo five-shot CoT prompting achieved higher precision, recall, balanced accuracy and F1 than zero-shot prompting for detecting teacher questions. → Lehesvuori 2026

Evidence¶

Lehesvuori 2026¶

Lehesvuori, S., Urrutia, F., Heilala, V., Araya, R., & Hämäläinen, R. (2026). Discovering technology-aided possibilities for automatic analysis of science teacher questions. Education Tech Research Dev. https://doi.org/10.1007/s11423-025-10545-3

q2 · i? · design · r2

Seven-lesson test of GPT-3.5-turbo prompting variants (Table 2). The five-shot CoT model reached F1 0.4667 and balanced accuracy 0.8399, versus F1 0.2931 for zero-shot CoT, showing "the GPT-3.5-turbo five-shot approach was the best-performing classification model."

"the GPT-3.5-turbo five-shot approach was the best-performing classification model. Five-shot prompting approach technique involved five real question examples from the data."

Discussion¶

Related Claims¶

• In detecting teacher questions with zero-shot GPT-3.5-turbo, chain-of-thought prompting had slightly higher F1 and precision than Self-Ask but slightly lower recall and balanced accuracy — related
• GPT-4 outperforms GPT-3.5-turbo in a single-lesson question-detection test — related
• In-context examples improve LLM Bloom classification: three-shot and ten-shot prompting outperform zero-shot across datasets — related
• Few-shot LLM approaches outperform standalone supervised ML approaches, yet hand-crafted rules remain relevant — related