Prompt repetition boosts LLM accuracy

Research Update: 8 papers on AI in Education you need to know for 2026 In this episode, Ray and Dan provide a rapid-fire rundown of the most significant research papers hitting the AI in Education space so far in 2026. After a series of news-heavy episodes, the hosts catch up on the data behind synthetic avatars, grading accuracy, and the psychological biases we hold against AI. Key highlights include: Synthetic Lecturers: Exploring stakeholder perspectives on digital twins and the emotional reaction to the term Deepfake in academia. The Grading Gap: Why ChatGPT tends to be more sycophantic and generous with weak work compared to human instructors. The Disclosure Penalty: New findings from 16 experiments showing why humans devalue creative writing the moment they know AI is involved. Prompting Hacks: The "Groundhog Day" method 😂 Why simply repeating your prompt twice can boost accuracy across 70 different AI systems. Tools for Researchers: Insights into Jasper Roe's research checklist and the "Paper Banana" tool for automating scientific diagrams. Links to all the research papers discussed

• Can synthetic avatars replace lecturers? An exploratory international study of higher education stakeholder perceptions| https://link.springer.com/article/10.1186/s41239-025-00568-4
• Who grades best? Comparing ChatGPT, peer, and instructor evaluations across varying levels of student project quality https://www.tandfonline.com/doi/full/10.1080/02602938.2025.2588682?src=
• The Artificial Intelligence Disclosure Penalty: Humans Persistently Devalue AI-Generated Creative Writing https://doi.org/10.1037/xge0001889
• The older "Transparency Dilemma" paper referenced too: https://www.sciencedirect.com/science/article/pii/S0749597825000172
• Asking generative artificial intelligence the right questions improves writing performance https://www.sciencedirect.com/science/article/pii/S2666920X25000141?via%3Dihub
• When AI only asks: how question-driven dialogue shapes prewriting in the classroom https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2026.1740044/full
• Prompt Repetition Improves Non-Reasoning LLMs https://arxiv.org/html/2512.14982v1
• How to Use Generative AI in Educational Research https://www.cambridge.org/core/elements/how-to-use-generative-ai-in-educational-research/916142E735B678F86A59240BFE651F5C
• PaperBanana: Automating Academic Illustration for AI Scientists https://dwzhu-pku.github.io/PaperBanana/ https://arxiv.org/abs/2601.23265