Applying PET raises accuracy for every user cohort, with the largest gain for the Low Literacy cohort¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? All user cohorts benefit from PET, with the largest improvement for the Low Literacy cohort, whose accuracy increases from 82.4% to 83.4%. → Lier Jin 2026 q2 i? Smaller improvements are observed for the Everyday User (+1.1 percentage points), Educated User (+0.7 percentage points), and Prompt Engineer (+0.5 percentage points) cohorts. → Lier Jin 2026 (2)

Evidence¶

Lier Jin 2026¶

Lier Jin, Lan Hu, Binqi Shen, Hanyu Cai, and Yuting Xin. (2026). Same Question, Different Answer? Measuring and Mitigating Prompt Privilege for Equitable AI Access. https://arxiv.org/abs/2608.08942 doi:10.48550/arxiv.2608.08942

q2 · i? · causal · r2

Post-normalization evaluation in which PET is applied as a preprocessing step before GPT-5.4-mini inference on MedQA. All user cohorts benefit from PET, with the largest improvement observed for the Low Literacy cohort, whose accuracy increases from 82.4% to 83.4%.

"All user cohorts benefit from PET, with the largest improvement observed for the Low Literacy cohort, whose accuracy increases from 82.4% to 83.4%."

Lier Jin 2026 (2)¶

Lier Jin, Lan Hu, Binqi Shen, Hanyu Cai, and Yuting Xin. (2026). Same Question, Different Answer? Measuring and Mitigating Prompt Privilege for Equitable AI Access. https://arxiv.org/abs/2608.08942

q2 · i? · causal · r2

Same post-PET evaluation, per-cohort gains in percentage points. Smaller improvements are observed for the Everyday User (+1.1 percentage points), Educated User (+0.7 percentage points), and Prompt Engineer (+0.5 percentage points) cohorts, which the authors attribute to less room for improvement.

"Smaller improvements are observed for the Everyday User (+1.1 percentage points), Educated User (+0.7 percentage points), and Prompt Engineer (+0.5 percentage points) cohorts, suggesting that users already employing well-structured prompts have comparatively less room for improvement."

Discussion¶

Related Claims¶

• After PET normalization, no statistically significant performance differences remain between user cohorts — related
• LLM accuracy increases monotonically with prompt sophistication across semantically equivalent medical questions, with Low Literacy users scoring significantly below Prompt Engineer users — related
• Training improved accuracy in every digital-forensics experience group and most generative-AI familiarity groups, with the largest estimated gains among analysts with no forensics experience and frequent generative-AI users — related