LLM Conformity Protocol Measures Evaluator Bias and Peer Influence
August 6, 2026
A new experimental protocol separates ordinary re-answering from peer-driven conformity in open-ended LLM revisions. Testing across four open-weight models shows that incorrect peer input reduces revision quality and that both human and GPT-4o evaluators exhibit directional sensitivity to visible peer context.
HOW THIS AFFECTS YOU
●
researcherYou can use this decomposition method to isolate whether model errors stem from training data or evaluator bias.
●
policyThis highlights the need for more robust evaluation frameworks to mitigate social bias in model alignment.