Tag
This paper introduces an experimental protocol to measure open-ended LLM conformity, showing that wrong peer input degrades revision quality and that evaluators are not neutral when shown peer context, highlighting the need for anchor calibration.
This paper reveals a confound in LLM conformity benchmarks: standard prompts mix a speaker cue with repeated wrong answers. By introducing a 'no-source' condition that removes the explicit speaker, the authors show that most apparent conformity persists even without a speaker, challenging previous interpretations of social influence.