VoirDire is an intervention-calibrated aggregation method for LLM-judge panels that varies surface attributes to identif...

VoirDire is an intervention-calibrated aggregation method for LLM-judge panels that varies surface attributes to identify nuisance-sensitive directions.Source: arXiv cs.LGhttps://arxiv.org/abs/2611.00491#MachineLearning

Read Original

Related