Scoring, Reasoning, and Selecting the Best! Ensembling Large Language Models via a Peer-Review Process

A BERTology View of LLM Orchestrations: Token- and Layer-Selective Probes for Efficient Single-Pass Classification