Pairwise or Pointwise? Evaluating Feedback Protocols for Bias in LLM-Based Evaluation

Open in new window