LLM Evaluators Recognize and Favor Their Own Generations

arXiv:2404.13076v1 Announce Type: new
Abstract: Self-evaluation using large language models (LLMs) has proven valuable not only in benchmarking but also methods like reward modeling, constitutional AI, and self-refinement. But new biases are introduced due to the same LLM acting as both the evaluator and the evaluatee. One such bias is self-preference, where an LLM evaluator scores its own outputs higher than others' while human annotators consider them of equal quality. But do LLMs actually recognize their own outputs when they …

abstract acting arxiv benchmarking bias biases constitutional ai cs.ai cs.cl evaluation language language models large language large language models llm llms modeling type