Enhancing RLHF in LLMs: Comparing BERT, XGBoost, and Deep Models for Bias Detection

Summary

A comparison of BERT, XGBoost, and deep model families for bias detection in the context of reinforcement learning from human feedback, examining how reward-model choice affects bias signals in LLM alignment pipelines.

Topics