DPO Fine-Tuning on Anthropic HH-RLHF with TRL and LoRA

A tutorial builds an end-to-end preference-learning workflow using the Anthropic HH-RLHF dataset and Direct Preference Optimization. It audits dataset biases, runs lexical shortcut diagnostics, and constructs a version-robust DPO training pipeline with TRL and LoRA in Colab.