passdrill
AI & LLM Engineering · Evaluation, Testing & Safety · Card 007/010 medium

A team wants to align a pretrained language model's outputs with human preferences -- for example, favoring helpful, well-formatted answers over unhelpful or poorly formatted ones -- in a way that goes beyond what the original next-token-prediction pretraining objective optimizes for. They collect many pairs of candidate outputs for the same prompt and have human annotators indicate which output in each pair they prefer. How does RLHF (reinforcement learning from human feedback) typically use this preference data to align the model?

  1. It directly edits the pretrained model's weight matrices using the preference labels as targets in an ordinary supervised next-token-prediction loss, with no reward model or reinforcement learning step involved at all
  2. It uses the preference pairs only to filter the pretraining corpus, removing any documents similar in style to the less-preferred outputs, before retraining the model from scratch on the filtered corpus
  3. It first trains a separate reward model to predict which of two outputs a human would prefer, using the collected pairwise comparisons as training data, and then uses reinforcement learning (commonly PPO) to fine-tune the language model's policy to maximize the score that reward model assigns to its outputs
  4. It uses the preference pairs to build a retrieval index, and at inference time retrieves the most similar previously-preferred output to return as the answer instead of generating a new one
Next card → Shuffle