Bradley Terry

Reward model in RLHF: what does it score, and on what data does it train?

DPO: what's the closed-form trick that lets it skip the reward model?

Tradeoffs between pairwise LLM-judge comparisons and scalar scoring rubrics

What is the primary scalability limitation of pairwise LLM-judge evaluation compared to scalar scoring?

What is the primary validity threat to Chatbot Arena ELO rankings as a measure of general model quality?