The Essay-Marking Problem
bycloud's English teacher analogy makes benchmark scores sound self-evident, then admits the judgment behind them is not. AI Explained reaches the same pressure point from another direction: intelligence is hard to measure, yet the LMSYS Chatbot Arena leaderboard remains a public signal. LLM-as-a-Judge research formalizes one alternative, using models to assess outputs, but it does not erase the original question of whose judgment counts.






