arXiv:2610.19095 · v3
Similar Models, Broken Odds: Where Luce's Axiom Fails in Chatbot Arena
摘要
Chatbot Arena gives each language model one Bradley-Terry score, which fixes how often any model should beat any other. Choice theory predicts where this fails: near-copies are compared on the few details in which they differ, so votes between them are more decisive than their scores imply, exactly for the comparison a lab cares about, a new version against the old one. On 1.8 million Arena votes, votes between versions of one model are 1.37 times as decisive as Bradley-Terry predicts, between sizes of one generation 1.26 times and between models of one lab 1.12 times; a release a year later points the same way, and the effect survives tie, prompt, topic and timing checks. How alike two models' answers are worded and formatted tracks the effect. Plain scores leave the odds between relatives about 9% too close to even on held-out votes; a correction with four to six parameters roughly halves this and predicts held-out votes better, mostly by modeling ties, while the leaderboard order barely moves; frozen and carried to new releases, its gains in head-to-head odds are too small for our test to establish. With language-model judges and three-answer menus, a near-copy takes votes mostly from its twin when choices are pooled over prompts (log-odds shift +0.10, 95% interval 0.05 to 0.16), less so within a prompt.
作者备注:20 pages, 5 figures, 10 tables; v3: adds a test on human summarization preferences and a calculation of votes needed for a release decision; clearer abstract, introduction and transfer results; appendix regrouped