Back to AI News
arxiv

Study Finds Chatbot Arena Voters Grow Harsher Over Time, Even on Unchanged Models

Researchers analyzed 1,260,967 blind votes cast on Chatbot Arena between May 2023 and August 2024, restricting the analysis to pairs of models whose versions were fixed. For the same unchanged pair, the share of votes saying "both are bad" rose 1.71 percentage points per month against a mean of 17.1%, and 1.91 points per month in 2024. In fourteen pairs observed often in both years, the rate rose in every one, from 15.5% to 23.3%, while the stronger model's share of decisive votes stayed flat in 2024. The authors argue this reflects standards that track the category's progress, so absolute verdicts drift even as relative comparisons hold steady.

AI Brief

It gives concrete evidence that human preference data and LLM leaderboards drift over time, which matters for anyone interpreting Arena rankings or training on human feedback.

arXiv:2611.14333v1 Announce Type: cross Abstract: Users often complain that AI assistants such as ChatGPT are getting worse. Some versions do change, but human evaluation leaves a second possibility untested: the people judging may change. We separate the two on Chatbot Arena, a public platform where people compare answers from two anonymous large language models (LLMs) and pick the better one or say both are bad. We study 1,260,967 blind votes cast from May 2023 to August 2024 on pairs of models whose versions are fixed: dated vendor snapshots and models with fixed open weights. For the same pair of unchanged models, the share of votes saying "both are bad" rises by 1.71 percentage points a month on a mean of 17.1%, and by 1.91 points a month in 2024. In fourteen pairs observed often in both 2023 and 2024, it rose in every one, from 15.5% to 23.3%. The ranking does not trend: in 2024, the stronger model's share of decisive votes is flat. The rise holds within the same voter and on the same pair, and across arrival cohorts from their first recorded vote. It holds with prompt controls, in every major language, for dated vendor snapshots, open-weight models and models with stable answer length, and between interface changes. The drift is shared across models: month effects estimated from other models' pairs remove it from each model's own pairs. Platform-wide, the average model would be expected to score 78% against the 2023 average, yet the share of "both are bad" votes ended near where it began. The pattern fits a standard that follows the category's progress, so a product that stands still is judged as declining. Absolute verdicts on AI drifted while relative comparisons held steady, an evaluation drift that matters for LLM leaderboards and for the human feedback used to train models.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!