Kleinberg and Raghavan proved algorithmic monoculture can hurt social welfare
The 2021 theorem's worked case gives about a 4% welfare loss under three candidates and Gaussian noise. In the same paper's Plackett-Luce case the effect vanishes.
At the top of the RewardBench leaderboard, higher scores predicted lower Chatbot Arena rank
AI2 built the benchmark in March 2024 to score a reward model on 2,985 fixed chosen-versus-rejected pairs. Berkeley's follow-up training work that year found the top of the leaderboard sorting the wrong way against Chatbot Arena rank.
The wait before a chatbot's first word barely tracks the prompt's length
Anyscale measured a Llama 2 70b prompt token at about a hundredth of what an output token costs in end-to-end latency, well below the tens-of-thousands regime where quadratic attention drives the pause.
NYC's MyCity chatbot told ten Markup staffers landlords can refuse Section 8 vouchers
Adams put the chatbot on chat.nyc.gov in October 2023 pointed at 2,000 NYC Business pages. Mamdani took it down 27 months later, after wrong answers on housing, tips, and cash payments.
Today's RAG pipelines fine-tune neither piece the paper trained together
Lewis et al. froze DPR's document encoder and fine-tuned its query encoder alongside BART. The RAG that ships today wires two frozen off-the-shelf pieces.