LMSYS Chatbot Arena
lmarena.ai
A live human preference leaderboard where anonymous models compete head-to-head in blind evaluations, producing the most credible ranking of conversational AI quality available.
Why it is useful
Elo rankings based on millions of human blind comparisons are harder to game than benchmark suites, which makes this the most trustworthy signal for which models people actually prefer in practice. Useful both for choosing models and for understanding the gap between self-reported benchmark performance and real-world user preference.
More in Research & Data
Other research & data you might not have heard of.
- World Bank Open DataFree access to global development data covering hundreds of indicators across economics, health, education, and environment for every country over decades.
- AlternativeToA crowdsourced directory that finds free and open-source alternatives to any software, app, or web service — covering desktop, mobile, and web platforms with community ratings and filtering by platform and licence.
- Academic TorrentsDistributes large research datasets — from the Enron email corpus to machine learning benchmarks — via BitTorrent instead of a single slow server.
- KaggleCommunity platform for datasets, machine learning competitions, and collaborative notebooks.
- Hugging Face PapersHugging Face daily ML papers feed linking research to open-source implementations, demos, and community discussion.
- Harvard DataverseOpen research data repository from Harvard's Institute for Quantitative Social Science, hosting 75,000+ citable datasets across every discipline.