LiveBench AI (2024年)（ライブベンチエーアイ）

概要

LiveBench AI は2024年6月Abacus.AI + NYU発表・Industry-leading contamination-free monthly-refreshed 2024年LiveBench AI dynamic benchmark position確立。Industry-leading 6 categories reasoning/math/coding/data analysis/instruction following/language + Yann LeCun co-author + objective ground-truth scoring + livebench.ai。

主な特徴・仕組み

• Industry-leading Abacus.AI (Bindu Reddy CEO) + NYU (Yann LeCun) + collaborators 2024年6月 release
• Industry-leading LiveBench AI dynamic benchmark signature
• Industry-leading contamination-free monthly-refreshed signature
• Industry-leading 6 categories: reasoning + math + coding + data analysis + instruction following + language signature
• Industry-leading objective ground-truth scoring signature (no LLM-as-judge)
• Industry-leading Yann LeCun co-author signature (Meta Chief AI Scientist)
• Industry-leading new questions monthly signature (avoids contamination)
• Industry-leading public leaderboard livebench.ai signature
• Industry-leading 2024年6月 ArXiv paper "LiveBench A Challenging, Contamination-Free LLM Benchmark"
• Industry-leading o1 + Claude 3.5 Sonnet + GPT-4o frequent top-3

まとめ

LiveBench AI = 2024年6月Abacus.AI + NYU発表のpremium dynamic benchmark。Industry-leading contamination-free monthly-refreshed + 6 categories reasoning + math + coding + data analysis + instruction following + language + objective ground-truth scoring no LLM-as-judge + Yann LeCun co-author Meta Chief AI Scientist + new questions monthly avoids contamination + public leaderboard livebench.ai + Abacus.AI Bindu Reddy CEO + NYU collaboration position確立。