Chatbot Arena ELO rating on difficult, complex prompts — filters for the hardest user queries to differentiate top-tier models.
Filters for difficult prompts where frontier models are less likely to tie; useful for separating top-tier systems that cluster on easier categories.
Hardness labels are heuristic rather than rigorously calibrated by solved-rate statistics; niche prompt style can drive outcomes independently of true reasoning difficulty. Data: Arena (lmarena.ai) leaderboard dataset, https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset, licensed CC-BY-4.0; rows are converted to the ELO fields shown here.
Higher is better. ELO-like rating for hard prompts.
Hardness labels are heuristic rather than rigorously calibrated by solved-rate statistics.
No scores available for this benchmark yet.