Arena Introduces Factuality Leaderboard
WeijiaShi2 · x · 2026-07-16
Arena has incorporated factuality into its model rankings, with the new leaderboard displaying a weighted result of "human preference + factuality".
The team explained that they randomly sample battle conversations, extract verifiable web-based claims, and fact-check them individually. To do this, they have already annotated over 2 million LLM claims, with 1.3 million+ from Text Arena and 700,000 from Search Arena.
Results mentioned in the quoted tweet: GPT-5.6 holds a significant lead in factuality, followed closely by Grok-4.5; Opus-4.8 scored higher than Fable-5, while Muse-Spark lost notable points as the factuality weight increased.
Related event: Arena Adds Factuality to Model Rankings(12 posts)→
More from Models
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11
- User Hails Unconfirmed 'DeepSeek 4.1 Flash' as an Inflection Point in LLMs — himanshustwts · 2026-09-11
- Terminal Bench v4: GLM-5.3 Leads at 41.9%, Kimi-K3 Underwhelms at 12.6% — Ok_Warning2146 · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11