EQ-Bench Creative Writing Leaderboard Introduces Slop Score for Style Evaluation
sam_paech · x · 2026-08-25
Users discussed the EQ-Bench Creative Writing v3 leaderboard, designed to evaluate LLMs' emotional intelligence. The newly introduced Slop Score metric measures the frequency of clichés ("GPT-isms") by matching text against a master list of over-represented words and phrases. The leaderboard also features a Repetition metric to assess output redundancy and allows viewing detailed style profiles for models.
Related event: EQ-Bench Creative Writing Leaderboard Introduces Slop Score(2 posts)→
More from Models
- Model identified as GLM-based via tokenization analysis — peterjliu · 2026-08-25
- Arcee AI's Base Models Shine: 5.6 Pro Still Impresses — code_star · 2026-08-25
- DeepSeek R1 alleged to use Claude trajectories for performance boost — teortaxesTex · 2026-08-25
- DeepSeek Flash Vision: New Open Source King, 10x Cheaper than Kimi K3 — bindureddy · 2026-08-25
- Local AI user praises Qwen as a local coding monster, dunks on DeepSeek's UI awareness — tat_tvam_asshole · 2026-08-25
- Insider claims next-gen models will cause an 'ontological shock' — Neurogence · 2026-08-25