Creative-writing judge discrimination
Judgemark v4
Github | Paper | | Twitter | About
💙EQ-Bench3 🌀Spiral-Bench v1.2 ✍️Longform Writing 🎨Creative Writing v3 ☢️Slop Score ⚖️Judgemark v4 🎤BuzzBench 🌍DiploBench
Judgemark v4 rates LLM judges on how discriminative they are at scoring creative writing. The score is based on separability metrics.
| Model | Judgemark Score / 95% CI | Cost | Chart |
|---|---|---|---|
| Loading... | |||
How to read this page
- Judgemark Score / 95% CI: final v4 score multiplied by 100, with the prompt-bootstrap interval shown as a whisker on the same bar.
- Chart: the generated multichart for the run, including raw means, score distribution heatmap, and overall score-use histogram.