WritingBench

WritingBench evaluates open-ended writing tasks: speeches, marketing copy, academic drafts, creative text. Rubrics measure whether the piece actually fulfils its brief — a step beyond 'sounds good' towards measurable writing quality.

WritingBench measures how precisely a model obeys the instructions it was given. In detail, it checks:

• fulfilling the specific writing brief (genre, audience, goal)
• verifiable constraints inside the text
• quality judged by rubric criteria
• stylistic breadth across very different text formats

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

What is the WritingBench benchmark?

WritingBench evaluates open-ended writing tasks: speeches, marketing copy, academic drafts, creative text. Rubrics measure whether the piece actually fulfils its brief — a step beyond 'sounds good' towards measurable writing quality.

What does WritingBench test?

WritingBench mainly measures how precisely a model obeys the instructions it was given. Concretely, it checks: fulfilling the specific writing brief (genre, audience, goal); verifiable constraints inside the text; quality judged by rubric criteria.

How is the WritingBench score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get WritingBench data and scores?

The official source is https://arxiv.org/abs/2503.05244. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/writingbench/scores) or pull it directly from the Open LLM Radar API with any HTTP client.