Auto-Rubric Evaluator
Describe any task and the system auto-builds a weighted rubric, runs two models, and grades each response against every criterion — visualized as an animated head-to-head score matrix.
How it works
- Rubric: a grader model designs 3 weighted criteria for your task.
- Generate: two models (gpt-5.4, grok-4.3) each attempt the task.
- Grade: each response is scored 1–5 per criterion; weighted totals decide the winner.
- Scores render as animated A-vs-B bars per criterion.
Build rubric
→
2 modelsanswer
→
Gradeper criterion
→
Rank