Trust score (Kojen)
Kojen's trust score is a 0–100 roll-up of an evaluation; for comprehensive evaluations it weights accuracy 50%, hallucination 30% and bias 20%.
It is meant as a ship gate next to p95 latency, not a replacement for the components: model cards recommend reporting performance broken down rather than as one number, so Kojen shows each part, with confidence intervals, under the score.
Set minimums per component first (for example accuracy at least 85% and no critical safety failures), then a threshold on the overall score.
Learn more on Kojen
Related terms
- HallucinationA hallucination is model output that is fluent and confident but unsupported by its source material or false about the world..
- LLM-as-a-judgeLLM-as-a-judge is the practice of using a language model to grade another model's outputs against a rubric or a reference answer, instead of a human grader..
- Confidence intervalA confidence interval is a range computed from a sample that, across repeated samples, would contain the true value a stated share of the time, such as 95%.
Sources
- Mitchell et al., 2019. Model Cards for Model Reporting (arXiv:1810.03993)
Put the definitions to work.
Kojen evaluates, tunes and red-teams your agents with these methods built in: Wilson intervals, pass^k, LLM judges with visible components. Free to start.
Start training free