When models disagree, it’s a red flag that an input might be risky or unusual....

https://raymondsinterestingchat.quantlynix.com/posts/when-should-a-model-abstain-instead-of-guessing-understanding-selective-prediction-and-escalation-strategies

When models disagree, it’s a red flag that an input might be risky or unusual. Measuring ensemble variance or entropy helps spot these tricky cases