5 ms·
Unfortunately not, as it captures such a wide spectrum of use cases and scenarios. There are some benchmarks to measure this quality in specific settings, e.g.
by amilios 2y ago
Unfortunately not, as it captures such a wide spectrum of use cases and scenarios. There are some benchmarks to measure this quality in specific settings, e.g. summarization, but AFAIK nothing general.
- rkwz 2y agoThanks, any ideas why it's not possible to build a generic eval for this? Since it's about asking a set of questions that's not public knowledge (or making stuff up) and check if the model says "I don't know"?