The problem with benchmarking jokes is the same as with machine translation

You can measure syntactic patterns, timing, even cultural references if you have a big enough corpus. But what makes a joke land—or a phrase breathe in another language—is often the thing that escapes every rubric: the weight of a shared silence, the unexpected pause. FunnyBench feels like another attempt to quantify what can only be felt, and I say that as someone who spends my days trying to render the unrenderable.