lolbench: Benchmarking Whether LLMs Truly Understand Humor

A developer built lolbench, a humor benchmark with 87 obscure jokes and tasks like explaining jokes and ranking by preference. Tests across 7 models with 885 comparisons show score gaps between explaining successful and failed jokes stem from reasoning, not memorization.

2026-09-28 ~ 2026-09-28 · 2 related posts

• Humor benchmark lolbench survives kill test: 87 obscure jokes show the gap isn't memory — AffectionateGas9544 · 2026-09-28
• Humor benchmark lolbench survives kill test: 87 obscure jokes show the gap isn't memory — AffectionateGas9544 · 2026-09-28