Benchmarks become stories before they become understanding

LLM capability claims rarely travel as raw tables. They become shorthand: a score mentioned in a launch post, a demo retold without its constraints, a leaderboard screenshot detached from its date, or a benchmark name used as proof of general intelligence. Benchmark Folklore keeps those stories visible while asking what they actually support. The page’s stance is simple: public lore matters because it changes adoption, but lore should never be confused with durable evidence.