Matt challenges the notion of fair comparisons in leaderboards, highlighting the potential bias in model tuning on dev sets. Siva acknowledges the inherent disadvantages of leaderboards and discusses the need for addressing these issues to ensure fairness in evaluations.
