# TabBench-Bio paper snapshot

43 original datasets and 19 predictive configurations, plus DUMMY and the separate AUTOGLUON reference. Results are drawn from the latest database restricted to the original roster. Strict is the primary analysis; adaptive and failures-omitted views are sensitivities. Rankings use 2,000 target-bootstrap draws.

Read data/dashboard.json, data/leaderboard.json and data/snapshot-manifest.json.
