AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

A Theory of Dynamic Benchmarks

Ali Shirali, Rediet Abebe, Moritz Hardt

ICLR 2023 · Probabilistic and scientific modeling › Molecules and proteins · approach: Prediction-powered and conformal inference · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Continue sequential path benchmarking beyond three roundsPath dynamic benchmark · Realizable classificationmajority-vote risk: \frac{\epsilon^2}{8}No clear effect
Run additional rounds of path benchmarkingCIFAR-10 path dynamic benchmark · CIFAR-10majority-vote riskNo clear effect
Run additional rounds of path benchmarkingSNLI path dynamic benchmark · SNLImajority-vote riskNo clear effect
Have rollout error overlaps resemble the constructed negative example more closelyPath dynamic benchmark rollouts · CIFAR-10final risk correlation with z_4: Pearson r=0.46, p < 0.001Mixed
Have rollout error overlaps resemble the constructed negative example more closelyPath dynamic benchmark rollouts · SNLIfinal risk correlation with z_6: Pearson r=0.68, p < 0.001Mixed
Use exponential-loss dynamic gradient-based updatesDynamic gradient-based updating · Classificationrisk upper bound: R_\gD(h_t) \le \exp(-\frac{(1 - 2\epsilon)^2 t}{2})Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy