AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Epistemic Uncertainty Quantification To Improve Decisions From Black-Box Models

Sébastien Melo, Gaël Varoquaux, Marine Le Morvan

ICLR 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Epistemic versus aleatoric decomposition · 0 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
use binning-free honest-tree grouping loss estimator instead of binned estimatorGrouping loss estimation · Semi-synthetic Weather (TabReD)grouping loss captured per sample sizeImproved
increase LLM model sizeLLM zero-shot confidence scores · Folktexts (27 LLMs, 5 ACS tasks)normalized grouping loss: grouping loss decreases with model sizeImproved
apply instruction tuning to base LLMLLM zero-shot confidence scores · Folktexts (27 LLMs, 5 ACS tasks)normalized grouping loss: lower (or equal) grouping loss than base counterpartsImproved
use epistemic risk cascade instead of using largest modelLLM pool · ACSIncome (Llama 3 instruct)accuracy: +4% accuracy at about 30% of costImproved
use epistemic risk cascade instead of largest modelLLM pool · Folktexts 5 datasets, 120 poolsaccuracy: 6% accuracy gain at 46% of the costImproved
add grouping risk (full epistemic risk) on top of calibration risk for deferral (ablation)Cascade · ACSEmploymentaccuracy: up to 15%Improved
use epistemic risk cascade instead of predictive routerCascade · Folktexts 5 datasetsaccuracy and cost: strict Pareto optimality in 60% of casesImproved
use full epistemic risk instead of calibration risk with more than 6000 training samples (ablation)Cascade · All Folktexts datasetsaccuracy or cost: consistently improves above 6,000 samplesImproved
use full epistemic risk with fewer than 6000 training samples (ablation)Cascade · Folktexts datasetsaccuracy: up to +12% with 2,000 samples on ACSEmploymentMixed

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