AI Paper Atlas

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

Epistemic Uncertainty Quantification for Pre-trained VLMs via Riemannian Flow Matching

Li Ju, Mayank Nautiyal, Andreas Hellander, Ekta Vats, Prashant Singh

ICML 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 Riemannian (geodesic) flow matching instead of Euclidean variants (ablation)REPVLM · ImageNet-1K, Conceptual Captions proxyaccuracy-rejection curveImproved
increase proxy dataset size from 50K to 1M pairs (ablation)REPVLM · ImageNet-1K, Conceptual Captions proxyAcc at 90% rejectionImproved
increase proxy dataset beyond 500K samples (ablation)REPVLM · ImageNet-1K, Conceptual Captions proxyAcc at 90% rejectionNo clear effect
increase ODE integration steps from 1-2 to 5 (ablation)REPVLM · ImageNet-1K, Conceptual Captions proxyaccuracy-rejection curveImproved
increase ODE steps from 5 to 8 (ablation)REPVLM · ImageNet-1K, Conceptual Captions proxyaccuracy-rejection curveNo clear effect
use embedding density instead of MC dropout / ProbVLM for epistemic UQREPVLM · ImageNet-1K, Conceptual Captions proxyAcc@90% rejection: 0.710 -> 0.733 (vs MC-Dropout)Improved
use REPVLM instead of embedding norm on Food101REPVLM · Food101Acc@90% rejection: 0.934 (Emb. Norm) vs 0.904Hurt
use REPVLM instead of Conformity on ImageNet-SketchREPVLM · ImageNet-SketchAcc@90% rejection: 0.672 (Conformity) vs 0.656 (CCaps proxy)Hurt
train on noisy LAION proxy instead of cleaner proxiesREPVLM · ImageNet-SketchAcc@90% rejection: 0.656 -> 0.618Hurt
use density-based scoring with REPVLM versus a Euclidean GMM densityREPVLM vs GMM · Food101Spearman S: -0.479 -> 1.000Improved
use REPVLM with 5 ODE steps instead of MC dropout with 10 passesREPVLM · Computational costGFLOPs: 92.50 -> 9.80Improved

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