AI Science Tracker
EN

University of Utah

3 discoveries from 2020 to 2026

Discoveries

Agentic AI Reasoning and Learned Sampling for Task-Aware Projection Selection in Sparse-View CT: Predicting Hemodynamics

Source check: Source checked. Not independent scientific confirmation.

Finding: On simulated aneurysm CTA, task-aware view masks lower spatially resolved TAWSS errors at extreme sparsity, while prediction-optimal masks reconstruct images poorly. This is the result-bearing preprint version dated 2026-10-28; peer-review status is not established by the reviewed source.

Technology: A Fourier neural operator predicts wall-shear stress from simulated sparse CTA projections; Gumbel, Soft top-K and an agentic selector choose views.

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A Fourier neural operator predicts wall-shear stress from simulated sparse CTA projections; Gumbel, Soft top-K and an agentic selector choose views.
Test and result
On simulated aneurysm CTA, task-aware view masks lower spatially resolved TAWSS errors at extreme sparsity, while prediction-optimal masks reconstruct images poorly. CFD-derived synthetic aneurysm CTA uses five-run comparisons and independent agent searches across view budgets; results do not establish patient or clinical validity.
Limitation / still unknown
not establish patient or clinical validity.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/39003v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Robot model activations signaled failures before recovery became unlikely in simulation

Source check: Source checked. Not independent scientific confirmation.

Finding: Researchers found that OpenVLA activation features could distinguish eventual failures from successful rollouts before recovery became unlikely in a simulated robot benchmark. The predictor reached an AUROC of 0.71 before the study’s estimated point of no return.

Technology: OpenVLA-7B, sparse autoencoders

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Fitted sparse autoencoders decompose OpenVLA activation patterns and fitted outcome predictors compute early warnings on held-out robot-policy traces.
Test and result
The study used 500 rollouts over ten simulated manipulation tasks, held out 95 trajectories, and estimated recoverability with alternative branched rollouts.
Limitation / still unknown
Recovery was defined using the study’s simulated branching protocol. The result does not establish reliable physical-robot intervention or superiority over raw activation predictors. Recoverability used only four alternative continuations per probe and depended on the tested policy and remaining time budget. Across reported alarm thresholds, false alarms ranged from 64% to 14% while actionable detections fell from 88% to 33%.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2610.32208 · https://arxiv.org/html/2610.32208v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 21 Nov 2026. Review scope: arXiv v1 §2 OpenVLA rollouts, §2.2/Table 1 and Appendix E recoverability estimates.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Clinical predictors for etiology of acute diarrhea in children in resource-limited settings

Source check: Source checked. Not independent scientific confirmation.

Finding: Random forests screened clinical predictors and random-forest/logistic-regression models predicted viral versus other known pediatric diarrhea etiologies in GEMS data. Five factors yielded cross-validated AUC 0.825 (sensitivity 0.59, specificity 0.85); this retrospective analysis does not establish bedside benefit.

Technology: Random-forest feature screening and random-forest/logistic-regression models predict viral diarrhea etiology.

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Random-forest feature screening and random-forest/logistic-regression models predict viral diarrhea etiology.
Test and result
Five clinical factors yielded cross-validated AUC 0.825 for viral etiology (sensitivity 0.59, specificity 0.85) using molecular-attribution reference data.
Limitation / still unknown
The model was retrospectively evaluated with five-fold cross-validation; no prospective clinical benefit was tested.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://www.medrxiv.org/content/10.1101/2020.01.27.20016725v1.full.pdf · https://doi.org/10.1371/journal.pntd.0008677 · https://pubmed.ncbi.nlm.nih.gov/33035209/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open the interactive profile for the timeline, or browse all organizations and model families.