[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86133-en":3,"doc-seo-86133-105":30,"detail-sidebar-cat-0-en-105":84},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86133,962075114765,"Quinn","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Adapting Evidential Neural Networks to Test-Time Neighbor Fusion Improves Molecular Property Prediction","A trained molecular property model can be refined at test time by adjusting each prediction using measured labels of the most similar training molecules, a retraining-free procedure termed neighbor fusion. Evidential neural networks enable a principled Bayesian update by parameterizing aleatoric and epistemic uncertainty. The proposed PG-EVIKAL improves fusion by learning a property-distance metric that re-ranks structurally similar neighbors for property relevance, reducing RMSE and improving calibration across 16 datasets, including sequential-assay settings.","Adapting Evidential Neural Networks to Test-Time Neighbor Fusion Improves Molecular Property Prediction  \nCameron Gruich†* Weichi Yao‡, § Yixin Wang‡ Bryan Goldsmith†  \n† Department of Chemical Engineering, University of Michigan, Ann Arbor, Michigan 48109-2136, USA § Michigan Institute for Data & AI in Society, University of Michigan, Ann Arbor, Michigan 48109-1042, USA ‡ Department of Statistics, University of Michigan, Ann Arbor, Michigan 48109-1107 USA  \narXiv :2607 . 1 109 1v 1 [ cs .LG] 13 Jul 2026  \nAbstract  \nA trained molecular property model can be refined at test time by correcting each prediction with the measured labels of the most similar training molecules, a retraining-free procedure we call neighbor fusion; evidential neural networks make it principled by using their aleatoric and epistemic uncertainty to parameterize a Bayesian update.  \nOur main contribution, PG-EVIKAL, learns a property-distance metric to re-rank structurally similar neighbors by their property relevance before fusion, building on EVIKAL (scalar Kalman filter) and GP-EVIKAL (Gaussian process variant handling correlated neighbors) . Evaluated on 16 molecular datasets, PG-EVIKAL reduces RMSE relative to the evidential model baseline on 14 of them, with a median reduction of 19 .4%, and improves calibration; in sequential-assay scenariosit further incorporates newly measured molecules, refining predictions as they arrive without retraining. This work demonstrates that evidential uncertainty decomposition is not merely a calibration objective but an actionable inference resource that enables test-time refinement of molecular property predictions.  \n1. Introduction  \nDeep learning has emerged as a powerful tool for molecular property prediction (Wu et al., 2018; Yang et al., 2019), enabling rapid in-silico screening of chemical libraries that would be prohibitively expensive to evaluate experimentally or from first principles modeling. However, point predictions alone are insufficient for many molecular discovery workflows. Knowing how uncertain a prediction is can beas valuable as the prediction itself (Hirschfeld et al., 2020), directing resources toward the most informative experi-  \n*Corresponding author. E-mail: [cameron.gruich@gmail.com](cameron.gruich@gmail.com)  \nments and flagging unreliable estimates before they misdirect costly assays. This has motivated a line of work on Bayesian and uncertainty-aware molecular models that aim to augment predictions with calibrated uncertainty intervals (Scalia et al., 2020; Soleimany et al., 2021) .  \nUncertainty quantification (UQ) methods typically provide prediction-centered intervals that encode the model’s confidence in its own prediction (Kuleshov et al., 2018; Abdar et al., 2021) . A natural question is whether this confidence signal can be used not just to rank predictions but also to actively improve them. If a model is uncertain about a query molecule of interest, structurally similar training molecules with known property labels could, in principle, correct or sharpen the query prediction. This line of thinking motivates test-time neighbor fusion: using the training set as a source of a corrective signal at inference time, without any model updates (Figure 1) .  \nThe natural alternative is active learning, which directs new measurements toward molecules where predictive uncertainty is high and then retrains the model on the expanded dataset (Reker & Schneider, 2015; Graff et al., 2021) . Active learning is principled and well-studied, but requires retraining after each acquisition cycle (Settles, 2009) . Such retraining is impractical when training takes hours, assigning new labels is prohibitively expensive, or the model is already in active deployment. Test-time neighbor fusion sidesteps this entirely. The training set is queried at inference, and neighbor labels are fused with the model’s prediction to refine the estimate without gradient updates.  \nEffective neighbor fusion ideally ut","cbCaic5ZNTozueIP","https://ap.wps.com/l/cbCaic5ZNTozueIP","pdf",10860608,7,1,45,"English","en",105,"# Introduction\n## Test-time neighbor fusion concept\n## Relationship to uncertainty quantification and active learning\n## Overview of evidential regression approach","[{\"question\":\"What does PG-EVIKAL add over EVIKAL and GP-EVIKAL?\",\"answer\":\"PG-EVIKAL learns a property-distance metric to re-rank structurally similar neighbors by their property relevance before performing fusion, improving RMSE and calibration.\"}]",1784208806,113,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":79,"head_meta":81,"extra_data":83,"updated_unix":28},"adapting-evidential-neural-networks-to-test-time-neighbor-fusion-improves-molecular-property-prediction","",{"@graph":36,"@context":78},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/adapting-evidential-neural-networks-to-test-time-neighbor-fusion-improves-molecular-property-prediction/86133/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72],{"name":73,"@type":74,"acceptedAnswer":75},"What does PG-EVIKAL add over EVIKAL and GP-EVIKAL?","Question",{"text":76,"@type":77},"PG-EVIKAL learns a property-distance metric to re-rank structurally similar neighbors by their property relevance before performing fusion, improving RMSE and calibration.","Answer","https://schema.org",{"og:url":52,"og:type":80,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":82,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":85},[86,90,94,98,103,108,112,115,120,123,127],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":87,"show_sort_weight":88,"slug":89},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":91,"show_sort_weight":92,"slug":93},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Exam",70,"exam",{"id":99,"doc_module":4,"doc_module_name":46,"category_name":100,"show_sort_weight":101,"slug":102},5,"Comic",60,"comic",{"id":104,"doc_module":4,"doc_module_name":46,"category_name":105,"show_sort_weight":106,"slug":107},6,"Technology",50,"technology",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":109,"show_sort_weight":110,"slug":111},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":46,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":99,"slug":130},19,"General","general"]