Evidence-Guided Recovery for AI Agents with Ambiguous Tool Outcomes and Partial Execution

研究论文 Hugging Face Papers 原文语言:中文

论文摘要(作者报告,未经本站复测):Tool timeouts can leave AI agents facing ambiguous tool outcomes and partial execution: an action may have failed, committed or remained in flight while other workflow steps have already completed.

全文

TooltimeoutscanleaveAIagentsfacingambiguoustooloutcomesandpartialexecution:anactionmayhavefailed,committedorremainedinflightwhileotherworkflowstepshavealreadycompleted.Weformulateevidence-guidedtaskrecoveryoverapprovedobligations,providercapabilitiesandstatescompatiblewithpublicevidence.Asharedadmissibilitycheckhandleskeyexpiry,authorizationtiming,dependenciesandresidualeffects;aterminalcertificateseparatesactualgoalsatisfactionfrompubliclysupportedsafecompletion.Theknown-modelimplementationcombinesstandardobservation-contingentplanningwithaseparatelyimplementedeffect-historyadjudicator.Inafinitestudyof37assignments,planningandthreeactivecontrolseachcertify28completions;foursavedsyntheticfeeunitsincuradditionalcomputation.Apreregisteredassignment-held-outcomparisonthenevaluatesplanningadviceagainstevidencepromptingwithtwolocalmodelbundles.Across58pairedunitsin19previouslydevelopedtaskfamilies,advicecertifies30completionsversus9,a36.2-percentage-pointdifference(exactfamilysign-flipp=0.0078),using272proposalsversus399.All116trajectoriespassindependentreplay.Bothmodelsexhibitapositivecompletiondifference.Thecomparisonmeasurestheevaluatedadviceinterfaceunderacommonshield;search-versus-ruleadvicerequiresaseparatematchedcomparison.Unknownandhandoffoutcomesremainexplicit,withguaranteestiedtothedeclaredprovidercapabilitiesratherthanarbitrarythird-partyexactly-onceexecution.

View arXiv pageView PDF Add to collection

正文以来源语言显示。

参考来源

← 返回每日简报