AI decision benchmarks and research
Browse TypeSafe AI’s WorkflowEvals datasets and third-party preprint summaries about Jev behavior. Every entry identifies its source, verification date, and limits.
WorkflowEvals datasets
Provider-published workflow cases, question instances, and run results. These pages explain the data and do not represent an independent Type Safe ranking.
TypeSafe WorkflowEvals
Four provider-published datasets for testing model decisions inside invoice, customer-service, security, and agent-review workflows.
View evidenceWorkflowEvals Invoice Processing
A 150-case dataset for invoice approval and fraud-triage decisions under three policy settings.
View evidenceWorkflowEvals Customer Service
A 204-case dataset for customer-service routing and actions under three policy settings.
View evidenceWorkflowEvals Security Incidents
A 240-case dataset for security-alert triage, containment, and action selection.
View evidenceWorkflowEvals Agent Trace Observability
A 111-case dataset for selecting review actions from agent traces under two review profiles.
View evidenceThird-party Jev research
Research summaries based on author-reported results. The preprints have not completed formal peer review, and their metrics apply only to the described tasks and settings.
October 2026 Jev Research Overview
A source-backed overview of six independent, third-party Jev preprints covering theory of mind, application security, adversarial robustness, medicine, and decision bias.
View evidenceJev Arithmetic Rejection Bottleneck
A third-party preprint reporting a sharp drop in Jev rejection behavior when arithmetic answer menus omit the correct option.
View evidenceJev Ordinal-Scale Bias
A third-party preprint examining how ordered answer scales affect option use and errors in JEV-like direct-decision models.
View evidenceNo independent leaderboard results currently meet the publication standard
The resources above include provider datasets and third-party research, but their tasks, model access, hardware, and scoring methods are not uniform enough for a fair overall ranking.
Last reviewed: 2026-09-23
Read the methodology and eligibility rules