Abstract
The evaluation of automated, multi-step digital workflows may face challenges such as context loss in long sequences, visual detail degradation from image processing, and non-deterministic results. A described technology can utilize a hybrid system architecture that may decouple visual analysis from state aggregation. A perception component can process visual captures of a workflow, such as screenshots, independently and potentially in parallel. This component can employ a high-resolution tiling method to analyze large images as smaller, overlapping segments, which can help preserve fine details for extraction by a machine learning model. A separate, programmatic aggregation component can then process the structured outputs in a specified order, for example, chronologically, using state-tracking logic to generate an assessment. This separation of components may be a factor in the reliability, scalability, and detail of automated workflow evaluations.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Patel, Shravan and Ge, Neil, "Hybrid Architecture for Digital Workflow Evaluation via Parallel Analysis and Deterministic Aggregation", Technical Disclosure Commons, (June 23, 2026)
https://www.tdcommons.org/dpubs_series/10544