Structured checks improve trustworthiness of scientific workflow outputs
Structure Before Intelligence: Gating What Agentic Scientific Workflows Release
Artificial Intelligence
Summary
Scientific workflows can sometimes produce wrong results without anyone noticing. The authors looked at using a safety system that checks and controls which results are allowed to be shared. This system includes rechecking calculations, validating outputs, and estimating risk before releasing results. They found that while these controls helped reduce mistakes, language-model agents used to fix errors didn't make a big difference. Overall, this approach shows how organizing the release process can help keep scientific results reliable.
What this means in practice
- •For data engineering teams: Implement workflow control systems that enforce error checks and risk assessments before releasing data outputs.
- •For software quality assurance teams: Use structured containment and validation to prevent propagation of faults in complex data pipelines.
Authors
Ashish Acharya, Aman Al Amin, Hong Qin
Abstract
An agentic scientific workflow can run to completion and still release wrong results: a silent numerical error, a stale input, or a defect in a shared upstream step can contaminate several outputs while a sibling branch stays correct. We treat trust in such results as a release decision. A harness around the workflow's directed acyclic graph (DAG) decides which outputs may be used: it enforces mandatory checks, contains failures to the affected branch, recomputes from tracked versions, adds validators, and releases a result only if a risk model estimates its probability of being invalid as low. Language-model agents work inside the harness, choosing corrective actions and reviewing releases, but cannot override it. On 60 held-out paired instances of two workflows with injected faults, we switched components on and off. Each supplied a different property. Containment and recovery raised the number of correct results. Validation supplied safety: fixed rules without validation released 37 invalid results, while validating every candidate released none. Risk estimation supplied efficiency, matching that safety with 31% fewer validators at somewhat lower yield, although better targeting than random validation was not shown. Lineage evidence helped on one workflow. The language model added no measurable benefit: whether invalid results escaped depended on the harness, not the model. This proof of concept points to a practical path: placing agents inside explicit release-control structure can keep released scientific results trustworthy, and it gives more capable agents a measurable baseline to improve on.