Structure Before Intelligence: Gating What Agentic Scientific Workflows Release
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.
Related
- TIAgent: From Natural Language to Grounded Agentic Workflows for Computational Pathology
- MedMCP: An Open Agentic Ecosystem for Medical Imaging Workflows
- The Findings Workflow: An agentic, collaborative, provenance-tracked framework for scientific data analysis with Claude Code
- Agentic cryo-EM structure determination with CryoPilot
- Debate, Not Voting: Agentic Arbitration over Weak Optical Chemical Structure Recognition Models
BibTeX
@misc{acharya2026structure,
title = {{Structure Before Intelligence: Gating What Agentic Scientific Workflows Release}},
author = {Ashish Acharya and Aman Al Amin and Hong Qin},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.19507},
eprint = {2610.19507},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.19507},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Structure Before Intelligence: Gating What Agentic Scientific Workflows Release
AU - Ashish Acharya
AU - Aman Al Amin
AU - Hong Qin
PY - 2026
DA - 2026/10/14
PB - arXiv
DO - 10.48550/arxiv.2610.19507
UR - https://arxiv.org/abs/2610.19507
AB - 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.
ER -