Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
This paper introduces a "harness" system that treats the release of results from agentic scientific workflows as a decision, enforcing mandatory checks, failure containment, recomputation from tracked versions, and validation via a risk model to estimate invalidity probability. The system successfully improved the number of correct results through containment and recovery, and achieved perfect safety by validating every candidate output, outperforming fixed rules. While language models did not demonstrably improve safety or efficiency, the harness itself provides a robust control mechanism for trustworthy scientific output.
Agentic workflows can be made trustworthy by gating their output with a risk-aware harness, even when the LLM agent itself fails to detect errors.
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.