🧪 Test?View on arXiv
CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction
Not provided in the abstract
clinical AIcohort curationsemantic correction
2610.36100
Builder Relevance
1h ago80%
Abstract
CohortFlow is a steerable agent for clinical cohort curation that efficiently accommodates evolving eligibility criteria through localized updates.
Reality Card
Core Claim
Localized steering in CohortFlow reduces mean correction latency by 64.8% and model calls by 38.1% while maintaining requirement-level accuracy.
Method / Result
Achieved 77.0% requirement micro-F1 with localized steering compared to 76.4% for full regeneration.
Limitations
The statistical significance of the performance improvement is not established (p=0.493).
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.