AI

CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/10/2026, 04:03:24 UTC.

Briefings

CohortFlow is a steerable clinical cohort curation agent that replaces full regeneration with localized semantic corrections. On 100 scenarios from clinical-trial eligibility criteria, it achieved 77.0% micro-F1 against 76.4% for full regeneration, with a non-significant p-value of 0.493. The system reduced mean correction latency by 64.8% and execution latency by 44.3%, while using 38.1% fewer model calls during the 126 correction turns.

Key points

• CohortFlow uses localized semantic updates to replace full regeneration of clinical cohort specifications.
• On 100 scenarios, the agent achieved 77.0% micro-F1 against 76.4% for full regeneration, with a non-significant p-value of 0.493.

Context

This research addresses the challenge of clinical trial eligibility criteria evolving without reprocessing unchanged requirements. The method uses explicit semantic dependencies to selectively invalidate affected computations while preserving valid prior results.

Limitations

The evaluation covers only 100 scenarios derived from clinical-trial eligibility criteria. Results are based on an 8B model and a separate proof-of-concept over MIMIC data. Statistical significance was not found for the primary metric, and the study is a proof-of-concept evaluation.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 28/10/2026, 05:04:35 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeiao4qigck2p6elkuri55bwrhrtixelmnvcunv2jzytzdihj533uwy