PanVasc Research for AI assisted evidence analysis in panvascular intervention
- 1Zhongshan Hospital, Fudan University
Abstract
Panvascular intervention research requires evidence workflows that preserve source identity, outcome definitions and observation windows. We developed PanVasc Research, an executable research framework, and evaluated a fixed local Qwen3-4B model using two complementary tasks. Fifty ClinicalTrials.gov records from five vascular query strata generated 200 source-fidelity tests with intact evidence or controlled removal of the requested source, primary outcome or timeframe, plus 50 clean controls. A separate 100-statement sample from the official NLI4CT test set assessed clinical-trial entailment and evidence selection against existing expert labels. Generic and checklist prompts used identical evidence and maximum generation budgets; each output was also evaluated with an input-only deterministic contract. Registry exact accuracy was 51/200 (25.5%) with the generic prompt and 77/200 (38.5%) with the checklist; intact-case agreement was 50/50 and 48/50. The rule baseline recovered 200/200 tuples. NLI label accuracy was 48/100 (48.0%) and 50/100 (50.0%), respectively. The source contract retained 35 and 31 incorrect NLI labels in the two arms. Registry references establish fidelity to a registration snapshot, while NLI4CT concerns breast-cancer trials and does not validate vascular expertise. The framework also retains provenance-recorded literature retrieval, structured research planning and local numerical analysis. These experiments support a bounded assessment of source handling and semantic failure, rather than a new foundation model or autonomous scientific discovery. A proposed endpoint representation identifies the additional domain annotation and independent validation required for a panvascular research model.
Related
- Panvascular-JEPA: Unseen-Structure Evaluation of Factorized Predictive States for Hemodynamic Networks
- Evidence-informed dynamic modelling and intervention prioritisation in diabetic bone marrow
- An AI-ready compositional framework for mechanistic aging research and in silico intervention testing
- LLM-assisted analytical review catches statistical errors that human researchers miss: documented evidence from two computational biology case studies
- Perceived factors associated with self-reported AI-assisted academic cheating among undergraduate students: Evidence from ethical ambiguity, technological affordances, institutional gaps, academic pressure, and peer comparison
BibTeX
@misc{you2026panvasc,
title = {{PanVasc Research for AI assisted evidence analysis in panvascular intervention}},
author = {L. You and Y. Guo and W. Wang and Z. Peng and X. Zhong and L. Shen and J. Ge},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.22.752857},
url = {https://www.biorxiv.org/content/10.64898/2026.09.22.752857v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - PanVasc Research for AI assisted evidence analysis in panvascular intervention
AU - L. You
AU - Y. Guo
AU - W. Wang
AU - Z. Peng
AU - X. Zhong
AU - L. Shen
AU - J. Ge
PY - 2026
DA - 2026/09/22
PB - bioRxiv
DO - 10.64898/2026.09.22.752857
UR - https://www.biorxiv.org/content/10.64898/2026.09.22.752857v1
AB - Panvascular intervention research requires evidence workflows that preserve source identity, outcome definitions and observation windows. We developed PanVasc Research, an executable research framework, and evaluated a fixed local Qwen3-4B model using two complementary tasks. Fifty ClinicalTrials.gov records from five vascular query strata generated 200 source-fidelity tests with intact evidence or controlled removal of the requested source, primary outcome or timeframe, plus 50 clean controls. A separate 100-statement sample from the official NLI4CT test set assessed clinical-trial entailment and evidence selection against existing expert labels. Generic and checklist prompts used identical evidence and maximum generation budgets; each output was also evaluated with an input-only deterministic contract. Registry exact accuracy was 51/200 (25.5%) with the generic prompt and 77/200 (38.5%) with the checklist; intact-case agreement was 50/50 and 48/50. The rule baseline recovered 200/200 tuples. NLI label accuracy was 48/100 (48.0%) and 50/100 (50.0%), respectively. The source contract retained 35 and 31 incorrect NLI labels in the two arms. Registry references establish fidelity to a registration snapshot, while NLI4CT concerns breast-cancer trials and does not validate vascular expertise. The framework also retains provenance-recorded literature retrieval, structured research planning and local numerical analysis. These experiments support a bounded assessment of source handling and semantic failure, rather than a new foundation model or autonomous scientific discovery. A proposed endpoint representation identifies the additional domain annotation and independent validation required for a panvascular research model.
ER -