AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

KnowGuard: Knowledge-Driven Abstention for Multi-Round Clinical Reasoning

Xilin Dang, Kexin Chen, Xiaorui Su, Ayush Noori, Iñaki Arango, Lucas Vittor, LONG XINYI, Yuyang Du, Marinka Zitnik, Pheng-Ann Heng

ICLR 2026 · Large language models › Retrieval and knowledge · approach: Domain-specific QA datasets

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace LLM self-assessment abstention with knowledge-graph investigationGPT-4 doctor agent · ioAFRIMEDQA/ioMEDQA/ioCRAFT-MDaccuracy: ioMEDQA 64.23 -> 70.98 vs Scale RatingImproved
add evidence evaluation stage (ranking) to KG retrieval (ablation)KnowGuard · ioAFRIMEDQAaccuracy: 66.22 -> 72.60Improved
add evidence evaluation stage (ablation)KnowGuard · ioAFRIMEDQAavg. turns: 2.69 -> 7.03Mixed
add multi-modal KG triplet evidence on top of text evidence (ablation)KnowGuard · ioMEDQAaccuracy: 64.79 -> 70.66Improved
add text evidence retrieval (no KG) (ablation)Scale Rating-style baseline · ioCRAFT-MDaccuracy: 65.40 -> 62.73No clear effect
add patient population reasoning (PPR) (ablation)KnowGuard · ioAFRIMEDQAaccuracy: 72.60 -> 73.20Improved
add patient population reasoning (PPR) (ablation)KnowGuard · ioMEDQA and ioCRAFT-MDaccuracy: 74.29 -> 74.12; 71.92 -> 71.96No clear effect
add rationale generation and self-consistencyKnowGuard · ioAFRIMEDQAaccuracy: 68.70 -> 73.20Improved
add rationale generation and self-consistencyBinary Decision · ioMEDQAavg. turns: 7.69 -> 13.00Hurt
add rationale generation and self-consistencyScale Rating · ioAFRIMEDQAaccuracy: 63.06 -> 61.36Hurt
use Long Context retrieval instead of KG explorationLong Context · three datasetsaccuracy: ioMEDQA 59.95 vs 70.98Hurt
vary evidence weights around optimum (e.g. LLM relevance 0.5/0.6/0.7) (ablation)KnowGuard · ioMEDQAaccuracy: 70.59 / 74.12 / 71.64Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy