Saved in:
Bibliographic Details
Main Authors: Askin, Deniz, Hadar, Gal, Conway-Smith, Brendan
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.16676
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909049264537600
author Askin, Deniz
Hadar, Gal
Conway-Smith, Brendan
author_facet Askin, Deniz
Hadar, Gal
Conway-Smith, Brendan
contents Metacognition-the ability to monitor one's own knowledge state, spot gaps, and autonomously fill them--remains largely absent from modern AI. Here, we present MetaKGEnrich, a fully automated pipeline that endows large language model (LLM) applications with self-directed knowledge repair. The system (i) builds knowledge graphs from a seed query, (ii) detects sparse regions via seven graph metrics, (iii) has GPT-4o generate targeted questions, (iv) retrieves web evidence with Tavily and ingests it into Neo4j, and (v) re-answers the query with GraphRAG for GPT-4 to evaluate improvement. Tested on 30 queries from each of three widely-used datasets: Google Research Natural Questions, MS MARCO, and Hot-potQA. MetaKGEnrich improved answer quality in 80% of HotpotQA questions, 87% of Google Research Natural Questions and 83% of MS MARCO questions, while preserving well-supported regions. This proof of concept demonstrates how topological self-diagnosis plus targeted retrieval can advance AI toward humanlike metacognitive learning.
format Preprint
id arxiv_https___arxiv_org_abs_2605_16676
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Enhancing Metacognitive AI: Knowledge-Graph Population with Graph-Theoretic LLM Enrichment
Askin, Deniz
Hadar, Gal
Conway-Smith, Brendan
Artificial Intelligence
Metacognition-the ability to monitor one's own knowledge state, spot gaps, and autonomously fill them--remains largely absent from modern AI. Here, we present MetaKGEnrich, a fully automated pipeline that endows large language model (LLM) applications with self-directed knowledge repair. The system (i) builds knowledge graphs from a seed query, (ii) detects sparse regions via seven graph metrics, (iii) has GPT-4o generate targeted questions, (iv) retrieves web evidence with Tavily and ingests it into Neo4j, and (v) re-answers the query with GraphRAG for GPT-4 to evaluate improvement. Tested on 30 queries from each of three widely-used datasets: Google Research Natural Questions, MS MARCO, and Hot-potQA. MetaKGEnrich improved answer quality in 80% of HotpotQA questions, 87% of Google Research Natural Questions and 83% of MS MARCO questions, while preserving well-supported regions. This proof of concept demonstrates how topological self-diagnosis plus targeted retrieval can advance AI toward humanlike metacognitive learning.
title Enhancing Metacognitive AI: Knowledge-Graph Population with Graph-Theoretic LLM Enrichment
topic Artificial Intelligence
url https://arxiv.org/abs/2605.16676