Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The EpisTwin: A Knowledge Graph-Grounded Neuro-Symbolic Architecture for Personal AI

Giovanni Servedio, Potito Aghilar, Alessio Mattiace, Gianni Carmosino, Francesco Musicco, Gabriele Conte, Vito Walter Anelli, Tommaso Di Noia, Francesco Maria Donini

TL;DR

EpisTwin is introduced, a neuro-symbolic framework that grounds generative reasoning in a verifiable, user-centric Personal Knowledge Graph and demonstrates robust results across a suite of state-of-the-art judge models, offering a promising direction for trustworthy Personal AI.

Abstract

Personal Artificial Intelligence is currently hindered by the fragmentation of user data across isolated silos. While Retrieval-Augmented Generation offers a partial remedy, its reliance on unstructured vector similarity fails to capture the latent semantic topology and temporal dependencies essential for holistic sensemaking. We introduce EpisTwin, a neuro-symbolic framework that grounds generative reasoning in a verifiable, user-centric Personal Knowledge Graph. EpisTwin leverages Multimodal Language Models to lift heterogeneous, cross-application data into semantic triples. At inference, EpisTwin enables complex reasoning over the personal semantic graph via an agentic coordinator that combines Graph Retrieval-Augmented Generation with Online Deep Visual Refinement, dynamically re-grounding symbolic entities in their raw visual context. We also introduce PersonalQA-71-100, a synthetic benchmark designed to simulate a realistic user's digital footprint and evaluate EpisTwin performance. Our framework demonstrates robust results across a suite of state-of-the-art judge models, offering a promising direction for trustworthy Personal AI.

The EpisTwin: A Knowledge Graph-Grounded Neuro-Symbolic Architecture for Personal AI

TL;DR

EpisTwin is introduced, a neuro-symbolic framework that grounds generative reasoning in a verifiable, user-centric Personal Knowledge Graph and demonstrates robust results across a suite of state-of-the-art judge models, offering a promising direction for trustworthy Personal AI.

Abstract

Personal Artificial Intelligence is currently hindered by the fragmentation of user data across isolated silos. While Retrieval-Augmented Generation offers a partial remedy, its reliance on unstructured vector similarity fails to capture the latent semantic topology and temporal dependencies essential for holistic sensemaking. We introduce EpisTwin, a neuro-symbolic framework that grounds generative reasoning in a verifiable, user-centric Personal Knowledge Graph. EpisTwin leverages Multimodal Language Models to lift heterogeneous, cross-application data into semantic triples. At inference, EpisTwin enables complex reasoning over the personal semantic graph via an agentic coordinator that combines Graph Retrieval-Augmented Generation with Online Deep Visual Refinement, dynamically re-grounding symbolic entities in their raw visual context. We also introduce PersonalQA-71-100, a synthetic benchmark designed to simulate a realistic user's digital footprint and evaluate EpisTwin performance. Our framework demonstrates robust results across a suite of state-of-the-art judge models, offering a promising direction for trustworthy Personal AI.
Paper Structure (35 sections, 6 equations, 3 figures, 2 tables)

This paper contains 35 sections, 6 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Knowledge Graph Construction.
  4. Personal Knowledge Graphs.
  5. Graph-based Reasoning.
  6. Personal AI and Data Sovereignty.
  7. Problem Definition and Preliminaries
  8. Personal Knowledge Graph (PKG)
  9. KG Construction from Text
  10. Thematic Context Extraction via Community Detection
  11. Leiden Algorithm.
  12. Graph Retrieval-Augmented Generation
  13. LLM-based Agents
  14. Epistemic Twin Constructor
  15. PKG Population
  16. ...and 20 more sections

Figures (3)

  • Figure 1: PKG population when the Information Object is a photo: triples are extracted from both metadata and visual content.
  • Figure 2: Communities over the PKG: (a) The topologically disjoint entities "Alarm" and "Football Match Event" could be grouped into a shared community that reveals an implicit consequentiality, improving reasoning. (b) A macroscopic visualization of a PKG populated by entities, relationships, and thematic communities.
  • Figure 3: Judicial Panel Evaluation.Left: Distribution of scores assigned by LLM Judges on PersonalQA-71-100 to EpisTwin answers. Right: LLM judgments distribution after vote aggregation.

Theorems & Definitions (8)

  • Example 1: Sarah call
  • Definition 1: Information Object
  • Example 3
  • Definition 2: Personal Knowledge
  • Remark 1
  • Definition 3: Personal Knowledge Graph
  • Definition 4: KG Construction Function
  • Definition 5: Community Structure