What question did this study set out to answer?

The aim is to reduce hallucination in large language models by incorporating epistemic metadata into training data.

April 7, 2026Open Access

VKB-Training: Epistemically Categorized Training Data for Hallucination Reduction in Large Language Models

Key Points

The aim is to reduce hallucination in large language models by incorporating epistemic metadata into training data.
Developed VKB-Training with five epistemic tags for training data.
Utilized a four-stage hybrid annotation pipeline for categorizing training samples.
Implemented training mechanisms like confidence-weighted loss and provenance-aware attention.
Demonstrated a reduction in hallucination rates during model evaluation.
Established a formalized experimental protocol to test VKB-Training effects effectively.

Abstract

NOTE: ILLUSTRATIVE NUMBERS - WIPLarge language models hallucinate because their training data carries no epistemic metadata: facts, hypotheses, value judgments, and acknowledged unknowns occupy the same embedding space with identical weight. We propose VKB-Training (Verified Knowledge Base Training), a data-centric approach that assigns each training sample a five-category epistemic tag (Fact, Model, Value, Hypothesis, BlindSpot), a calibrated confidence score, and a provenance chain. We introduce a four-stage hybrid annotation pipeline:(1) AI triangulation — multiple LLMs classify independently; inter-model disagreement signals normative content (the "Caesar/God boundary")(2) Human sampling with axiom extraction — domain annotators resolve high-disagreement cases; recurrent decision principles are extracted as reusable rules(3) Expert calibration with reputation weighting — formalized Galton's ox-weighing insight (per S.V.E. XI, DOI: 10.5281/zenodo.18109198)(4) Logical consistency filters — contradiction detection and symmetry verification via the CGS Method (DOI: 10.5281/zenodo.18776172) Three training mechanisms are proposed: confidence-weighted loss, provenance-aware attention, and a BlindSpot training objective that maximizes output entropy at known knowledge gaps. VKB-Training was first described as part of the CogOS framework (DOI: 10.5281/zenodo.18109244). This paper extracts and formalizes the VKB component as a standalone, empirically testable proposal with a falsifiable experimental protocol and pre-specified success thresholds. Section 7 (Ethical Data Sourcing: Author Revenue Sharing, 10-50%) is included in the preprint but will be omitted from the workshop submission. Prepared for submission to NeurIPS 2026 Workshop.

AI에게 질문

Bookmark

View Full Paper