Hallucination as output-boundary misclassification: a composite abstention architecture for language models

TL;DR

This paper introduces a composite abstention architecture combining instruction-based refusal and a structural support deficit gate to reduce hallucinations in large language models.

Contribution

It proposes a novel composite intervention that effectively combines two mechanisms to control hallucinations, outperforming individual methods.

Findings

The composite architecture achieves high accuracy with low hallucination across multiple models.

Instruction prompting reduces hallucination but causes over-caution and residual errors.

Structural gating maintains answer accuracy and provides a baseline abstention capability.

Abstract

Large language models often produce unsupported claims. We frame this as a misclassification error at the output boundary, where internally generated completions are emitted as if they were grounded in evidence. This motivates a composite intervention that combines instruction-based refusal with a structural abstention gate. The gate computes a support deficit score, St, from three black-box signals: self-consistency (At), paraphrase stability (Pt), and citation coverage (Ct), and blocks output when St exceeds a threshold. In a controlled evaluation across 50 items, five epistemic regimes, and three models, neither mechanism alone was sufficient. Instruction-only prompting reduced hallucination sharply, but still showed over-cautious abstention on answerable items and residual hallucination for GPT-3.5-turbo. The structural gate preserved answerable accuracy across models but missed confident confabulation on conflicting-evidence items. The composite architecture achieved high overall accuracy with low hallucination, while also inheriting some over-abstention from the instruction component. A supplementary 100-item no-context stress test derived from TruthfulQA showed that structural gating provides a capability-independent abstention floor. Overall, instruction-based refusal and structural gating show complementary failure modes, which suggests that effective hallucination control benefits from combining both mechanisms.