pith. sign in

Constrained Paraphrase Consistency for LLM Hallucination Detection

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
abstract

Large language models (LLMs) can generate factually inconsistent claims, motivating accurate and scalable hallucination detectors. Prior work largely enlarges training sets via synthesis or new annotations, introducing increasing cost and potential bias while underusing the consistency implied by semantically equivalent paraphrases. We propose Consistency-Constrained Hallucination Detector (CCHD), which formulates training as a constrained optimization problem. The standard cross-entropy on original document-claim pairs is complemented by (i) paraphrase-consistency constraints bounding divergence across paraphrased views, and (ii) label-preservation constraints tying paraphrases to ground truth. We solve the problem by gradient descent-ascent over model parameters and per-view Lagrange multipliers, adding only a few scalar dual variables and no inference-time overhead. With DeBERTa and Flan-T5 backbones, CCHD consistently outperforms strong baselines (FactCG, MiniCheck, and AlignScore) on standard factuality benchmarks, demonstrating its superiority on hallucination detection.

fields

cs.CL 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Constrained Paraphrase Consistency for LLM Hallucination Detection

cs.CL · 2026-06-06 · unverdicted · novelty 6.0

CCHD formulates hallucination detector training as constrained optimization with paraphrase-consistency and label-preservation rules solved via gradient descent-ascent, outperforming baselines on factuality benchmarks.

citing papers explorer

Showing 1 of 1 citing paper.

  • Constrained Paraphrase Consistency for LLM Hallucination Detection cs.CL · 2026-06-06 · unverdicted · none · ref 1 · internal anchor

    CCHD formulates hallucination detector training as constrained optimization with paraphrase-consistency and label-preservation rules solved via gradient descent-ascent, outperforming baselines on factuality benchmarks.