The reliability of large language models remains a critical challenge, particularly due to their susceptibility to hallucinations and factual inaccuracies during text generation. Existing solutions either underutilize models' self-correction with preemptive strategies or use costly post-hoc verification. To further explore the potential of real-time self-verification and correction, we present Dynamic Self-Verify Decoding (DSVD), a novel decoding framework that enhances generation reliability through real-time hallucination detection and efficient error correction. DSVD integrates two key components: (1) parallel self-verification architecture for continuous quality assessment, (2) dynamic rollback mechanism for targeted error recovery. Extensive experiments across five benchmarks demonstrate DSVD's effectiveness, achieving significant improvement in truthfulness (Quesetion-Answering) and factual accuracy (FActScore). Results show the DSVD can be further incorporated with existing faithful decoding methods to achieve stronger performance. Our work establishes that real-time self-verification during generation offers a viable path toward more trustworthy language models without sacrificing practical deployability.
View on arXiv@article{guo2025_2503.03149,
title={ DSVD: Dynamic Self-Verify Decoding for Faithful Generation in Large Language Models },
author={ YiQiu Guo and Yuchen Yang and Zhe Chen and Pingjie Wang and Yusheng Liao and Ya Zhang and Yanfeng Wang and Yu Wang },
journal={arXiv preprint arXiv:2503.03149},
year={ 2025 }
}