DVD: Dynamic Contrastive Decoding for Knowledge Amplification in Multi-Document Question Answering

Jing Jin, Houfeng Wang, Hao Zhang, Xiaoguang Li, Zhijiang Guo


Abstract
Large language models (LLMs) are widely used in question-answering (QA) systems but often generate information with hallucinations. Retrieval-augmented generation (RAG) offers a potential remedy, yet the uneven retrieval quality and irrelevant contents may distract LLMs.In this work, we address these issues at the generation phase by treating RAG as a multi-document QA task.We propose a novel decoding strategy, Dynamic Contrastive Decoding, which dynamically amplifies knowledge from selected documents during the generation phase. involves constructing inputs batchwise, designing new selection criteria to identify documents worth amplifying, and applying contrastive decoding with a specialized weight calculation to adjust the final logits used for sampling answer tokens. Zero-shot experimental results on ALCE-ASQA, NQ, TQA and PopQA benchmarks show that our method outperforms other decoding strategies. Additionally, we conduct experiments to validate the effectiveness of our selection criteria, weight calculation, and general multi-document scenarios. Our method requires no training and can be integrated with other methods to improve the RAG performance. Our codes will be publicly available at https://github.com/JulieJin-km/Dynamic_Contrastive_Decoding.
Anthology ID:
2024.emnlp-main.266
Volume:
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
4624–4637
Language:
URL:
https://aclanthology.org/2024.emnlp-main.266
DOI:
Bibkey:
Cite (ACL):
Jing Jin, Houfeng Wang, Hao Zhang, Xiaoguang Li, and Zhijiang Guo. 2024. DVD: Dynamic Contrastive Decoding for Knowledge Amplification in Multi-Document Question Answering. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 4624–4637, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
DVD: Dynamic Contrastive Decoding for Knowledge Amplification in Multi-Document Question Answering (Jin et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.266.pdf
Software:
 2024.emnlp-main.266.software.zip