An Analysis of Decoding Methods for LLM-based Agents for Faithful Multi-Hop Question Answering

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Murphy, Alexander, Rizvi, Mohd Sanad Zaki, Haussmann, Aden, Nie, Ping, Liu, Guifu, Gema, Aryo Pradipta, Minervini, Pasquale
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909558117498880
author Murphy, Alexander
Rizvi, Mohd Sanad Zaki
Haussmann, Aden
Nie, Ping
Liu, Guifu
Gema, Aryo Pradipta
Minervini, Pasquale
author_facet Murphy, Alexander
Rizvi, Mohd Sanad Zaki
Haussmann, Aden
Nie, Ping
Liu, Guifu
Gema, Aryo Pradipta
Minervini, Pasquale
contents Large Language Models (LLMs) frequently produce factually inaccurate outputs - a phenomenon known as hallucination - which limits their accuracy in knowledge-intensive NLP tasks. Retrieval-augmented generation and agentic frameworks such as Reasoning and Acting (ReAct) can address this issue by giving the model access to external knowledge. However, LLMs often fail to remain faithful to retrieved information. Mitigating this is critical, especially if LLMs are required to reason about the retrieved information. Recent research has explored training-free decoding strategies to improve the faithfulness of model generations. We present a systematic analysis of how the combination of the ReAct framework and decoding strategies (i.e., DeCoRe, DoLa, and CAD) can influence the faithfulness of LLM-generated answers. Our results show that combining an agentic framework for knowledge retrieval with decoding methods that enhance faithfulness can increase accuracy on the downstream Multi-Hop Question Answering tasks. For example, we observe an F1 increase from 19.5 to 32.6 on HotpotQA when using ReAct and DoLa.
format Preprint
id arxiv_https___arxiv_org_abs_2503_23415
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle An Analysis of Decoding Methods for LLM-based Agents for Faithful Multi-Hop Question Answering
Murphy, Alexander
Rizvi, Mohd Sanad Zaki
Haussmann, Aden
Nie, Ping
Liu, Guifu
Gema, Aryo Pradipta
Minervini, Pasquale
Computation and Language
Artificial Intelligence
Large Language Models (LLMs) frequently produce factually inaccurate outputs - a phenomenon known as hallucination - which limits their accuracy in knowledge-intensive NLP tasks. Retrieval-augmented generation and agentic frameworks such as Reasoning and Acting (ReAct) can address this issue by giving the model access to external knowledge. However, LLMs often fail to remain faithful to retrieved information. Mitigating this is critical, especially if LLMs are required to reason about the retrieved information. Recent research has explored training-free decoding strategies to improve the faithfulness of model generations. We present a systematic analysis of how the combination of the ReAct framework and decoding strategies (i.e., DeCoRe, DoLa, and CAD) can influence the faithfulness of LLM-generated answers. Our results show that combining an agentic framework for knowledge retrieval with decoding methods that enhance faithfulness can increase accuracy on the downstream Multi-Hop Question Answering tasks. For example, we observe an F1 increase from 19.5 to 32.6 on HotpotQA when using ReAct and DoLa.
title An Analysis of Decoding Methods for LLM-based Agents for Faithful Multi-Hop Question Answering
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2503.23415