Teaching Large Language Models to Maintain Contextual Faithfulness via Synthetic Tasks and Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Si, Shuzheng, Zhao, Haozhe, Gao, Cheng, Bai, Yuzhuo, Wang, Zhitong, Gao, Bofei, Luo, Kangyang, Li, Wenhao, Huang, Yufei, Chen, Gang, Qi, Fanchao, Zhang, Minjia, Chang, Baobao, Sun, Maosong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912702878711808
author Si, Shuzheng
Zhao, Haozhe
Gao, Cheng
Bai, Yuzhuo
Wang, Zhitong
Gao, Bofei
Luo, Kangyang
Li, Wenhao
Huang, Yufei
Chen, Gang
Qi, Fanchao
Zhang, Minjia
Chang, Baobao
Sun, Maosong
author_facet Si, Shuzheng
Zhao, Haozhe
Gao, Cheng
Bai, Yuzhuo
Wang, Zhitong
Gao, Bofei
Luo, Kangyang
Li, Wenhao
Huang, Yufei
Chen, Gang
Qi, Fanchao
Zhang, Minjia
Chang, Baobao
Sun, Maosong
contents Teaching large language models (LLMs) to be faithful in the provided context is crucial for building reliable information-seeking systems. Therefore, we propose a systematic framework, CANOE, to reduce faithfulness hallucinations of LLMs across different downstream tasks without human annotations. Specifically, we first synthesize short-form question-answering (QA) data with four diverse tasks to construct high-quality and easily verifiable training data without human annotation. Also, we propose Dual-GRPO, a rule-based reinforcement learning method that includes three tailored rule-based rewards derived from synthesized short-form QA data, while simultaneously optimizing both short-form and long-form response generation. Notably, Dual-GRPO eliminates the need to manually label preference data to train reward models and avoids over-optimizing short-form generation when relying only on the synthesized short-form QA data. Experimental results show that CANOE greatly improves the faithfulness of LLMs across 11 different tasks, even outperforming the most advanced LLMs, e.g., GPT-4o and OpenAI o1.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16483
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Teaching Large Language Models to Maintain Contextual Faithfulness via Synthetic Tasks and Reinforcement Learning
Si, Shuzheng
Zhao, Haozhe
Gao, Cheng
Bai, Yuzhuo
Wang, Zhitong
Gao, Bofei
Luo, Kangyang
Li, Wenhao
Huang, Yufei
Chen, Gang
Qi, Fanchao
Zhang, Minjia
Chang, Baobao
Sun, Maosong
Computation and Language
Artificial Intelligence
Teaching large language models (LLMs) to be faithful in the provided context is crucial for building reliable information-seeking systems. Therefore, we propose a systematic framework, CANOE, to reduce faithfulness hallucinations of LLMs across different downstream tasks without human annotations. Specifically, we first synthesize short-form question-answering (QA) data with four diverse tasks to construct high-quality and easily verifiable training data without human annotation. Also, we propose Dual-GRPO, a rule-based reinforcement learning method that includes three tailored rule-based rewards derived from synthesized short-form QA data, while simultaneously optimizing both short-form and long-form response generation. Notably, Dual-GRPO eliminates the need to manually label preference data to train reward models and avoids over-optimizing short-form generation when relying only on the synthesized short-form QA data. Experimental results show that CANOE greatly improves the faithfulness of LLMs across 11 different tasks, even outperforming the most advanced LLMs, e.g., GPT-4o and OpenAI o1.
title Teaching Large Language Models to Maintain Contextual Faithfulness via Synthetic Tasks and Reinforcement Learning
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.16483