NAACL: Noise-AwAre Verbal Confidence Calibration for Robust LLMs in RAG Systems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Jiayu, Wang, Rui, Zong, Qing, Wang, Yumeng, Qian, Cheng, Zeng, Qingcheng, Zheng, Tianshi, Shi, Haochen, Guo, Dadi, Xu, Baixuan, Li, Chunyang, Song, Yangqiu
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918398715232256
author Liu, Jiayu
Wang, Rui
Zong, Qing
Wang, Yumeng
Qian, Cheng
Zeng, Qingcheng
Zheng, Tianshi
Shi, Haochen
Guo, Dadi
Xu, Baixuan
Li, Chunyang
Song, Yangqiu
author_facet Liu, Jiayu
Wang, Rui
Zong, Qing
Wang, Yumeng
Qian, Cheng
Zeng, Qingcheng
Zheng, Tianshi
Shi, Haochen
Guo, Dadi
Xu, Baixuan
Li, Chunyang
Song, Yangqiu
contents Accurately assessing model confidence is essential for deploying large language models (LLMs) in mission-critical factual domains. While retrieval-augmented generation (RAG) is widely adopted to improve grounding, confidence calibration in RAG settings remains poorly understood. We conduct a systematic study across four benchmarks, revealing that LLMs exhibit poor calibration performance due to noisy retrieved contexts. Specifically, contradictory or irrelevant evidence tends to inflate the model's false certainty, leading to severe overconfidence. To address this, we propose NAACL Rules (Noise-AwAre Confidence CaLibration Rules) to provide a principled foundation for resolving overconfidence under noise. We further design NAACL, a noise-aware calibration framework that synthesizes supervision from about 2K HotpotQA examples guided by these rules. By performing supervised fine-tuning (SFT) with this data, NAACL equips models with intrinsic noise awareness without relying on stronger teacher models. Empirical results show that NAACL yields substantial gains, improving ECE scores by 10.9% in-domain and 8.0% out-of-domain. By bridging the gap between retrieval noise and verbal calibration, NAACL paves the way for both accurate and epistemically reliable LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2601_11004
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle NAACL: Noise-AwAre Verbal Confidence Calibration for Robust LLMs in RAG Systems
Liu, Jiayu
Wang, Rui
Zong, Qing
Wang, Yumeng
Qian, Cheng
Zeng, Qingcheng
Zheng, Tianshi
Shi, Haochen
Guo, Dadi
Xu, Baixuan
Li, Chunyang
Song, Yangqiu
Computation and Language
Accurately assessing model confidence is essential for deploying large language models (LLMs) in mission-critical factual domains. While retrieval-augmented generation (RAG) is widely adopted to improve grounding, confidence calibration in RAG settings remains poorly understood. We conduct a systematic study across four benchmarks, revealing that LLMs exhibit poor calibration performance due to noisy retrieved contexts. Specifically, contradictory or irrelevant evidence tends to inflate the model's false certainty, leading to severe overconfidence. To address this, we propose NAACL Rules (Noise-AwAre Confidence CaLibration Rules) to provide a principled foundation for resolving overconfidence under noise. We further design NAACL, a noise-aware calibration framework that synthesizes supervision from about 2K HotpotQA examples guided by these rules. By performing supervised fine-tuning (SFT) with this data, NAACL equips models with intrinsic noise awareness without relying on stronger teacher models. Empirical results show that NAACL yields substantial gains, improving ECE scores by 10.9% in-domain and 8.0% out-of-domain. By bridging the gap between retrieval noise and verbal calibration, NAACL paves the way for both accurate and epistemically reliable LLMs.
title NAACL: Noise-AwAre Verbal Confidence Calibration for Robust LLMs in RAG Systems
topic Computation and Language
url https://arxiv.org/abs/2601.11004