Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lim, Byeonggeuk, Yun, JungMin, Kwon, Junehyoung, Kim, Kyeonghyun, Kim, YoungBin
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910169553698816
author Lim, Byeonggeuk
Yun, JungMin
Kwon, Junehyoung
Kim, Kyeonghyun
Kim, YoungBin
author_facet Lim, Byeonggeuk
Yun, JungMin
Kwon, Junehyoung
Kim, Kyeonghyun
Kim, YoungBin
contents Large Vision-Language Models (LVLMs) frequently suffer from hallucinations. Existing preference learning-based approaches largely rely on proprietary models to construct preference datasets. We identify that this reliance introduces a distributional mismatch between the proprietary and target models that hinders efficient alignment. To address this, we propose Alignment via VErified Self-correction DPO (AVES-DPO), a framework that aligns LVLMs using in-distribution data derived from the model's intrinsic knowledge. Our approach employs a consensus-based verification mechanism to diagnose diverse hallucinations and guides the model to self-correct, thereby generating preference pairs strictly compatible with its internal distribution. Extensive experiments demonstrate that AVES-DPO surpasses existing baselines in hallucination mitigation while requiring only 5.2k samples.
format Preprint
id arxiv_https___arxiv_org_abs_2604_24395
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs
Lim, Byeonggeuk
Yun, JungMin
Kwon, Junehyoung
Kim, Kyeonghyun
Kim, YoungBin
Artificial Intelligence
Large Vision-Language Models (LVLMs) frequently suffer from hallucinations. Existing preference learning-based approaches largely rely on proprietary models to construct preference datasets. We identify that this reliance introduces a distributional mismatch between the proprietary and target models that hinders efficient alignment. To address this, we propose Alignment via VErified Self-correction DPO (AVES-DPO), a framework that aligns LVLMs using in-distribution data derived from the model's intrinsic knowledge. Our approach employs a consensus-based verification mechanism to diagnose diverse hallucinations and guides the model to self-correct, thereby generating preference pairs strictly compatible with its internal distribution. Extensive experiments demonstrate that AVES-DPO surpasses existing baselines in hallucination mitigation while requiring only 5.2k samples.
title Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs
topic Artificial Intelligence
url https://arxiv.org/abs/2604.24395