Saved in:
Bibliographic Details
Main Authors: Wu, Xixian, Ou, Yang, Tian, Pengchao, Yang, Zian, Zhang, Jielei, Li, Peiyi, Gao, Longwen
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.14770
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914205170401280
author Wu, Xixian
Ou, Yang
Tian, Pengchao
Yang, Zian
Zhang, Jielei
Li, Peiyi
Gao, Longwen
author_facet Wu, Xixian
Ou, Yang
Tian, Pengchao
Yang, Zian
Zhang, Jielei
Li, Peiyi
Gao, Longwen
contents Vision-language models (VLMs) have demonstrated significant potential in Visual Question Answering (VQA). However, the susceptibility of VLMs to hallucinations can lead to overconfident yet incorrect answers, severely undermining answer reliability. To address this, we propose Dual-Assessment for VLM Reliability (DAVR), a novel framework that integrates Self-Reflection and Cross-Model Verification for comprehensive uncertainty estimation. The DAVR framework features a dual-pathway architecture: one pathway leverages dual selector modules to assess response reliability by fusing VLM latent features with QA embeddings, while the other deploys external reference models for factual cross-checking to mitigate hallucinations. Evaluated in the Reliable VQA Challenge at ICCV-CLVL 2025, DAVR achieves a leading $Φ_{100}$ score of 39.64 and a 100-AUC of 97.22, securing first place and demonstrating its effectiveness in enhancing the trustworthiness of VLM responses.
format Preprint
id arxiv_https___arxiv_org_abs_2512_14770
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Improving VQA Reliability: A Dual-Assessment Approach with Self-Reflection and Cross-Model Verification
Wu, Xixian
Ou, Yang
Tian, Pengchao
Yang, Zian
Zhang, Jielei
Li, Peiyi
Gao, Longwen
Computer Vision and Pattern Recognition
Artificial Intelligence
Vision-language models (VLMs) have demonstrated significant potential in Visual Question Answering (VQA). However, the susceptibility of VLMs to hallucinations can lead to overconfident yet incorrect answers, severely undermining answer reliability. To address this, we propose Dual-Assessment for VLM Reliability (DAVR), a novel framework that integrates Self-Reflection and Cross-Model Verification for comprehensive uncertainty estimation. The DAVR framework features a dual-pathway architecture: one pathway leverages dual selector modules to assess response reliability by fusing VLM latent features with QA embeddings, while the other deploys external reference models for factual cross-checking to mitigate hallucinations. Evaluated in the Reliable VQA Challenge at ICCV-CLVL 2025, DAVR achieves a leading $Φ_{100}$ score of 39.64 and a 100-AUC of 97.22, securing first place and demonstrating its effectiveness in enhancing the trustworthiness of VLM responses.
title Improving VQA Reliability: A Dual-Assessment Approach with Self-Reflection and Cross-Model Verification
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2512.14770