AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Anatomy of a Lie: A Multi-Stage Diagnostic Framework for Tracing Hallucinations in Vision-Language Models

Lexiang Xiong, QI LI, Jingwen Ye, Xinchao Wang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Decoding-time hallucination mitigation · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
combine all three probes in a GMM instead of using a single probe (ablation)CAD on Idefics2 · POPE adversarialAUC: 0.75-0.78 -> 0.947Improved
use unsupervised CAD instead of supervised probeCAD vs Supervised Probe · POPE adversarial (average)AUC: 0.791 -> 0.858Improved
use CAD instead of 10x-cost Semantic EntropyCAD vs Semantic Entropy · POPE adversarial (average)AUC: 0.709 -> 0.858Improved
use CAD instead of supervised probeCAD vs Supervised Probe · MME Qwen2-VLAUC: 0.7408 -> 0.7233Hurt
use CAD instead of supervised probeCAD vs Supervised Probe · MME (average)AUC: 0.7509 -> 0.7960Improved
contaminate calibration set with up to 30% hallucinations (ablation)CAD on Idefics2 · POPE adversarialAUC: above 0.91 at 30% contaminationNo clear effect
contaminate calibration set with hallucinations (ablation)CAD on Llava-v1.6 · POPE adversarialAUCHurt
contaminate calibration set with hallucinations (ablation)CAD on Qwen2-VL and DeepSeek-VL2 · POPE adversarialAUCNo clear effect
use H_Evi probe on open-ended captioning (ablation)H_Evi · MS-COCO captions (N=1000)H_Evi of hallucinated vs non-hallucinated captions: p << 0.001Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy