AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

When Diffusion Language Models Hesitate: Detecting and Correcting Visual Hallucinations via Confidence Fluctuation

Wenzheng Song, Pei Chen, Yichen Tan, Zejian Li, Lingyun Sun

ICML 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Preference tuning against hallucination

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Studies visual hallucinations in multi-modal diffusion language models (MDLMs) and proposes detecting and correcting them via confidence fluctuation during decoding.

Problem

Multi-modal diffusion language models suffer from visual hallucinations because their visual perception is static; unlike autoregressive models, they lack sequential dependency for dynamically attending to visual input.

Approach

Uses fluctuations in model confidence during the iterative diffusion decoding process as a signal to detect hallucinated tokens and correct them. Details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add VGR (fine-tuning plus refinement loop)LLaDA-V · CapArenaCapArena-Auto: -77.17 -> -52.17Improved
add VGRLLaDA-V · AMBER-gHal: 44.9 -> 37.2Improved
add VGRLLaDA-V · DetailCaps-4870CAPTURE: 59.62 -> 62.35Improved
add VGRLLaDA-V · CapMASAvg. score: 58.61 -> 60.84Improved
add VGR-Instruct fine-tuning (ablation)LLaDA-V · CapArenaCapArena-Auto: -77.17 -> -61.33Improved
add the inference-time refinement loop after fine-tuning (ablation)LLaDA-V (SFT) · AMBER-gCover: 63.7 -> 64.5Improved
localize with high CF instead of low confidence (ablation)VGR refinement · CapArenaCapArena-Auto: -57.83 -> -52.17Improved
combine high CF and low confidence in a hybrid (ablation)VGR refinement · CapArenaCapArena-Auto: -52.17 -> -56.50Hurt
add local view to global view (dual-view) (ablation)VGR refinement · CapArenaCapArena-Auto: -53.67 -> -52.17Improved
use syntax-aware span remasking instead of fixed window (ablation)VGR refinement · CapArenaCapArena-Auto: -59.50 -> -52.17Improved
raise visual gate threshold from 0.4 to 0.6 (ablation)VGR refinement · AMBERCorrection Rate: 97.4% -> 93.3%Hurt
apply VGR to MMaDAMMaDA · AMBER-gHal: 72.8 -> 39.6Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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