AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Adaptive Hallucination Alleviation in Multimodal Large Language Models: From Strategic Data Selection to Severity-Guided Training

Yuanyi Xu, Xiangru Zhu, Sihang Jiang, Zhixu Li, Bei Yang, Xiaoxiao Xu, Yanghua Xiao, Wei Wang

AAAI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Benchmarks and analysis of MLLMs · 0 citations (OpenAlex)

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.

Proposes an adaptive framework for alleviating hallucination in multimodal LLMs that strategically selects hallucination-related training samples and applies severity-guided training to them.

Problem

MLLMs suffer from multiple forms of hallucination, but prior methods struggle to efficiently build high-quality hallucination-related samples and to process them at fine granularity.

Approach

Combines strategic data selection for constructing hallucination-related samples with a severity-guided training scheme that treats samples according to their hallucination severity rather than uniformly.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add image selection (CLIP-S and PPL under noise) (ablation)DPO · AMBER discriminativeaccuracy: 77.6 -> 78.4Improved
add image selection (ablation)DPO · HallusionBenchaAcc: 43.72 -> 44.29Improved
add explicit severity weighting in the DPO loss (ablation)DPO · MMHal-Benchscore: 2.37 -> 2.51Improved
add explicit severity weighting in the DPO loss (ablation)DPO · AMBER generativeCHAIR: 2.5 -> 2.1Improved
add severity-weighted visual attention reallocation (ablation)HD-DPO (explicit) · MMHal-Benchscore: 2.51 -> 2.60Improved
add severity-weighted visual attention reallocation (ablation)HD-DPO (explicit) · HallusionBenchaAcc: 44.64 -> 45.46Improved
add severity-weighted visual attention reallocation (ablation)HD-DPO (explicit) · MMBenchscore: 64.3 -> 64.0Hurt
train on Object-type hallucinations only (vs Secondary types) (ablation)HD-DPO · AMBER/MMHal/CHAIRCHAIR_S: 47.3 -> 36.7 (Object); Action 43.0Improved
apply HD-DPO to other MLLMsQwen-VL / InstructBLIP / InternLM-XComposer2 · AMBER, CHAIRCHAIR_S: Qwen-VL 39.7 -> 10.3; InstructBLIP 21.0 -> 4.2Improved
apply HD-DPO instead of V-DPOV-DPO · HallusionBenchaAcc: 46.20 -> 45.46Hurt
apply HD-DPO instead of CSRCSR · MMBenchscore: 64.1 -> 64.0Hurt
apply HD-DPO to LLaVA-1.5-7BLLaVA-1.5-7B · AMBER discriminativeaccuracy: 72.3 -> 79.7Improved

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