AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

AdViP: Aligning Multi-modal LLMs via Adaptive Vision-enhanced Preference Optimization

Jinda Lu, Jinghan Li, Yuan Gao, Junkang Wu, Jiancan Wu, Xiang Wang, Xiangnan He

IJCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Bias-aware multi-image robustness · 8 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
add vision-based preference with equal weight to DPO (ablation)LLaVA-1.5 7B DPO · AMBER discriminativeaccuracy: 64.8 -> 51.8Hurt
add vision-based preference with equal weight to DPO (ablation)LLaVA-1.5 7B DPO · AMBER discriminativeF1: 78.8 -> 72.6Hurt
replace equal weighting with the adaptive strategy (ablation)LLaVA-1.5 7B · AMBER discriminativeF1: 72.6 -> 85.8Improved
replace equal weighting with the adaptive strategy (ablation)LLaVA-1.5 7B · AMBER discriminativeaccuracy: 51.8 -> 79.9Improved
replace equal weighting with the adaptive strategy (ablation)LLaVA-1.5 7B · AMBER discriminativeF1 relation: 47.4 -> 67.0Improved
apply AdaViP instead of vanilla DPOLLaVA-1.5 7B · Object HalBenchnon-hallucination rate (response): 80.1 -> 93.7Improved
apply AdaViP instead of vanilla DPOLLaVA-1.5 13B · Object HalBenchnon-hallucination rate (response): 82.3 -> 95.4Improved
apply AdaViP instead of vanilla DPOLLaVA-1.5 7B · AMBER generativeCHAIR: 3.5 -> 1.8Improved
apply AdaViP instead of vanilla DPOLLaVA-1.5 7B · MMHal-Benchscore: 2.00 -> 3.12Improved
apply AdaViP instead of vanilla DPOLLaVA-1.5 13B · MMHal-Benchhallucination rate: 30.0 -> 34.0Hurt
apply AdaViP instead of vanilla DPOLLaVA-1.5 7B · AMBER discriminativeaccuracy: 64.8 -> 79.9Improved

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