AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Robust Vision-Language Models via Manifold-Adversarial Adapters

Hao Li, Zeyu Xiao, Junhao Zhou, Peng Liu, Yang Zhao, Wei Jia

ICML 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis

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 a manifold-adversarial adapter approach for making vision-language models robust to real-world image corruptions via parameter-efficient adaptation.

Problem

Vision-language models degrade under real-world image corruptions, affecting both visual recognition and language-grounded reasoning; Euclidean feature alignment of corrupted and clean representations is argued to be insufficient.

Approach

Parameter-efficient adapters that align corrupted features with clean references using a manifold-aware (rather than purely Euclidean) alignment objective, combined with adversarial training, as an alternative to cascaded image restoration.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add adversarial manifold constraint to feature distillation (ablation)MAA adapters · R-Bench Disscore: 58.70 -> 59.39Improved
naive MSE-only adapter adaptation vs zero-shot (ablation)LLaVA-1.6-Mistral-7B · R-Bench Disscore: 58.79 -> 57.98Hurt
add adversarial constraint to full fine-tuning (ablation)feature distillation · R-Bench Disscore: 32.93 -> 34.34No clear effect
add adversarial constraint to LoRA (ablation)feature distillation · R-Bench Disscore: 56.77 -> 57.98Improved
add adversarial constraint to prefix tuning (ablation)feature distillation · R-Bench Disscore: 55.35 -> 58.99Improved
add window self-attention branch (ablation)MAA · R-Bench Disscore: 58.79 -> 59.39Improved
add max pooling branch (ablation)MAA · R-Bench Disscore: 59.39 -> 59.39No clear effect
train on hybrid multi-source data instead of single source (ablation)MAA · R-Bench Disscore: 58.99 -> 59.39Improved
prepend Real-ESRGAN restoration before VLMLLaVA-1.6-Mistral-7B · MMVetscore: 42.33 -> 38.12Hurt
prepend external image restoration models before VLMLLaVA-1.6-Mistral-7B · R-Bench Disscore: 58.79 -> 56.97 to 58.38Hurt
use InfoNCE as feature objective (ablation)MAA adapter · R-Bench Disscore: 58.79 -> 0.00Hurt
apply MAA to InternVL3.5-30B-A3BInternVL3.5-30B-A3B · R-Bench Disscore: 71.51 -> 72.93Improved

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