AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

Is Your (Reasoning) Multimodal Language Model Vulnerable Toward Distractions?

Ming Liu, Hao Chen, Jindong Wang, Liwen Wang, Jingchen Sun, Wensheng Zhang

AAAI 2026 · Vision-language and multimodal › Visual question answering · approach: Multimodal LLM question answering · 1 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
insert irrelevant hint text (Add Hint)InternVL2 (2B) · I-ScienceQAexact match: 91.40 -> 82.35Hurt
insert irrelevant hint textVL-Rethinker (7B) · I-ScienceQA Add Hintsexact match: 88.05 -> 82.30Hurt
add an unrelated imageGPT-4o · I-ScienceQA Add Imageexact match: 93.50 -> 93.00No clear effect
add an unrelated imageGPT-o4 mini · I-ScienceQA Add Imageexact match: 98.15 -> 98.75No clear effect
insert a distracting image alongside the originalVL-Rethinker (7B) · I-ScienceQA Insert Imageexact match: 82.75 -> 88.90Improved
insert a distracting image alongside the originalInternVL2 (1B) · I-ScienceQA Add Imageexact match: 85.60 -> 79.70Hurt
use textual rather than visual distractionsVLMs in general · I-ScienceQAexact match dropHurt
scale InternVL2 from 8B to 26BInternVL2 · I-ScienceQA Add Imageexact match: 94.45 -> 93.40No clear effect
scale LLaVA from 7B to 34BLLaVA · I-ScienceQA Add Hintexact match: 63.80 -> 82.65Improved
add prompt telling model to ignore irrelevant information (ablation)Qwen2-VL (2B) · I-ScienceQA Add Imageexact match: 63.30 -> 73.80Improved
add prompt telling model to ignore irrelevant information (ablation)Qwen2-VL (7B) · I-ScienceQA Add Imageexact match: 83.10 -> 81.35Hurt
swap CLIP for robust-CLIP vision encoder (ablation)LLaVA-7B · I-ScienceQAexact match: +0.85 on Add ImageNo clear effect

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