AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Dive into the implicit biases of low-rank vision-language alignment

Mingjia Shi, Shuo Wang, Xiaobo Wang, Sifan Zhou, Kai Wang, Tianyu Fu, Chenxu Zhao, Anyang Su, Ping Jiang, Minghui Wu

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 0 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
use low-rank instead of full-parameter LLM updates in alignment (ablation)VLM alignment (1.4B-13B) · MME-perceptionMME-P: 236.6 -> 1079.7 (13B)Improved
use low-rank instead of full-parameter alignment (ablation)VLM alignment 13B · Training timehours: 35.32 -> 13.23Improved
use low-rank alignment on Qwen3-8B (ablation)Qwen3-8B · MMBenchaccuracy: 18.1 -> 48.7Improved
use low-rank alignment on Qwen3-8B (ablation)Qwen3-8B · POPEPOPE: 92.7 -> 25.1Hurt
use low-rank alignment on Qwen3-8B (ablation)Qwen3-8B · MMMU-Proaccuracy: 12.1 -> 5.7Hurt
use LoHa instead of LoRA as low-rank operator (ablation)low-rank alignment · MME-perception (Vicuna-13B)MME-P: 375.9 -> 1079.7Improved
use LoKr at 13B instead of LoHa (ablation)low-rank alignment · Table 5 13B POPEPOPE: 67.4 -> 37.6Hurt
freeze vision encoder during alignment (ablation)LoRA alignment (CLIP) · GQAaccuracy: 60.18 -> 62.47Improved
apply LoRA to both attention and MLP instead of one (ablation)LoRA alignment · MME-PMME-P: 1384.71 -> 1511.61 (vs MLP-only)Improved
use low-rank instead of full with tuned weight decay (ablation)alignment · six benchmarksaccuracy: 1432.68 -> 1511.61 (MME-P)Improved
use low-rank alignment with DINOv2 or SigLIPv2 encoder (ablation)alignment · MME-P / GQAMME-P: 1503.69 -> 1481.45 (DINOv2 MME-P); 1406.61 -> 1474.70 (SigLIPv2)Mixed
increase LoRA rank (ablation)low-rank alignment · knowledge benchmarks (GQA, SQA, TextVQA)accuracyNo 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