---
id: 20260401-T0-11
title: "视觉演示选择新方法提升MLLM性能"
title_en: "New Method Improves MLLM Performance via Visual Demo Selection"
url: https://ai.daily.yangsir.net/daily/20260401-T0-11
issue_date: 2026-04-01
publish_date: 2026-03-31T04:00:00.000Z
category: research
source_name: "arXiv cs.LG (ML)"
source_url: https://arxiv.org/abs/2603.26775
---

# 视觉演示选择新方法提升MLLM性能

针对多模态大模型的视觉上下文学习演示质量问题，arXiv论文提出演示选择新方法。研究指出当前主流的kNN搜索策略效率低下，新方法通过优化演示选择，显著提升了MLLM在视觉任务中的适应能力。

## English Version

**New Method Improves MLLM Performance via Visual Demo Selection**

An arXiv paper addresses poor demonstration selection in multimodal LLMs by introducing a new method to replace inefficient kNN search. The approach enhances MLLM adaptability for visual tasks through optimized demonstration selection.

---

**来源**：[arXiv cs.LG (ML)](https://arxiv.org/abs/2603.26775)

**详情页**：https://ai.daily.yangsir.net/daily/20260401-T0-11

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*