AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

What Makes Good Examples for Visual In-Context Learning?

Yuanhan Zhang, Kaiyang Zhou, Ziwei Liu

NeurIPS 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Human-like visual cognition in VLMs · 18 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
Replace random prompt selection with unsupervised nearest-example retrievalVisual in-context learning · Pascal-5ᶦ segmentation AvgmIoU: 27.56 -> 33.56Improved
Replace random prompt selection with supervised prompt retrievalVisual in-context learning · Pascal-5ᶦ segmentation AvgmIoU: 27.56 -> 35.56Improved
Replace random prompt selection with unsupervised nearest-example retrievalVisual in-context learning · Pascal VOC single object detectionmIoU: 25.45 -> 26.84Improved
Replace random prompt selection with supervised prompt retrievalVisual in-context learning · Pascal VOC single object detectionmIoU: 25.45 -> 28.22Improved
Replace random prompt selection with prompt retrievalVisual in-context learning · ImageNet-2012 colorizationmse: 0.67 -> 0.63Improved
Use supervised rather than unsupervised prompt retrievalVisual in-context learning · Pascal-5ᶦ segmentation AvgmIoU: 33.56 -> 35.56Improved
Use supervised rather than unsupervised prompt retrievalVisual in-context learning · MSCOCO-5ᶦ segmentation Avg (Pascal to MSCOCO shift)mIoU: 18.02 -> 19.95Improved
Use a larger retrieval set (ablation)Prompt retrieval methods · Pascal-5ᶦ segmentationmIoUImproved
Increase the number of in-context examples (ablation)Random, UnsupPR, and SupPR · Pascal-5ᶦ segmentationmIoUImproved
Change the order of the in-context examples (ablation)Visual in-context learning · Pascal-5ᶦ segmentationmIoUNo clear effect
Replace cosine distance with Euclidean or Manhattan distance (ablation)Prompt retrieval score function · Pascal-5ᶦ segmentationmIoUNo clear effect
Change the retrieval backbone among CLIP, EVA, and ViT (ablation)UnsupPR and SupPR · Pascal-5ᶦ segmentation AvgmIoU: 33.56; 33.64; 33.83; 35.56; 35.71; 36.12No 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