Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Beyond Model Size: Probing the Gaps in Visual in-Context Learning by Training a Tiny Model

About

Visual in-Context Learning (VICL) aims at making progress towards adaptive vision models, that can -- based on a few examples -- adapt to a new task at test-time. With the history of in-context learning in natural language processing research, where large, parameter-heavy models are in use, one pathway that current VICL methods take is model- and data-scaling as key ingredients. Yet, it is not clear, whether these ingredients are the key for in-context learning to take shape in vision models. To stress-test such large models, we challenge them with an extreme counterexample: we train a tiny visual in-context model with merely $1$ million parameters and a modest amount of $70,000$ images. We compare the results of this severely capacity capped tiny model to $7,000\times$ larger VICL models in different adaptive settings, (1) on image data with small distribution shifts, (2) on unseen task encodings and (3) on a completely new task, i.e., the setting VICL envisions. With the chasm of training resources between the tiny- and large models, our experiments showcase a lack in how adaptive capabilities are measured, with respect to how tasks are encoded, which tasks were used in pre-training and the choice of metrics. These gaps in current VICL benchmarking underscore a need for innovation in evaluation of adaptive capabilities.

Sunil Khatri, Steven Landgraf, Markus Ulrich, Simon Rei{\ss}• 2026

Related benchmarks

TaskDatasetResultRank
Depth EstimationNYU Depth V2
RMSE0.5526
226
Edge DetectionBSDS500
Mean Precision0.2587
15
De-rainingDID-MDN
LPIPS0.1247
7
Image ColorizationImageNet
LPIPS0.2672
7
Semantic segmentationPascal VOC (split 0)
mIoU13.19
7
Semantic segmentationPASCAL VOC Split 1
mIoU16.69
7
Semantic segmentationPASCAL VOC Split 3
mIoU13.5
7
Semantic segmentationPASCAL VOC Split 2
mIoU12.57
7
Showing 8 of 8 rows

Other info

Follow for update