billpsomas/efficient-probing

This repo contains the official implementation of the ICLR 2026 paper "Attention, Please! Revisiting Attentive Probing Through the Lens of Efficiency"

32
/ 100
Emerging

This project helps machine learning researchers efficiently evaluate the performance of frozen pre-trained image recognition models on new datasets. It takes a pre-trained model and image data as input and produces a performance score along with visual attention maps. This is primarily for researchers or ML engineers who need to quickly assess how well a pre-trained vision model understands specific visual concepts without the lengthy process of full fine-tuning.

Use this if you need to quickly and efficiently evaluate the classification capability of a frozen pre-trained vision encoder on various image datasets and gain insights into its decision-making through interpretable attention maps.

Not ideal if you are looking to fully fine-tune a model for maximum performance on a specific task or if your primary goal is to train a model from scratch.

machine-learning-research computer-vision model-evaluation image-classification deep-learning
No Package No Dependents
Maintenance 10 / 25
Adoption 7 / 25
Maturity 15 / 25
Community 0 / 25

How are scores calculated?

Stars

29

Forks

Language

Python

License

Apache-2.0

Last pushed

Feb 23, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/billpsomas/efficient-probing"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.