# Can this Model Also Recognize Dogs? Zero-Shot Model Search from Weights

arXiv: https://arxiv.org/abs/2502.09619 · HTML: https://mlsanity.com/paper/2502.09619/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 13 Feb 2025.

## Built on by (most cited first, 5 of 5)

- [We Should Chart an Atlas of All the World's Models](https://mlsanity.com/paper/2503.10633/) (2025)
- [W2T: LoRA Weights Already Know What They Can Do](https://mlsanity.com/paper/2603.15990/) (2026)
- [Learning Model Representations Using Publicly Available Model Hubs](https://mlsanity.com/paper/2510.02096/) (2025)
- [Discovering Hidden Gems in Model Repositories](https://mlsanity.com/paper/2601.22157/) (2026)
- [Delta Activations: A Representation for Finetuned Large Language Models](https://mlsanity.com/paper/2509.04442/) (2025)

## Abstract (arXiv)

With the increasing numbers of publicly available models, there are probably pretrained, online models for most tasks users require. However, current model search methods are rudimentary, essentially a text-based search in the documentation, thus users cannot find the relevant models. This paper presents ProbeLog, a method for retrieving classification models that can recognize a target concept, such as "Dog", without access to model metadata or training data. Differently from previous probing methods, ProbeLog computes a descriptor for each output dimension (logit) of each model, by observing its responses on a fixed set of inputs (probes). Our method supports both logit-based retrieval ("find more logits like this") and zero-shot, text-based retrieval ("find all logits corresponding to dogs"). As probing-based representations require multiple costly feedforward passes through the model, we develop a method, based on collaborative filtering, that reduces the cost of encoding repositories by 3x. We demonstrate that ProbeLog achieves high retrieval accuracy, both in real-world and fine-grained search tasks and is scalable to full-size repositories.