Pith. sign in

REVIEW 5 major objections 4 minor 68 references

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

T0 review · 5 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read This paper introduces ProbeLog, which represents each output dimension of a classifier by its responses to fixed probe images, enabling both logit-based and zero-shot text-based model search using only weights.

desk verdict ProbeLog is a fresh, useful take on model search with some evaluation gaps; the zero-shot text alignment needs harder evidence, but the paper deserves a real review. read the letter →

arxiv 2502.09619 v1 pith:GFSOAZW7 submitted 2025-02-13 cs.LG cs.CV

classification cs.LGcs.CV MSC 68T0768P2068T45
keywords modelsearchzoosprobinglogitdescriptorszero-shotretrievalcollaborativefilteringweight-spacelearningCLIP
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper introduces ProbeLog, a method for searching large repositories of classification models by what the models actually do, using only their weights. The claim is that every output dimension (logit) of a classifier can be described by its pattern of responses to a fixed set of probe images, and that these logit descriptors can be matched both to other logits and to a text prompt such as "Dog" without any metadata or training data. A user could therefore find every publicly available model that recognizes a target concept even when the models are undocumented. On the paper's curated hubs, ProbeLog reaches 72.8% top-1 retrieval accuracy when searching in-distribution, and 43.8% top-1 accuracy when searching a 1,000-class hub from a text prompt, against a random baseline of 0.1%. The authors also show that probing cost can be cut by about three times using collaborative filtering to impute unprobed entries.

What carries the argument

The central objects are ProbeLog descriptors defined per logit (Eq. 2), mean-standard-deviation normalization (Eq. 5), a top-$k$ asymmetric discrepancy (Eq. 3), and zero-shot text descriptors built from CLIP dot products (Eq. 4). Together they turn model search into nearest-neighbor search over logit-level functional signatures, with no access to weights beyond forward passes. Collaborative probing (Eq. 6) completes missing probe entries by low-rank matrix factorization, making the gallery construction tractable.

What would settle it

Run the exact text-retrieval pipeline on a held-out repository of classifiers trained on concept sets not overlapping the probe distribution (for instance medical image classes) while keeping COCO probes; if top-1 accuracy does not stay far above the random baseline after normalization, the claimed shared metric space for zero-shot search is refuted.

Watch

Extended reading notes

Core claim

ProbeLog represents each logit $i$ of a model $f$ by the vector $\varphi(f,i) = [f(x_1)[i], f(x_2)[i], \ldots, f(x_n)[i]]$ of its outputs on $n$ fixed ordered probe images. The paper shows that after subtracting each descriptor's mean and dividing by its standard deviation, these vectors form a usable semantic index that is invariant to output-class order and to the other classes the model sees. Retrieval uses an asymmetric discrepancy: for a query descriptor, only the $k$ probe positions where the query logit responds most strongly are compared against gallery descriptors in $L_2$. For zero-shot text search, the probe images and the target text are embedded with a joint text-image model such as CLIP; the text descriptor is the vector of dot products between the text embedding and each probe embedding, and the same normalization makes it comparable to real logit descriptors. A final component completes partially probed galleries via truncated-SVD matrix factorization, so repositories can be encoded with a small fraction of probes per model.

Load-bearing premise

The load-bearing premise is that after subtracting the mean and dividing by the standard deviation, CLIP text-probe similarities and raw logit responses occupy the same metric space; the paper's own ablation shows that removing this normalization drops text-to-INet top-1 accuracy from 43.8% to 0%.

Editorial extensions

If this is right

  • Model repositories can be searched by capability even when model cards are empty or templates.
  • Users can issue a text query for a concept never seen during indexing and retrieve models that recognize it.
  • The per-logit representation is naturally invariant to class ordering, so identical models with permuted outputs do not confuse retrieval.
  • Gallery construction cost scales down materially: roughly one third of the probes per model suffices when combined with matrix-factorization imputation.
  • The representation is lightweight relative to model weights, so a large repository's descriptors can be stored and searched cheaply.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the method indexes functions rather than architectures, it could plausibly extend to matching adapters, LoRA modules, or generative model concepts through activation or attention probes, though the paper only claims classification logits.
  • The dependence of text retrieval on z-score normalization suggests an empirical correspondence between CLIP's semantic geometry and logit confidence geometry; testing this correspondence across more text models could either broaden or constrain the zero-shot claim.
  • A testable design improvement the paper leaves implicit: choosing probes adaptively per concept (coreset selection) rather than uniformly at random could reduce the 4,000-probe requirement while keeping accuracy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 4 minor

Summary. This paper introduces ProbeLog, a method for retrieving classification models by semantic concept using only model weights. For each logit, it records responses to a fixed ordered set of probe images, normalizes the resulting vector (Eq. 5), and compares descriptors with an asymmetric discrepancy that focuses on the query's highest-response probes (Eq. 3). A zero-shot variant computes a probe-level text descriptor from CLIP similarities between probe images and the query concept (Eq. 4), normalizes it like logit descriptors, and applies the same discrepancy measure. Collaborative probing (Eq. 6) uses truncated-SVD imputation to reduce the number of forward passes needed to encode a repository. The method is evaluated on two new hubs, INet-Hub (1,500 models trained on ImageNet subsets) and HF-Hub (71 real Hugging Face models), in both logit-to-logit and text-to-logit retrieval settings, with reported top-1 accuracies up to 72.8% and 43.8%, respectively.

Significance. The problem is timely and important: model hubs are growing quickly, and searching by weights rather than documentation could be practically valuable. The logit-level descriptor is a sensible functional representation that addresses class-order permutation, and collaborative probing is an interesting efficiency idea. If the zero-shot text-to-logit alignment is robust, the paper would make a useful contribution to model retrieval. However, the current evidence for the central text-search claim is incomplete, and several implementation details needed for reproduction are missing. The strengths are the clean problem formulation and the new evaluation hubs, but the quantitative claims need additional support before the paper can be accepted.

major comments (5)
  1. [Sec. 4.3, Eq. (5), Table 3] This is the load-bearing assumption for the zero-shot contribution; if the revision does not address it, the text-search claim is not supported.
  2. [Sec. 4.2, Eq. (3); Sec. 4.4, Eq. (6)] I also ask for a sensitivity curve for k, since the ablation in Table 3 only compares top-k against other selection rules, not against different k values.
  3. [App. A; App. B; Sec. 5.2] The authors should also state how many random seeds or probe-set draws the reported standard deviations correspond to.
  4. [Sec. 5.2, Tables 1 and 4] The model-level and full-query baselines are useful, but they do not address the paper's stated alternative of metadata-based search.
  5. [Sec. 4.4, Fig. 7] Please also clarify whether the 'same accuracy' comparison in Fig. 7 uses the same k and normalization as the full-probe experiments.
minor comments (4)
  1. [Sec. 4.4, Eq. (6)] The sentence defining the mask matrix says 'all ones expect for zeros'; it should read 'all ones except for zeros'.
  2. [Sec. 4.3] It should be stated explicitly that Eq. (5) is applied to both the logit descriptor and the zero-shot text descriptor; the current notation only shows the logit descriptor, which is a source of confusion.
  3. [Sec. 5.2] There are typographical issues: 'over40%' should be 'over 40%' and 'IN ettask' should be 'INet task'.
  4. [Sec. 5.4, Table 2] The conclusion that 'even out-of-distribution probes sampled from COCO retrieve relevant logits with high accuracy' would be easier to evaluate if the table also reported the mean number of relevant logits per query for each task, so the reader can interpret top-k accuracy against chance.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation-level circularity: ProbeLog descriptors are measured logit responses, and the zero-shot text comparison rests on an empirical CLIP-alignment assumption rather than on an equation that equates inputs with outputs.

full rationale

The derivation chain is self-contained. Logit-level descriptors are defined directly as observed model responses to fixed probes (Eq. 2), not derived from the retrieval target or from the evaluation labels. The discrepancy measure (Eq. 3) is an asymmetric comparison rule over the query's highest-response probes; it is a heuristic design choice, not a definitional encoding of class identity. The zero-shot text descriptor (Eq. 4) is constructed from CLIP image-text similarities, an externally trained representation, and Eq. 5 is a per-vector standardization that is applied identically to text and logit descriptors; no parameter is fitted to the retrieval labels, and no subset of the ground-truth data is used to tune the method's core mapping. The ablations in Tables 2 and 3 select probe distributions, normalization, and top-k selection by observing performance on the same benchmark sets, which is a validation-and-overfitting concern rather than a circularity: the reported numbers are genuine outputs of the fixed procedure, not quantities forced by construction. Self-citations appear only as related-work pointers to the authors' earlier probing and weight-space papers; none is invoked as a load-bearing uniqueness theorem or as justification for the central retrieval claim. Collaborative Probing is an efficiency technique whose matrix factorization is fit to sparse probe responses, and it is evaluated against full probing rather than used to define the retrieval objective. The limitations section honestly flags out-of-distribution concepts and the extension beyond classifiers, which are scope limitations, not admissions of circularity. Overall, no prediction reduces by definition to its inputs, and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 4 free parameters · 4 assumptions · 2 invented entities

The central claim relies on the assumption that a fixed probe set elicits semantic fingerprints from arbitrary classifiers, and that CLIP text vectors can be bridged to logit vectors by simple normalization. The top-k parameter and probe distribution are tuned on the evaluation tasks. No new physical entities are introduced; the ProbeLog descriptors are representational constructs.

free parameters (4)
  • k (number of top probes in discrepancy metric) = not reported
    Sec. 4.2 defines the discrepancy using top k probe entries; Sec. 5.4 ablates top-k vs alternatives but never states the value of k used in Tables 1 and 3.
  • Number of probes n = 4,000 (8,000 in some ablations)
    Sec. 5.4 shows accuracy grows with n; 4,000 used for main results, 8,000 gives higher accuracy (47.8% top-1 text-to-INet).
  • Probe distribution = COCO images
    Chosen after ablations in Sec. 5.4 showing ImageNet and Stable Diffusion probes work better; COCO is used for generality.
  • Collaborative probing sampling fraction and SVD rank = Not fully specified
    Sec. 5.3 uses p% per model, with 4% giving good results; SVD rank and iterations are not reported.
assumptions (4)
  • domain assumption A logit's responses to a fixed set of probe inputs are a meaningful descriptor of the concept it recognizes.
    Core assumption underlying Eq. 2; Figure 4 provides empirical support but no theoretical justification.
  • ad hoc to paper After mean/std normalization, CLIP text similarities and raw logit responses share a common metric space.
    Sec. 4.3 introduces normalization (Eq. 5) as a scaling fix; it is load-bearing because Table 3 shows near-zero accuracy without it.
  • domain assumption The probe set can be fixed and reused across all models in the repository.
    Required by the matrix-completion framing in Sec. 4.4 and by the definition of descriptors.
  • domain assumption The models are accessible for arbitrary forward passes with the chosen probes.
    Implicit throughout; the method does not address models that restrict inference or require special pre-processing.
invented entities (2)
  • ProbeLog descriptor
    purpose: Per-logit functional representation used for retrieval
    A new representation introduced by the paper; it is the method itself, not a falsifiable physical entity.
  • Zero-shot ProbeLog descriptor
    purpose: Text-based query descriptor built from CLIP similarities
    Defined in Eq. 4 as the vector of dot products between probe embeddings and target text embedding; relies on CLIP alignment.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Can this Model Also Recognize Dogs? Zero-Shot Model Search from Weights." pith.science (2026). https://pith.science/paper/GFSOAZW7

@misc{pith2026250209619,
  author       = {Pith},
  title        = {Pith review of: Can this Model Also Recognize Dogs? Zero-Shot Model Search from Weights},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GFSOAZW7}},
  note         = {Machine review of arXiv:2502.09619}
}
read the original abstract

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.

Figures

Figures reproduced from arXiv: 2502.09619 by the authors.

Figure 1
Figure 1. Hugging Face Documentation. We analyze the model cards of 1.2M Hugging Face models. We discover that the major￾ity of models are either undocumented or poorly documented. for their specific task and use it directly without additional training. With the rise of large public model repositories, this is becoming feasible. For instance, Hugging Face, the largest existing model repository, hosts over 1 million mod￾els, w… view at source ↗
Figure 2
Figure 2. Classification Model Search. We present a new task of Classification Model Search, where the goal is to find classifiers that can recognize a target concept. Concretely, given an input prompt, such as “Dog”, we wish to retrieve all classifiers that one of their classes is “Dog”. The search space is a large model repository, that contains many models and concepts to search from. The retrieved models can replace model… view at source ↗
Figure 3
Figure 3. ProbeLog Descriptors. Our method generates a descriptor for individual output dimensions (logits) of models. First, we sample and a set of inputs (e.g., from the COCO dataset), and fix them as our set of probes. Then, to create a new ProbeLog descriptor for a model logit, we feed the set of ordered probes nto the model and observe their outputs. Finally, we take all values of the logit we wish to represent, and norm… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: CIFAR10 Logit Similarities.(a) Ground truth label. (b) ProbeLog representations using 1, 000 out-of-distribution COCO image probes. (c) ProbeLog representations using 1, 000 in￾distribution CIFAR10 image probes. Both find meaningful similar￾ities, although in-distribut…
Figure 5
Figure 5. Figure 5: Text-Aligned ProbeLog Representation. We present a method to create ProbeLog-like representations for text prompts. We encode and store each of our ordered probes using the CLIP image encoder. At inference time, we embed the target text prompt, and compute its similari…
Figure 6
Figure 6. Figure 6: Collaborative Probing. We pass a random subset of probes through each model in the repository to obtain partial logit representations. By performing factorization based matrix imputation we can complete the missing information. This saves a substantial part of the comp…
Figure 7
Figure 7. Figure 7: Collaborative Probing. We test our method using col￾laborative probing on the text → INet-Hub retrieval task. While the full size of the dataset is 8, 000 COCO probes, we show cases where each model is probed by less than 15% of these probes. We can see that for the li…
Figure 8
Figure 8. Figure 8: Number of Probes. We test our zero-shot retrieval ap￾proach on INet-Hub with increasing numbers of probes. While more probes lead to higher accuracy, the gains are diminishing. Scaling-up to entire repositories. While our model hubs already have up to 1, 500 large mode…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 30 canonical work pages

  1. [1]

    NeRN -- Learning Neural Representations for Neural Networks

    Ashkenazi, M., Rimon, Z., Vainshtein, R., Levi, S., Richardson, E., Mintz, P., and Treister, E. Nern--learning neural representations for neural networks. arXiv preprint arXiv:2212.13554, 2022

  2. [2]

    Learning to see by looking at noise

    Baradad Jurjo, M., Wulff, J., Wang, T., Isola, P., and Torralba, A. Learning to see by looking at noise. Advances in Neural Information Processing Systems, 34: 0 2556--2569, 2021

  3. [3]

    Network dissection: Quantifying interpretability of deep visual representations

    Bau, D., Zhou, B., Khosla, A., Oliva, A., and Torralba, A. Network dissection: Quantifying interpretability of deep visual representations. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 6541--6549, 2017

  4. [4]

    D., Steinke, T., Hayase, J., Cooper, A

    Carlini, N., Paleka, D., Dvijotham, K. D., Steinke, T., Hayase, J., Cooper, A. F., Lee, K., Jagielski, M., Nasr, M., Conmy, A., et al. Stealing part of a production language model. arXiv preprint arXiv:2403.06634, 2024

  5. [5]

    Spann: Highly-efficient billion-scale approximate nearest neighborhood search

    Chen, Q., Zhao, B., Wang, H., Li, M., Liu, C., Li, Z., Yang, M., and Wang, J. Spann: Highly-efficient billion-scale approximate nearest neighborhood search. Advances in Neural Information Processing Systems, 34: 0 5199--5212, 2021

  6. [6]

    Where to start? Analyzing the potential value of intermediate models

    Choshen, L., Venezian, E., Don-Yehia, S., Slonim, N., and Katz, Y. Where to start? analyzing the potential value of intermediate models. arXiv preprint arXiv:2211.00107, 2022

  7. [7]

    Z., Salti, S., and Di Stefano, L

    De Luigi, L., Cardace, A., Spezialetti, R., Ramirez, P. Z., Salti, S., and Di Stefano, L. Deep learning on implicit neural representations of shapes. arXiv preprint arXiv:2302.05438, 2023

  8. [8]

    Imagenet: A large-scale hierarchical image database

    Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pp.\ 248--255. Ieee, 2009

Show all 68 references
  1. [9]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Devlin, J. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

  2. [10]

    and Loynd, R

    Diao, C. and Loynd, R. Relational attention: Generalizing transformers for graph-structured tasks. arXiv preprint arXiv:2210.05062, 2022

  3. [11]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Dosovitskiy, A. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020

  4. [12]

    A., and Shocher, A

    Dravid, A., Gandelsman, Y., Efros, A. A., and Shocher, A. Rosetta neurons: Mining the common units in a model zoo. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 1934--1943, 2023

  5. [13]

    A., and Aberman, K

    Dravid, A., Gandelsman, Y., Wang, K.-C., Abdal, R., Wetzstein, G., Efros, A. A., and Aberman, K. Interpreting the weight space of customized diffusion models. arXiv preprint arXiv:2406.09413, 2024

  6. [14]

    From data to functa: Your data point is a function and you can treat it like one

    Dupont, E., Kim, H., Eslami, S., Rezende, D., and Rosenbaum, D. From data to functa: Your data point is a function and you can treat it like one. arXiv preprint arXiv:2201.12204, 2022

  7. [15]

    Classifying the classifier: dissecting the weight space of neural networks

    Eilertsen, G., J \"o nsson, D., Ropinski, T., Unger, J., and Ynnerman, A. Classifying the classifier: dissecting the weight space of neural networks. In ECAI 2020, pp.\ 1119--1126. IOS Press, 2020

  8. [16]

    Hyperdiffusion: Generating implicit neural fields with weight-space diffusion

    Erko c , Z., Ma, F., Shan, Q., Nie ner, M., and Dai, A. Hyperdiffusion: Generating implicit neural fields with weight-space diffusion. In Proceedings of the IEEE/CVF international conference on computer vision, pp.\ 14300--14310, 2023

  9. [17]

    S., Riley, P

    Gilmer, J., Schoenholz, S. S., Riley, P. F., Vinyals, O., and Dahl, G. E. Neural message passing for quantum chemistry. In International conference on machine learning, pp.\ 1263--1272. PMLR, 2017

  10. [18]

    Knowledge is a region in weight space for fine-tuned language models

    Gueta, A., Venezian, E., Raffel, C., Slonim, N., Katz, Y., and Choshen, L. Knowledge is a region in weight space for fine-tuned language models. arXiv preprint arXiv:2302.04863, 2023

  11. [19]

    Ha, D., Dai, A., and Le, Q. V. Hypernetworks. arXiv preprint arXiv:1609.09106, 2016

  12. [20]

    Deep residual learning for image recognition

    He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 770--778, 2016

  13. [21]

    Learning useful representations of recurrent neural network weight matrices

    Herrmann, V., Faccio, F., and Schmidhuber, J. Learning useful representations of recurrent neural network weight matrices. arXiv preprint arXiv:2403.11998, 2024

  14. [22]

    Representing model weights with language using tree experts

    Horwitz, E., Cavia, B., Kahana, J., and Hoshen, Y. Representing model weights with language using tree experts. arXiv preprint arXiv:2410.13569, 2024 a

  15. [23]

    Recovering the pre-fine-tuning weights of generative models

    Horwitz, E., Kahana, J., and Hoshen, Y. Recovering the pre-fine-tuning weights of generative models. In ICML, 2024 b . URL https://openreview.net/forum?id=761UxjOTHB

  16. [24]

    On the origin of llamas: Model tree heritage recovery

    Horwitz, E., Shul, A., and Hoshen, Y. On the origin of llamas: Model tree heritage recovery. arXiv preprint arXiv:2405.18432, 2024 c

  17. [25]

    Lg-cav: Train any concept activation vector with language guidance

    Huang, Q., Song, J., Xue, M., Zhang, H., Hu, B., Wang, H., Jiang, H., Wang, X., and Song, M. Lg-cav: Train any concept activation vector with language guidance. arXiv preprint arXiv:2410.10308, 2024

  18. [26]

    Izmailov, P., Podoprikhin, D., Garipov, T., Vetrov, D., and Wilson, A. G. Averaging weights leads to wider optima and better generalization. arXiv preprint arXiv:1803.05407, 2018

  19. [27]

    V., Krishnawamy, R., and Kadekodi, R

    Jayaram Subramanya, S., Devvrit, F., Simhadri, H. V., Krishnawamy, R., and Kadekodi, R. Diskann: Fast accurate billion-point nearest neighbor search on a single node. Advances in Neural Information Processing Systems, 32, 2019

  20. [28]

    Billion-scale similarity search with gpus

    Johnson, J., Douze, M., and J \'e gou, H. Billion-scale similarity search with gpus. IEEE Transactions on Big Data, 7 0 (3): 0 535--547, 2019

  21. [29]

    Deep linear probe generators for weight space learning

    Kahana, J., Horwitz, E., Shuval, I., and Hoshen, Y. Deep linear probe generators for weight space learning. arXiv preprint arXiv:2410.10811, 2024

  22. [30]

    Scale equivariant graph metanetworks

    Kalogeropoulos, I., Bouritsas, G., and Panagakis, Y. Scale equivariant graph metanetworks. arXiv preprint arXiv:2406.10685, 2024

  23. [31]

    Lightgbm: A highly efficient gradient boosting decision tree

    Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., and Liu, T.-Y. Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems, 30, 2017

  24. [32]

    Kipf, T. N. and Welling, M. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907, 2016

  25. [33]

    J., Gavves, E., Snoek, C

    Kofinas, M., Knyazev, B., Zhang, Y., Chen, Y., Burghouts, G. J., Gavves, E., Snoek, C. G., and Zhang, D. W. Graph neural networks for learning equivariant representations of neural networks. arXiv preprint arXiv:2403.12143, 2024

  26. [34]

    Matrix factorization techniques for recommender systems

    Koren, Y., Bell, R., and Volinsky, C. Matrix factorization techniques for recommender systems. Computer, 42 0 (8): 0 30--37, 2009

  27. [35]

    Learning multiple layers of features from tiny images

    Krizhevsky, A., Hinton, G., et al. Learning multiple layers of features from tiny images. 2009

  28. [36]

    B., Mumford, D., and Huang, J

    Lee, A. B., Mumford, D., and Huang, J. Occlusion models for natural images: A statistical study of a scale-invariant dead leaves model. International Journal of Computer Vision, 41: 0 35--59, 2001

  29. [37]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    Li, J., Li, D., Savarese, S., and Hoi, S. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, pp.\ 19730--19742. PMLR, 2023

  30. [38]

    T., Lorraine, J., and Lucas, J

    Lim, D., Maron, H., Law, M. T., Lorraine, J., and Lucas, J. Graph metanetworks for processing diverse neural architectures. arXiv preprint arXiv:2312.04501, 2023

  31. [39]

    Learning on loras: Gl-equivariant processing of low-rank weight spaces for large finetuned models

    Lim, D., Gelberg, Y., Jegelka, S., Maron, H., et al. Learning on loras: Gl-equivariant processing of low-rank weight spaces for large finetuned models. arXiv preprint arXiv:2410.04207, 2024

  32. [40]

    Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Doll \'a r, P., and Zitnick, C. L. Microsoft coco: Common objects in context. In Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, pp....

  33. [41]

    A convnet for the 2020s

    Liu, Z., Mao, H., Wu, C.-Y., Feichtenhofer, C., Darrell, T., and Xie, S. A convnet for the 2020s. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 11976--11986, 2022

  34. [42]

    Content-based search for deep generative models

    Lu, D., Wang, S.-Y., Kumari, N., Agarwal, R., Tang, M., Bau, D., and Zhu, J.-Y. Content-based search for deep generative models. In SIGGRAPH Asia 2023 Conference Papers, pp.\ 1--12, 2023

  35. [43]

    E., Chen, Z., Salakhutdinov, R., and Stoica, I

    Luo, M., Wong, J., Trabucco, B., Huang, Y., Gonzalez, J. E., Chen, Z., Salakhutdinov, R., and Stoica, I. Stylus: Automatic adapter selection for diffusion models. arXiv preprint arXiv:2404.18928, 2024

  36. [44]

    Equivariant architectures for learning in deep weight spaces

    Navon, A., Shamsian, A., Achituve, I., Fetaya, E., Chechik, G., and Maron, H. Equivariant architectures for learning in deep weight spaces. In International Conference on Machine Learning, pp.\ 25790--25816. PMLR, 2023

  37. [45]

    A., and Malik, J

    Peebles, W., Radosavovic, I., Brooks, T., Efros, A. A., and Malik, J. Learning to learn with generative models of neural network checkpoints. arXiv preprint arXiv:2209.12892, 2022

  38. [46]

    W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al

    Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pp.\ 8748--8763. PMLR, 2021

  39. [47]

    P., Girshick, R., He, K., and Doll \'a r, P

    Radosavovic, I., Kosaraju, R. P., Girshick, R., He, K., and Doll \'a r, P. Designing network design spaces. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 10428--10436, 2020

  40. [48]

    Model ratatouille: Recycling diverse models for out-of-distribution generalization

    Ram \'e , A., Ahuja, K., Zhang, J., Cord, M., Bottou, L., and Lopez-Paz, D. Model ratatouille: Recycling diverse models for out-of-distribution generalization. In International Conference on Machine Learning, pp.\ 28656--28679. PMLR, 2023

  41. [49]

    You only look once: Unified, real-time object detection

    Redmon, J. You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, 2016

  42. [50]

    High-resolution image synthesis with latent diffusion models

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., and Ommer, B. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 10684--10695, 2022

  43. [51]

    Self-supervised representation learning on neural network weights for model characteristic prediction

    Sch \"u rholt, K., Kostadinov, D., and Borth, D. Self-supervised representation learning on neural network weights for model characteristic prediction. Advances in Neural Information Processing Systems, 34: 0 16481--16493, 2021

  44. [52]

    W., and Borth, D

    Sch \"u rholt, K., Mahoney, M. W., and Borth, D. Towards scalable and versatile weight space learning. arXiv preprint arXiv:2406.09997, 2024

  45. [53]

    Ziplora: Any subject in any style by effectively merging loras

    Shah, V., Ruiz, N., Cole, F., Lu, E., Lazebnik, S., Li, Y., and Jampani, V. Ziplora: Any subject in any style by effectively merging loras. arXiv preprint arXiv:2311.13600, 2023

  46. [54]

    Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face

    Shen, Y., Song, K., Tan, X., Li, D., Lu, W., and Zhuang, Y. Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face. Advances in Neural Information Processing Systems, 36, 2024

  47. [55]

    A., Gera, A., Sznajder, B., Choshen, L., Dor, L

    Tahan, S. A., Gera, A., Sznajder, B., Choshen, L., Dor, L. E., and Shnarch, E. Label-efficient model selection for text generation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp.\ 8384--8402, 2024

  48. [56]

    and Le, Q

    Tan, M. and Le, Q. Efficientnet: Rethinking model scaling for convolutional neural networks. In International conference on machine learning, pp.\ 6105--6114. PMLR, 2019

  49. [57]

    O., Houlsby, N., Kolesnikov, A., Beyer, L., Zhai, X., Unterthiner, T., Yung, J., Steiner, A., Keysers, D., Uszkoreit, J., et al

    Tolstikhin, I. O., Houlsby, N., Kolesnikov, A., Beyer, L., Zhai, X., Unterthiner, T., Yung, J., Steiner, A., Keysers, D., Uszkoreit, J., et al. Mlp-mixer: An all-mlp architecture for vision. Advances in neural information processing systems, 34: 0 24261--24272, 2021

  50. [58]

    Llama: Open and efficient foundation language models

    Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozi \`e re, B., Goyal, N., Hambro, E., Azhar, F., et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023

  51. [59]

    N., The, A

    Tran, V.-H., Vo, T. N., The, A. N., Huu, T. T., Nguyen-Nhat, M.-K., Tran, T., Pham, D.-T., and Nguyen, T. M. Equivariant neural functional networks for transformers. arXiv preprint arXiv:2410.04209, 2024

  52. [60]

    Predicting neural network accuracy from weights

    Unterthiner, T., Keysers, D., Gelly, S., Bousquet, O., and Tolstikhin, I. Predicting neural network accuracy from weights. arXiv preprint arXiv:2002.11448, 2020

  53. [61]

    Attention is all you need

    Vaswani, A. Attention is all you need. Advances in Neural Information Processing Systems, 2017

  54. [62]

    Pytorch image models

    Wightman, R. Pytorch image models. https://github.com/rwightman/pytorch-image-models, 2019

  55. [63]

    Y., Roelofs, R., Gontijo-Lopes, R., Morcos, A

    Wortsman, M., Ilharco, G., Gadre, S. Y., Roelofs, R., Gontijo-Lopes, R., Morcos, A. S., Namkoong, H., Farhadi, A., Carmon, Y., Kornblith, S., et al. Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time. In Internation...

  56. [64]

    A., and Bansal, M

    Yadav, P., Tam, D., Choshen, L., Raffel, C. A., and Bansal, M. Ties-merging: Resolving interference when merging models. Advances in Neural Information Processing Systems, 36, 2024

  57. [65]

    Phylolm: Inferring the phylogeny of large language models and predicting their performances in benchmarks

    Yax, N., Oudeyer, P.-Y., and Palminteri, S. Phylolm: Inferring the phylogeny of large language models and predicting their performances in benchmarks. 2025

  58. [66]

    Z., and Finn, C

    Zhou, A., Yang, K., Burns, K., Cardace, A., Jiang, Y., Sokota, S., Kolter, J. Z., and Finn, C. Permutation equivariant neural functionals. Advances in neural information processing systems, 36, 2024 a

  59. [67]

    Z., and Finn, C

    Zhou, A., Yang, K., Jiang, Y., Burns, K., Xu, W., Sokota, S., Kolter, J. Z., and Finn, C. Neural functional transformers. Advances in neural information processing systems, 36, 2024 b

  60. [68]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.