Pith. sign in

REVIEW 7 cited by

Predicting Neural Network Accuracy from Weights

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2002.11448 v4 pith:ZYSVPC5F submitted 2020-02-26 stat.ML cs.LG

Thomas Unterthiner , Daniel Keysers , Sylvain Gelly , Olivier Bousquet , Ilya Tolstikhin This is my paper · ORCID
classification stat.MLcs.LG
keywords neuralaccuracydifferentnetworknetworkstrainedweightsable
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We show experimentally that the accuracy of a trained neural network can be predicted surprisingly well by looking only at its weights, without evaluating it on input data. We motivate this task and introduce a formal setting for it. Even when using simple statistics of the weights, the predictors are able to rank neural networks by their performance with very high accuracy (R2 score more than 0.98). Furthermore, the predictors are able to rank networks trained on different, unobserved datasets and with different architectures. We release a collection of 120k convolutional neural networks trained on four different datasets to encourage further research in this area, with the goal of understanding network training and performance better.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. On the Expressive Power of Permutation-Equivariant Weight-Space Networks

    cs.LG 2026-02 conditional novelty 7.0 of 10

    Permutation-equivariant weight-space networks are all equally expressive, and universality holds when hidden-layer biases are pairwise distinct.

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

    cs.LG 2025-02 conditional novelty 7.0 of 10

    ProbeLog represents each classifier output by its responses to fixed probe images and uses CLIP to answer text queries, achieving 43.8% top-1 accuracy when searching 1,500 ImageNet-trained models for a concept.

  3. WeightCLIP: Aligning Datasets and Models for Weight Space Learning

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Contrastive dataset–weight alignment reshapes weight-space latents so dataset prompts retrieve, generate, and refine neural nets better than prior weight-space methods.

  4. Suitability Filter: A Statistical Framework for Classifier Evaluation in Real-World Deployment Settings

    cs.LG 2025-05 conditional novelty 6.0 of 10

    A statistical non-inferiority test on estimated per-sample correctness probabilities flags when a classifier's accuracy on unlabeled user data drops by more than a chosen margin relative to its test set.

  5. An open dataset of neural networks for hypernetwork research

    cs.LG 2025-07 reject novelty 5.0 of 10

    A public dataset of 10,000 LeNet-5 networks split into 10 Imagenette classes is released, with a 72% Naive Bayes baseline for classifying networks by their weights.

  6. NeuroVoxel-LM: Language-Aligned 3D Perception via Dynamic Voxelization and Meta-Embedding

    cs.CV 2025-07 conditional novelty 4.0 of 10

    NeuroVoxel-LM combines dynamic multi-resolution voxelization with attention-based pooling of NeRF weights, reporting faster 3D feature extraction and modestly better NeRF captioning than fixed-resolution and max-pooli...

  7. Predicting Deep Neural Network Training Outcomes from Early Training Telemetry

    cs.CL 2026-08 conditional

    After only a few epochs, a run's own loss, accuracy, gradient, and weight-norm telemetry predicts its final accuracy (R² = 0.92 to 0.99) and relative performance (AUC = 0.983 to 0.998) across six image-classification domains.

Pith tools