Pith. sign in

REVIEW 7 cited by

Learning to Learn with Generative Models of Neural Network Checkpoints

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 2209.12892 v1 pith:GCHLPNML submitted 2022-09-26 cs.LG cs.CV

William Peebles , Ilija Radosavovic , Tim Brooks , Alexei A. Efros , Jitendra Malik This is my paper
classification cs.LGcs.CV
keywords neurallearningnetworkparameterparametersapproachcheckpointsgenerative
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We explore a data-driven approach for learning to optimize neural networks. We construct a dataset of neural network checkpoints and train a generative model on the parameters. In particular, our model is a conditional diffusion transformer that, given an initial input parameter vector and a prompted loss, error, or return, predicts the distribution over parameter updates that achieve the desired metric. At test time, it can optimize neural networks with unseen parameters for downstream tasks in just one update. We find that our approach successfully generates parameters for a wide range of loss prompts. Moreover, it can sample multimodal parameter solutions and has favorable scaling properties. We apply our method to different neural network architectures and tasks in supervised and reinforcement learning.

Discussion (0). Sign in 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. Weight Space Representation Learning via Neural Field Adaptation

    cs.LG 2025-12 conditional novelty 6.0 of 10

    Multiplicative LoRA weights of pre-trained neural fields form structured, semantically meaningful representations that outperform prior weight-space methods for generation and classification.

  2. Hyper Diffusion Avatars: Dynamic Human Avatar Generation using Network Weight Space Diffusion

    cs.GR 2025-09 conditional novelty 6.0 of 10

    A diffusion model over per-person UNet weights generates new dynamic human avatars that render pose-dependent 3D Gaussians in real time.

  3. Unsupervised Behavioral Compression: Learning Low-Dimensional Policy Manifolds through State-Occupancy Matching

    cs.LG 2026-03 conditional novelty 5.0 of 10

    Occupancy-based Policy Compression (OPC) curates diverse policies via state-visitation entropy and trains an autoencoder by matching mixture occupancy distributions, yielding more useful low-dimensional policy manifol...

  4. Semantic-guided LoRA Parameters Generation

    cs.LG 2025-09 conditional novelty 5.0 of 10

    SG-LoRA generates LoRA parameters for unseen tasks from text descriptions alone, using semantic expert selection plus a conditional VAE, matching or exceeding oracle fine-tuning on retrieval benchmarks.

  5. Conflicting Scores, Confusing Signals: An Empirical Study of Vulnerability Scoring Systems

    cs.CR 2025-08 unverdicted novelty 5.0 of 10

    The abstract claims a first-of-kind, outcome-linked comparison of four vulnerability scoring systems showing major ranking disagreements, but the submitted full text is an unrelated paper, leaving the study unevaluable.

  6. Continual Adaptation: Environment-Conditional Parameter Generation for Object Detection in Dynamic Scenarios

    cs.CV 2025-06 conditional novelty 5.0 of 10

    A conditional diffusion model generates LoRA adapter parameters for object detectors at test time, improving continual domain adaptation accuracy by small margins over prior methods.

  7. Text2Weight: Bridging Natural Language and Neural Network Weight Spaces

    cs.LG 2025-08 conditional novelty 4.0 of 10

    A diffusion transformer generates the weights of a frozen-feature CLIP classifier head from text task descriptions, achieving moderate accuracy on unseen class subsets.

Pith tools