Pith. sign in

REVIEW 4 cited by

Getting ViT in Shape: Scaling Laws for Compute-Optimal Model Design

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 2305.13035 v5 pith:7H3ZZFSO submitted 2023-05-22 cs.CV cs.LG

Ibrahim Alabdulmohsin , Xiaohua Zhai , Alexander Kolesnikov , Lucas Beyer This is my paper · ORCID
classification cs.CVcs.LG
keywords modelscalingcompute-optimalvisionachievesacrosscomputelaws
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Scaling laws have been recently employed to derive compute-optimal model size (number of parameters) for a given compute duration. We advance and refine such methods to infer compute-optimal model shapes, such as width and depth, and successfully implement this in vision transformers. Our shape-optimized vision transformer, SoViT, achieves results competitive with models that exceed twice its size, despite being pre-trained with an equivalent amount of compute. For example, SoViT-400m/14 achieves 90.3% fine-tuning accuracy on ILSRCV2012, surpassing the much larger ViT-g/14 and approaching ViT-G/14 under identical settings, with also less than half the inference cost. We conduct a thorough evaluation across multiple tasks, such as image classification, captioning, VQA and zero-shot transfer, demonstrating the effectiveness of our model across a broad range of domains and identifying limitations. Overall, our findings challenge the prevailing approach of blindly scaling up vision models and pave a path for a more informed scaling.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

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

  1. RADIO1D: Elastic Representations for Condensed Vision Modeling

    cs.CV 2026-07 accept novelty 7.0 of 10

    RADIO1D produces elastic hierarchical 1D visual tokens via multi-teacher distillation that match or beat fixed 2D encoders in VLMs at lower token counts.

  2. Can Pretrained Vision-Language Embeddings Alone Guide Robot Navigation?

    cs.RO 2025-06 conditional novelty 6.0 of 10

    A behavior-cloning policy trained only on frozen SigLIP embeddings reaches 74% of language-specified targets in a simple simulator, versus 100% for a state-aware expert, and takes 3.2x more steps.

  3. Image Recognition with Vision and Language Embeddings of VLMs

    cs.CV 2025-09 conditional novelty 4.0 of 10

    A benchmark of dual-encoder VLMs finds text and image embeddings give complementary class accuracy, and a per-class precision fusion rule adds about 0.4% accuracy over either alone on ImageNet.

  4. Vector embedding of multi-modal texts: a tool for discovery?

    cs.IR 2025-09 conditional novelty 4.0 of 10

    Using ColPali embeddings of 3,600 textbook page images, cosine similarity beats dot product, Euclidean, and Manhattan distances on top-5 retrieval, but only reaches 0.51 precision@5 without a text-only baseline.

Pith tools