Pith. sign in

REVIEW 1 cited by

Text Rendering Strategies for Pixel Language Models

Not yet reviewed by Pith; the record is open.

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 2311.00522 v1 pith:BL5VTNRR submitted 2023-11-01 cs.CL

Text Rendering Strategies for Pixel Language Models

classification cs.CL
keywords renderinglanguagemodeltextmodelstasksapproachesbigram
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Pixel-based language models process text rendered as images, which allows them to handle any script, making them a promising approach to open vocabulary language modelling. However, recent approaches use text renderers that produce a large set of almost-equivalent input patches, which may prove sub-optimal for downstream tasks, due to redundancy in the input representations. In this paper, we investigate four approaches to rendering text in the PIXEL model (Rust et al., 2023), and find that simple character bigram rendering brings improved performance on sentence-level tasks without compromising performance on token-level or multilingual tasks. This new rendering strategy also makes it possible to train a more compact model with only 22M parameters that performs on par with the original 86M parameter model. Our analyses show that character bigram rendering leads to a consistently better model but with an anisotropic patch embedding space, driven by a patch frequency bias, highlighting the connections between image patch- and tokenization-based language models.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

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

  1. Pixels for Programs? A Cross-Provider Case Study of Input-Token Accounting for Source Code as Text and Images

    cs.SE 2026-07 conditional novelty 7.0

    Across 675 paired API calls, image input-token reductions are 86.5% (Anthropic), 80.6% (OpenAI), and 75.8% (Gemini) under token-volume weighting, with Gemini images costing far more than text below 200 lines.