How does content-adaptive tokenization affect the accuracy-efficiency trade-off on the MMMU and MathVista data

Abstract

Abstract: Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, well before a model is even invoked. The same text translated into different languages can have drastically different tokenization lengths, with differences up to 15 times in some cases. These disparities persist even for tokenizers that are intentionally trained for multilingual

Research Question

How does content-adaptive tokenization affect the accuracy-efficiency trade-off on the MMMU and MathVista datasets for multimodal models with 7B and 13B parameter language backbones, measured by FLOPs per sample and exact match scores?

Verification Level

Descriptive public verification status only; aggregate claim counts are public, but individual claim records are not exposed here.

Quality Tier

Descriptive public triage only; this tier does not alter current publication or DOI behavior.

Quality Dimensions

Automated triage signals derived from public fields; not human peer review or independent validation.

Correction Record

Public corrections are additive records. Current status does not claim the synthesis is error-free.

Provenance