Speculating LLMs' Chinese Training Data Pollution from Their Tokens

Tokens are basic elements in the datasets for LLM training. It is well-known that many tokens representing Chinese phrases in the vocabulary of GPT (4o/4o-mini/o1/o3/4.5/4.1/o4mini) 1 are indicating contents like pornography or online gambling. Based on this observation, our goal is to locate Polluted Chinese (PoC) tokens in LLMs and study the relationship between PoC tokens' existence and training data. (1) We give a formal definition and taxonomy of PoC tokens based on the GPT's vocabulary. (2) We build a PoC token detector via finetuning an LLM to label PoC tokens in vocabularies by considering each token's both semantics and related contents from the search engines. (3) We study how to speculate training data pollution via PoC tokens' appearances (token ID). Experiments on GPT and other 23 LLMs indicate that PoC tokens widely exist while GPT's vocabulary behaves the worst: more than 23% long Chinese tokens (i.e., a token with more than two Chinese characters) are either porn or online gambling. We validate our speculation method on famous pre-training datasets like C4 and Pile. Then, considering GPT-4o, we speculate the ratio of "波*野结衣" 2 related webpages in its training data is around 0.5%.