Improbable Bigrams Expose Vulnerabilities of Incomplete Tokens in Byte-Level Tokenizers
Tokenization is a crucial step that bridges human-readable text with model-readable discrete tokens. However, recent studies have revealed that tokenizers can be exploited to elicit unwanted model behaviors. In this work, we investigate incomplete tokens, i.e., undecodable tokens with stray bytes re...
Gespeichert in:
Hauptverfasser: | , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Tokenization is a crucial step that bridges human-readable text with
model-readable discrete tokens. However, recent studies have revealed that
tokenizers can be exploited to elicit unwanted model behaviors. In this work,
we investigate incomplete tokens, i.e., undecodable tokens with stray bytes
resulting from byte-level byte-pair encoding (BPE) tokenization. We hypothesize
that such tokens are heavily reliant on their adjacent tokens and are fragile
when paired with unfamiliar tokens. To demonstrate this vulnerability, we
introduce improbable bigrams: out-of-distribution combinations of incomplete
tokens designed to exploit their dependency. Our experiments show that
improbable bigrams are significantly prone to hallucinatory behaviors.
Surprisingly, alternative tokenizations of the same phrases result in
drastically lower rates of hallucination (93% reduction in Llama3.1). We
caution against the potential vulnerabilities introduced by byte-level BPE
tokenizers, which may impede the development of trustworthy language models. |
---|---|
DOI: | 10.48550/arxiv.2410.23684 |