AnomaLLMy -- Detecting anomalous tokens in black-box LLMs through low-confidence single-token predictions

Waligóra Witold
6/28/2024
cs.CLcs.AI

No official code found. Best implementation: wwa/AnomaLLMy (2 stars, MIT, last commit 2024-07-07).

Abstract

This paper introduces AnomaLLMy, a novel technique for the automatic detection of anomalous tokens in black-box Large Language Models (LLMs) with API-only access. Utilizing low-confidence single-token predictions as a cost-effective indicator, AnomaLLMy identifies irregularities in model behavior, addressing the issue of anomalous tokens degrading the quality and reliability of models. Validated on the cl100k_base dataset, the token set of GPT-4, AnomaLLMy detected 413 major and 65 minor anomalies, demonstrating the method's efficiency with just \$24.39 spent in API credits. The insights from this research are expected to be beneficial for enhancing the robustness of and accuracy of LLMs, particularly in the development and assessment of tokenizers.

View on arXivView PDF

Code Implementations(1)

wwa/AnomaLLMyCommunity100%
MIT

AnomaLLMy - detecting anomalous LLM tokens through low-confidence single-token predictions

20PythonMar 11, 20242 years agoMIT

Similar Papers

Anomaly Detection of Tabular Data Using LLMs

85% match2024

Visualizing token importance for black-box language models

84% match2025

Glitch Tokens in Large Language Models: Categorization Taxonomy and Effective Detection

84% match2024

The Challenge of Identifying the Origin of Black-Box Large Language Models

84% match2025

GlitchProber: Advancing Effective Detection and Mitigation of Glitch Tokens in Large Language Models

83% match2 repos32024

Detecting Hallucinations in Large Language Model Generation: A Token Probability Approach

83% match1 repo172024

Cite this paper

@article{witold2024anomallmy,
  title  = {AnomaLLMy -- Detecting anomalous tokens in black-box LLMs through low-confidence single-token predictions},
  author = {Waligóra Witold},
  year   = {2024},
  eprint = {2406.19840},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.19840v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2406.19840.svg)](https://codeofpaper.com/paper/2406.19840)

Discussion