# All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality

Source: [All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality](<https://www.sota2.com/research/paper/68fa6aba1e209131cea2a13f>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-25T23:32:17.810Z

Authors: William Timkey, Marten van Schijndel

Published: 2021-09-09T16:45:15.000Z

## Abstract

Similarity measures are a vital tool for understanding how language models represent and process language. Standard representational similarity measures such as cosine similarity and Euclidean distance have been successfully used in static word embedding models to understand how words cluster in semantic space. Recently, these measures have been applied to embeddings from contextualized models such as BERT and GPT-2. In this work, we call into question the informativity of such measures for contextualized language models. We find that a small number of rogue dimensions, often just 1-3, dominate these measures. Moreover, we find a striking mismatch between the dimensions that dominate similarity measures and those which are important to the behavior of the model. We show that simple postprocessing techniques such as standardization are able to correct for rogue dimensions and reveal underlying representational quality. We argue that accounting for rogue dimensions is essential for any similarity-based analysis of contextual language models.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Semantic Similarity Prediction on Human pile-sorting judgements (6-fold nested cross val)](<https://www.sota2.com/research/sota/semantic-similarity-prediction-on-human-pile-sorting-judgements-6-fold-nested-cross-val.md>): Semantic Similarity Prediction; Human pile-sorting judgements (6-fold nested cross val); Spearman ρ: 0.079.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2109.04404>)

[PDF](<https://arxiv.org/pdf/2109.04404.pdf>)
