To MRL or Not To MRL: Comparing Random Vector Truncation Against Matryoshka Embeddings as Cost Reduction Methods for Text Encoders

ACL ARR 2026 January Submission7192 Authors

Keywords: Matryoshka Representation Learning, Text Encoder, Text Retrieval, Text Classification

Abstract:

Matryoshka Representation Learning (MRL) is a widely adopted approach
for training text encoders so they provide useful text representations
at various sizes, available by simply truncating the resulting vectors
at sizes pre-determined at training time.
Recent works have shown that randomly truncating text embeddings has
minimal impact in downstream performance unless vectors are reduced in
size by at least 70\%.
However, random truncation has not yet been compared to MRL, so
that it is unclear to what extent it is useful at
reducing costs in applications that rely on text encoders.
In this short paper, we benchmark random truncation applied to models
that were trained with and without MRL.
Our results across several models and downstream tasks show that,
unless heavily truncating embeddings (i.e.\ reducing their size by at
least 80\%), randomly truncated embeddings of non-MRL models are at
least competitive, and often outperform models trained with MRL.
This suggests that random truncation is indeed a highly effective
method of embedding reduction, even compared to MRL, and that it is
unclear how to best train models with MLR, as the additional training
costs only become beneficial at very high truncation levels.
Our code is attached to our ARR submission.

Paper Type: Short

Research Area: Interpretability and Analysis of Models for NLP

Research Area Keywords: robustness,retrieval

Contribution Types: Model analysis & interpretability

Languages Studied: English

Submission Number: 7192