# On Initializing Transformers with Pre-trained Embeddings

arXiv: https://arxiv.org/abs/2407.12514 · HTML: https://mlsanity.com/paper/2407.12514/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 28 references that are in the map.
- First posted on arXiv 17 Jul 2024.

## Builds on (most cited first, 18 of 18)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [Efficient Estimation of Word Representations in Vector Space](https://mlsanity.com/paper/1301.3781/) (2013)
- [Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification](https://mlsanity.com/paper/1502.01852/) (2015)
- [A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference](https://mlsanity.com/paper/1704.05426/) (2017)
- [A Survey of Large Language Models](https://mlsanity.com/paper/2303.18223/) (2023)
- [mT5: A massively multilingual pre-trained text-to-text transformer](https://mlsanity.com/paper/2010.11934/) (2020)
- [Enriching Word Vectors with Subword Information](https://mlsanity.com/paper/1607.04606/) (2016)
- [Convolutional Neural Networks for Sentence Classification](https://mlsanity.com/paper/1408.5882/) (2014)
- [Self-Attention with Relative Position Representations](https://mlsanity.com/paper/1803.02155/) (2018)
- [Neural Architectures for Named Entity Recognition](https://mlsanity.com/paper/1603.01360/) (2016)
- [Multi30K: Multilingual English-German Image Descriptions](https://mlsanity.com/paper/1605.00459/) (2016)
- [When and Why are Pre-trained Word Embeddings Useful for Neural Machine Translation?](https://mlsanity.com/paper/1804.06323/) (2018)
- [WECHSEL: Effective initialization of subword embeddings for cross-lingual transfer of monolingual language models](https://mlsanity.com/paper/2112.06598/) (2021)
- [An Exploration of Word Embedding Initialization in Deep-Learning Tasks](https://mlsanity.com/paper/1711.09160/) (2017)

## Built on by (most cited first, 2 of 2)

- [ToolTok: Tool Tokenization for Efficient and Generalizable GUI Agents](https://mlsanity.com/paper/2602.02548/) (2026)
- [Cross-Granularity Representations for Biological Sequences: Insights from ESM and BiGCARP](https://mlsanity.com/paper/2603.20825/) (2026)

## Abstract (arXiv)

It has become common practice now to use random initialization schemes, rather than the pre-trained embeddings, when training transformer based models from scratch. Indeed, we find that pre-trained word embeddings from GloVe, and some sub-word embeddings extracted from language models such as T5 and mT5 fare much worse compared to random initialization. This is counter-intuitive given the well-known representational and transfer-learning advantages of pre-training. Interestingly, we also find that BERT and mBERT embeddings fare better than random initialization, showing the advantages of pre-trained representations. In this work, we posit two potential factors that contribute to these mixed results: the model sensitivity to parameter distribution and the embedding interactions with position encodings. We observe that pre-trained GloVe, T5, and mT5 embeddings have a wider distribution of values. As argued in the initialization studies, such large value initializations can lead to poor training because of saturated outputs. Further, the larger embedding values can, in effect, absorb the smaller position encoding values when added together, thus losing position information. Standardizing the pre-trained embeddings to a narrow range (e.g. as prescribed by Xavier) leads to substantial gains for Glove, T5, and mT5 embeddings. On the other hand, BERT pre-trained embeddings, while larger, are still relatively closer to Xavier initialization range which may allow it to effectively transfer the pre-trained knowledge.