# Layer-Wise Evolution of Representations in Fine-Tuned Transformers: Insights from Sparse AutoEncoders

arXiv: https://arxiv.org/abs/2502.16722 · HTML: https://mlsanity.com/paper/2502.16722/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in SAE features and their applications, one of 13 directions in Mechanistic interpretability.
- Builds on 1 of its 6 references that are in the map.
- First posted on arXiv 23 Feb 2025.

## Placed in

- Mechanistic interpretability / [SAE features and their applications](https://mlsanity.com/mechinterp/sae-features-and-their-applications/)

## Builds on (most cited first, 1 of 1)

- [Intelligent Learning Rate Distribution to reduce Catastrophic Forgetting in Transformers](https://mlsanity.com/paper/2404.01317/) (2024)

## Built on by (most cited first, 4 of 4)

- [MSPLoRA: A Multi-Scale Pyramid Low-Rank Adaptation for Efficient Model Fine-Tuning](https://mlsanity.com/paper/2503.21838/) (2025)
- [Beyond Temperature: Hyperfitting as a Late-Stage Geometric Expansion](https://mlsanity.com/paper/2605.22579/) (2026)
- [Distributed Multi-Layer Editing for Rule-Level Knowledge in Large Language Models](https://mlsanity.com/paper/2604.08284/) (2026)
- [Analysis of Quantized and Efficiently Adapted Protein Language Models](https://mlsanity.com/paper/2610.00665/) (2026)

## Abstract (arXiv)

Fine-tuning pre-trained transformers is a powerful technique for enhancing the performance of base models on specific tasks. From early applications in models like BERT to fine-tuning Large Language Models (LLMs), this approach has been instrumental in adapting general-purpose architectures for specialized downstream tasks. Understanding the fine-tuning process is crucial for uncovering how transformers adapt to specific objectives, retain general representations, and acquire task-specific features. This paper explores the underlying mechanisms of fine-tuning, specifically in the BERT transformer, by analyzing activation similarity, training Sparse AutoEncoders (SAEs), and visualizing token-level activations across different layers. Based on experiments conducted across multiple datasets and BERT layers, we observe a steady progression in how features adapt to the task at hand: early layers primarily retain general representations, middle layers act as a transition between general and task-specific features, and later layers fully specialize in task adaptation. These findings provide key insights into the inner workings of fine-tuning and its impact on representation learning within transformer architectures.