# Where Does Authorship Signal Emerge in Encoder-Based Language Models?

arXiv: https://arxiv.org/abs/2605.19908 · HTML: https://mlsanity.com/paper/2605.19908/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 36 references that are in the map.
- First posted on arXiv 19 May 2026.

## Builds on (most cited first, 25 of 25)

- [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)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere](https://mlsanity.com/paper/2005.10242/) (2020)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Text Embeddings by Weakly-Supervised Contrastive Pre-training](https://mlsanity.com/paper/2212.03533/) (2022)
- [ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT](https://mlsanity.com/paper/2004.12832/) (2020)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [BERT Rediscovers the Classical NLP Pipeline](https://mlsanity.com/paper/1905.05950/) (2019)
- [Probing Classifiers: Promises, Shortcomings, and Advances](https://mlsanity.com/paper/2102.12452/) (2021)
- [Designing and Interpreting Probes with Control Tasks](https://mlsanity.com/paper/1909.03368/) (2019)
- [Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference](https://mlsanity.com/paper/2412.13663/) (2024)
- [Localizing Model Behavior with Path Patching](https://mlsanity.com/paper/2304.05969/) (2023)
- [Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance?](https://mlsanity.com/paper/2005.00719/) (2020)
- [Same Author or Just Same Topic? Towards Content-Independent Style Representations](https://mlsanity.com/paper/2204.04907/) (2022)
- [Does It Capture STEL? A Modular, Similarity-based Linguistic Style Evaluation Framework](https://mlsanity.com/paper/2109.04817/) (2021)
- [Whodunit? Learning to Contrast for Authorship Attribution](https://mlsanity.com/paper/2209.11887/) (2022)
- [Why Moli`ere most likely did write his plays](https://mlsanity.com/paper/2001.01595/) (2020)
- [Git Blame Who?: Stylistic Authorship Attribution of Small, Incomplete Source Code Fragments](https://mlsanity.com/paper/1701.05681/) (2017)
- [Latent Space Interpretation for Stylistic Analysis and Explainable Authorship Attribution](https://mlsanity.com/paper/2409.07072/) (2024)
- [Layered Insights: Generalizable Analysis of Authorial Style by Leveraging All Transformer Layers](https://mlsanity.com/paper/2503.00958/) (2025)
- [IDIOLEX: Unified and Continuous Representations for Idiolectal and Stylistic Variation](https://mlsanity.com/paper/2604.04704/) (2026)
- [HALvest-Contrastive: Retrieval-Like Authorship Attribution with Patch-Level Late Interaction](https://mlsanity.com/paper/2407.20595/) (2024)

## Abstract (arXiv)

Authorship attribution models fine-tuned with the same pretrained encoder, data, and loss can differ four-fold in performance depending only on their scoring mechanism. We use mechanistic interpretability tools to explain this gap. Stylistic features such as word length, punctuation density, and function-word frequency are similarly available at every layer in every model we probe, including an off-the-shelf control encoder, suggesting that the gap is not explained by their linear readability. Instead, causal intervention shows that the scorer appears to determine where the encoder consolidates authorship signal. Mean pooling forces consolidation by early to mid layers, while late interaction defers it to later layers. We further derive this difference from the gradient structure of each scorer, and training dynamics reveal distinct learning trajectories that follow from that difference.