Abstract: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.
| Comments: | 12 pages, 6 figures. Under review |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2605.19908 [cs.CL] |
| (or arXiv:2605.19908v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.19908 arXiv-issued DOI via DataCite |
From: Francis Kulumba [view email]
[v1]
Tue, 19 May 2026 14:37:51 UTC (174 KB)
[v2]
Mon, 25 May 2026 18:03:03 UTC (174 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。