Amplifying, Not Learning: Fine-Tuned AI Text Detectors Amplify a Pretrained Direction
Mirrored from arXiv — Machine Learning for archival readability. Support the source by reading on the original site.
Computer Science > Machine Learning
Title:Amplifying, Not Learning: Fine-Tuned AI Text Detectors Amplify a Pretrained Direction
Abstract:AI text detectors amplify a pretrained typicality axis; they do not construct an AI-vs-human boundary. On raw encoders before any task supervision, projecting onto centroid(AI)-centroid(HC3) achieves NYT-vs-HC3 AUROC 0.806/0.944/0.834 across three architectures (86-106% of the fine-tuned discrimination ceiling: on RoBERTa-base, raw projection exceeds fine-tuning); on RoBERTa-base, full fine-tuning reduces discrimination below raw on both fluent-formal populations tested. The same axis inverts on non-native ESL writing (AUROC 0.06-0.20) -- a falsifiable prediction unique to the typicality reading. A 24-example frozen probe matches full fine-tuning (0.900 vs 0.895). A closed-form Jacobian predictor parameterises axis-manipulating interventions with R^2 = 1.000 universal, lifts ELECTRA-CE deployment TPR from 0.000 to 0.904 at FPR = 1%, and transfers to three independently-trained third-party RoBERTa detectors at 16/16 oracle-equivalence (57% NYT-FPR reduction on the OpenAI detector). Scope: encoder family; mechanism magnitude HC3-anchored; population-level shared axis with per-text mechanisms varying across architectures. Three operationally distinct probes -- text-surface caps_rate residualisation, geometric signed-epsilon ablation, closed-form text-pair predictor -- agree at cos 0.74/0.81/1.00 across three architectures, confirming observer-invariance. Under matched-TPR-0.90 evaluation, the published intervention zoo (CC, dealign-f2c) is calibration-equivalent across 27 cells (|Delta AUROC| <= 0.0081), and >= 97% of the LoRA->full-FT bias gap on ELECTRA is calibration shift, not learned representation -- the central claim's prediction confirmed.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2605.21653 [cs.LG] |
| (or arXiv:2605.21653v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.21653
arXiv-issued DOI via DataCite (pending registration)
|
Access Paper:
- View PDF
- HTML (experimental)
- TeX Source
Current browse context:
References & Citations
Bookmark
Bibliographic and Citation Tools
Code, Data and Media Associated with this Article
Demos
Recommenders and Search Tools
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
More from arXiv — Machine Learning
-
Temporal Contrastive Transformer for Financial Crime Detection: Self-Supervised Sequence Embeddings via Predictive Contrastive Coding
May 22
-
Teaching Language Models to Forecast Research Success Through Comparative Idea Evaluation
May 22
-
The Attribution Impossibility: No Feature Ranking Is Faithful, Stable, and Complete Under Collinearity
May 22
-
Don't Collapse Your Features: Why CenterLoss Hurts OOD Detection and Multi-Scale Mahalanobis Wins
May 22
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.