# T5-CSBoost: Adversarial Perturbation Resistant LLM Fingerprinting

arXiv: https://arxiv.org/abs/2607.14113 · HTML: https://mlsanity.com/paper/2607.14113/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 35 references that are in the map.
- First posted on arXiv 8 May 2026.

## Builds on (most cited first, 23 of 23)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [PaLM: Scaling Language Modeling with Pathways](https://mlsanity.com/paper/2204.02311/) (2022)
- [SimCSE: Simple Contrastive Learning of Sentence Embeddings](https://mlsanity.com/paper/2104.08821/) (2021)
- [DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature](https://mlsanity.com/paper/2301.11305/) (2023)
- [Contrastive Decoding: Open-ended Text Generation as Optimization](https://mlsanity.com/paper/2210.15097/) (2022)
- [How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs](https://mlsanity.com/paper/2401.06373/) (2024)
- [Release Strategies and the Social Impacts of Language Models](https://mlsanity.com/paper/1908.09203/) (2019)
- [How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection](https://mlsanity.com/paper/2301.07597/) (2023)
- [Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense](https://mlsanity.com/paper/2303.13408/) (2023)
- [RADAR: Robust AI-Text Detection via Adversarial Learning](https://mlsanity.com/paper/2307.03838/) (2023)
- [MAGE: Machine-generated Text Detection in the Wild](https://mlsanity.com/paper/2305.13242/) (2023)
- [Red Teaming Language Model Detectors with Language Models](https://mlsanity.com/paper/2305.19713/) (2023)
- [DeTeCtive: Detecting AI-generated Text via Multi-Level Contrastive Learning](https://mlsanity.com/paper/2410.20964/) (2024)
- [Your Large Language Models Are Leaving Fingerprints](https://mlsanity.com/paper/2405.14057/) (2024)
- [Token Prediction as Implicit Classification to Identify LLM-Generated Text](https://mlsanity.com/paper/2311.08723/) (2023)
- [DAMAGE: Detecting Adversarially Modified AI Generated Text](https://mlsanity.com/paper/2501.03437/) (2025)
- [Are AI-Generated Text Detectors Robust to Adversarial Perturbations?](https://mlsanity.com/paper/2406.01179/) (2024)
- [SCGC : Self-Supervised Contrastive Graph Clustering](https://mlsanity.com/paper/2204.12656/) (2022)
- [TOPFORMER: Topology-Aware Authorship Attribution of Deepfake Texts with Diverse Writing Styles](https://mlsanity.com/paper/2309.12934/) (2023)
- [Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection](https://mlsanity.com/paper/2501.03940/) (2025)
- [Efficient block contrastive learning via parameter-free meta-node approximation](https://mlsanity.com/paper/2209.14067/) (2022)
- [NBC-Softmax : Darkweb Author fingerprinting and migration tracking](https://mlsanity.com/paper/2212.08184/) (2022)

## Abstract (arXiv)

While many AI-generated text (AIGT) detectors achieve strong performance on clean inputs, their accuracy degrades significantly under light paraphrasing, word substitutions, character edits, and distribution shifts. We present T5 Contrastive Style Boosted Classifier (T5-CSBoost), an extension to the T5-Sentinel framework that keeps the original next-token prediction objective for source attribution while introducing an auxiliary margin-based triplet loss over decoder embeddings. This contrastive style regularization encourages the learning of compact, perturbation-resistant stylistic representations, offering a lightweight yet effective alternative to prior approaches that rely on architectural modifications, adversarial training, or complex multi-task objectives without altering the underlying T5-small backbone. T5-CSBoost achieves state-of-the-art multiclass source attribution and binary human-vs-LLM detection on OpenLLMText and HC3 AIGT benchmarks. More importantly, T5-CSBoost demonstrates enhanced robustness to word and character level adversarial perturbations of up to 90% intensity, achieving state-of-the-art on the challenging MAGE/Deepfake stress-test suite, including unseen models, unseen domains, and extreme paraphrasing scenarios. Our results highlight that explicitly regularizing stylistic embeddings via contrastive learning is a practical and effective strategy for building more robust LLM fingerprinting systems in real-world adversarial settings.