LLM Encoder vs. Decoder: Robust Detection of Chinese AI-Generated Text with LoRA

• URL: http://arxiv.org/abs/2509.00731v1
• Date: Sun, 31 Aug 2025 07:51:22 GMT
• Title: LLM Encoder vs. Decoder: Robust Detection of Chinese AI-Generated Text with LoRA
• Authors: Houji Jin, Negin Ashrafi, Armin Abdollahi, Wei Liu, Jian Wang, Ganyu Gui, Maryam Pishgar, Huanghao Feng,
• Abstract summary: We compare encoder-based Transformers (Chinese BERT-large and RoBERTa-wwm-ext-large), a decoder-only LLM (Alibaba's Qwen2.5-7B/Deep-R1-Distill-Qwen-7B fine-tuned via Low-Rank Adaptation, LoRA), and a FastText baseline.<n>Experiments reveal that although encoder models nearly memorize training data, they suffer significant performance degradation under distribution shifts.
• Score: 4.104443734934105
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The rapid growth of large language models (LLMs) has heightened the demand for accurate detection of AI-generated text, particularly in languages like Chinese, where subtle linguistic nuances pose significant challenges to current methods. In this study, we conduct a systematic comparison of encoder-based Transformers (Chinese BERT-large and RoBERTa-wwm-ext-large), a decoder-only LLM (Alibaba's Qwen2.5-7B/DeepSeek-R1-Distill-Qwen-7B fine-tuned via Low-Rank Adaptation, LoRA), and a FastText baseline using the publicly available dataset from the NLPCC 2025 Chinese AI-Generated Text Detection Task. Encoder models were fine-tuned using a novel prompt-based masked language modeling approach, while Qwen2.5-7B was adapted for classification with an instruction-format input and a lightweight classification head trained via LoRA. Experiments reveal that although encoder models nearly memorize training data, they suffer significant performance degradation under distribution shifts (RoBERTa: 76.3% test accuracy; BERT: 79.3%). FastText demonstrates surprising lexical robustness (83.5% accuracy) yet lacks deeper semantic understanding. In contrast, the LoRA-adapted Qwen2.5-7B achieves 95.94% test accuracy with balanced precision-recall metrics, indicating superior generalization and resilience to dataset-specific artifacts. These findings underscore the efficacy of decoder-based LLMs with parameter-efficient fine-tuning for robust Chinese AI-generated text detection. Future work will explore next-generation Qwen3 models, distilled variants, and ensemble strategies to enhance cross-domain robustness further.

Related papers

• A Lightweight Approach to Detection of AI-Generated Texts Using Stylometric Features [0.0]
  We introduce NEULIF, a lightweight approach that achieves best performance in the lightweight detector class.<n>In our approach, a text is first decomposed into stylometric and readability features which are then used for classification by a compact Convolutional Neural Network (CNN) or Random Forest (RF)<n>Models achieve 97% accuracy ( 0.95 F1) for CNN and 95% accuracy ( 0.94 F1) for the Random Forest, demonstrating high precision and recall, with ROC-AUC scores of 99.5% and 95%, respectively.
  arXiv  Detail & Related papers  (2025-11-22T08:08:03Z)
• LRANet++: Low-Rank Approximation Network for Accurate and Efficient Text Spotting [118.93173826110815]
  We propose a novel parameterized text shape method based on low-rank approximation for precise detection.<n>By exploiting the inherent shape correlation among different text contours, our method achieves consistency and compactness in shape representation.<n>We integrate the enhanced detection module with a lightweight recognition branch to form an end-to-end text spotting framework, termed LRANet++.
  arXiv  Detail & Related papers  (2025-11-08T03:08:03Z)
• Languages are Modalities: Cross-Lingual Alignment via Encoder Injection [0.8461674097042394]
  We present a compute efficient language-as-modality method that conditions an instruction-tuned decoder without changing the tokenizer or retraining the decoder.<n>LLINK substantially improves bilingual retrieval and achieves 81.3% preference over the base model.<n>We find that improvements can be attributed to reduced tokenization inflation and a stronger cross lingual alignment.
  arXiv  Detail & Related papers  (2025-10-31T07:43:21Z)
• Diversity Boosts AI-Generated Text Detection [51.56484100374058]
  DivEye is a novel framework that captures how unpredictability fluctuates across a text using surprisal-based features.<n>Our method outperforms existing zero-shot detectors by up to 33.2% and achieves competitive performance with fine-tuned baselines.
  arXiv  Detail & Related papers  (2025-09-23T10:21:22Z)
• Assessing Classical Machine Learning and Transformer-based Approaches for Detecting AI-Generated Research Text [0.0]
  Machine learning approaches can distinguish ChatGPT-3.5-generated texts from human-written texts.<n>DistilBERT achieves the overall best performance, while Logistic Regression and BERT-Custom offer solid, balanced alternatives.
  arXiv  Detail & Related papers  (2025-09-20T04:36:21Z)
• Group-Adaptive Threshold Optimization for Robust AI-Generated Text Detection [58.419940585826744]
  We introduce FairOPT, an algorithm for group-specific threshold optimization for probabilistic AI-text detectors.<n>We partitioned data into subgroups based on attributes (e.g., text length and writing style) and implemented FairOPT to learn decision thresholds for each group to reduce discrepancy.<n>Our framework paves the way for more robust classification in AI-generated content detection via post-processing.
  arXiv  Detail & Related papers  (2025-02-06T21:58:48Z)
• Text to Band Gap: Pre-trained Language Models as Encoders for Semiconductor Band Gap Prediction [5.812284760539713]
  We investigate the use of transformer-based language models, RoBERTa, T5, and LLaMA, for predicting the band gaps of semiconductor materials.<n>We construct material descriptions in two formats: structured strings that combine key features in a consistent template, and natural language narratives generated using the ChatGPT API.<n>Our results show that finetuned language models, particularly the decoder-only LLaMA-3 architecture, can outperform conventional approaches in prediction accuracy and flexibility.
  arXiv  Detail & Related papers  (2025-01-07T00:56:26Z)
• GigaCheck: Detecting LLM-generated Content [72.27323884094953]
  In this work, we investigate the task of generated text detection by proposing the GigaCheck. Our research explores two approaches: (i) distinguishing human-written texts from LLM-generated ones, and (ii) detecting LLM-generated intervals in Human-Machine collaborative texts. Specifically, we use a fine-tuned general-purpose LLM in conjunction with a DETR-like detection model, adapted from computer vision, to localize AI-generated intervals within text.
  arXiv  Detail & Related papers  (2024-10-31T08:30:55Z)
• ESPERANTO: Evaluating Synthesized Phrases to Enhance Robustness in AI Detection for Text Origination [1.8418334324753884]
  This paper introduces back-translation as a novel technique for evading detection. We present a model that combines these back-translated texts to produce a manipulated version of the original AI-generated text. We evaluate this technique on nine AI detectors, including six open-source and three proprietary systems.
  arXiv  Detail & Related papers  (2024-09-22T01:13:22Z)
• Are AI-Generated Text Detectors Robust to Adversarial Perturbations? [9.001160538237372]
  Current detectors for AI-generated text (AIGT) lack robustness against adversarial perturbations. This paper investigates the robustness of existing AIGT detection methods and introduces a novel detector, the Siamese Calibrated Reconstruction Network (SCRN) The SCRN employs a reconstruction network to add and remove noise from text, extracting a semantic representation that is robust to local perturbations.
  arXiv  Detail & Related papers  (2024-06-03T10:21:48Z)
• Technical Report on the Pangram AI-Generated Text Classifier [0.14732811715354457]
  We present Pangram Text, a transformer-based neural network trained to distinguish text written by large language models from text written by humans. We show that Pangram Text is not biased against nonnative English speakers and generalizes to domains and models unseen during training.
  arXiv  Detail & Related papers  (2024-02-21T17:13:41Z)
• Context Perception Parallel Decoder for Scene Text Recognition [52.620841341333524]
  Scene text recognition methods have struggled to attain high accuracy and fast inference speed. We present an empirical study of AR decoding in STR, and discover that the AR decoder not only models linguistic context, but also provides guidance on visual context perception. We construct a series of CPPD models and also plug the proposed modules into existing STR decoders. Experiments on both English and Chinese benchmarks demonstrate that the CPPD models achieve highly competitive accuracy while running approximately 8x faster than their AR-based counterparts.
  arXiv  Detail & Related papers  (2023-07-23T09:04:13Z)
• LLMDet: A Third Party Large Language Models Generated Text Detection Tool [119.0952092533317]
  Large language models (LLMs) are remarkably close to high-quality human-authored text. Existing detection tools can only differentiate between machine-generated and human-authored text. We propose LLMDet, a model-specific, secure, efficient, and extendable detection tool.
  arXiv  Detail & Related papers  (2023-05-24T10:45:16Z)
• DPIC: Decoupling Prompt and Intrinsic Characteristics for LLM Generated Text Detection [56.513637720967566]
  Large language models (LLMs) can generate texts that pose risks of misuse, such as plagiarism, planting fake reviews on e-commerce platforms, or creating inflammatory false tweets. Existing high-quality detection methods usually require access to the interior of the model to extract the intrinsic characteristics. We propose to extract deep intrinsic characteristics of the black-box model generated texts.
  arXiv  Detail & Related papers  (2023-05-21T17:26:16Z)
• Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense [56.077252790310176]
  We present a paraphrase generation model (DIPPER) that can paraphrase paragraphs, condition on surrounding context, and control lexical diversity and content reordering. Using DIPPER to paraphrase text generated by three large language models (including GPT3.5-davinci-003) successfully evades several detectors, including watermarking. We introduce a simple defense that relies on retrieving semantically-similar generations and must be maintained by a language model API provider.
  arXiv  Detail & Related papers  (2023-03-23T16:29:27Z)
• Code-Switching Text Generation and Injection in Mandarin-English ASR [57.57570417273262]
  We investigate text generation and injection for improving the performance of an industry commonly-used streaming model, Transformer-Transducer (T-T) We first propose a strategy to generate code-switching text data and then investigate injecting generated text into T-T model explicitly by Text-To-Speech (TTS) conversion or implicitly by tying speech and text latent spaces. Experimental results on the T-T model trained with a dataset containing 1,800 hours of real Mandarin-English code-switched speech show that our approaches to inject generated code-switching text significantly boost the performance of T-T models.
  arXiv  Detail & Related papers  (2023-03-20T09:13:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.