The rapid advancement of large language models (LLMs) such as ChatGPT, DeepSeek, and Claude has significantly increased the presence of AI-generated text in digital communication. This trend has heightened the need for reliable detection methods to distinguish between human-authored and machine-generated content. Existing approaches both zero-shot methods and supervised classifiers largely conceptualize this task as a binary classification problem, often leading to poor generalization across domains and models. In this paper, we argue that such a binary formulation fundamentally mischaracterizes the detection task by assuming a coherent representation of human-written texts. In reality, human texts do not constitute a unified distribution, and their diversity cannot be effectively captured through limited sampling. This causes previous classifiers to memorize observed OOD characteristics rather than learn the essence of `non-ID' behavior, limiting generalization to unseen human-authored inputs. Based on this observation, we propose reframing the detection task as an out-of-distribution (OOD) detection problem, treating human-written texts as distributional outliers while machine-generated texts are in-distribution (ID) samples. To this end, we develop a detection framework using one-class learning method including DeepSVDD and HRN, and score-based learning techniques such as energy-based method, enabling robust and generalizable performance. Extensive experiments across multiple datasets validate the effectiveness of our OOD-based approach. Specifically, the OOD-based method achieves 98.3\% AUROC and AUPR with only 8.9\% FPR95 on DeepFake dataset. Moreover, we test our detection framework on multilingual, attacked, and unseen-model and -domain text settings, demonstrating the robustness and generalizability of our framework. Code will be released openly and also available in the supplementary materials.

The advent of Large Language Models (LLMs) has revolutionized text generation, producing outputs that closely mimic human writing. This blurring of lines between machine- and human-written text presents new challenges in distinguishing one from the other – a task further complicated by the frequent updates and closed nature of leading proprietary LLMs. Traditional logits-based detection methods leverage surrogate models for identifying LLM-generated content when the exact logits are unavailable from black-box LLMs. However, these methods grapple with the misalignment between the distributions of the surrogate and the often undisclosed target models, leading to performance degradation, particularly with the introduction of new, closed-source models. Furthermore, while current methodologies are generally effective when the source model is identified, they falter in scenarios where the model version remains unknown, or the test set comprises outputs from various source models. To address these limitations, we present \textbf{D}istribution-\textbf{A}ligned \textbf{L}LMs \textbf{D}etection (DALD), an innovative framework that redefines the state-of-the-art performance in black-box text detection even without logits from source LLMs. DALD is designed to align the surrogate model's distribution with that of unknown target LLMs, ensuring enhanced detection capability and resilience against rapid model iterations with minimal training investment. By leveraging corpus samples from publicly accessible outputs of advanced models such as ChatGPT, GPT-4 and Claude-3, DALD fine-tunes surrogate models to synchronize with unknown source model distributions effectively. Our approach achieves SOTA performance in black-box settings on different advanced closed-source and open-source models. The versatility of our method enriches widely adopted zero-shot detection frameworks (DetectGPT, DNA-GPT, Fast-DetectGPT) with a `plug-and-play' enhancement feature. Extensive experiments validate that our methodology reliably secures high detection precision for LLM-generated text and effectively detects text from diverse model origins through a singular detector. Our method is also robust under the revised text attack and non-English texts.

BACKGROUND: Protein histidine phosphorylation (pHis) plays critical roles in prokaryotic signal transduction pathways and various eukaryotic cellular processes. It is estimated to account for 6-10% of the phosphoproteome, however only hundreds of pHis sites have been discovered to date. Due to the inherent disadvantages of experimental methods, it is an urgent task for developing efficient computational approaches to identify pHis sites. RESULTS: Here, we present a novel tool, pHisPred, for accurately identifying pHis sites from protein sequences. We manually collected the largest number of experimental validated pHis sites to build benchmark datasets. Using randomized tenfold CV, the weighted SVM-RBF model shows the best performance than other four commonly used classification models (LR, KNN, RF, and MLP). From ten thousands of features, 140 and 150 most informative features were individually selected out for eukaryotic and prokaryotic models. The average AUC and F1-score values of pHisPred were (0.81, 0.40) and (0.78, 0.46) for tenfold CV on the eukaryotic and prokaryotic training datasets, respectively. In addition, pHisPred significantly outperforms other tools on testing datasets, in particular on the eukaryotic one. CONCLUSION: We implemented a python program of pHisPred, which is freely available for non-commercial use at https://github.com/xiaofengsong/pHisPred . Moreover, users can use it to train new models with their own data.

Bacterial pathogens secrete numerous effector proteins via six secretion systems, type I to type VI secretion systems, to adapt to new environments or to promote virulence by bacterium-host interactions. Many computational approaches have been used in the identification of effector proteins before the subsequent experimental verification because they tolerate laborious biological procedures and are genome scale, automated and highly efficient. Prevalent examples include machine learning methods and statistical techniques. In this article, we summarize the computational progress toward predicting secreted effector proteins in bacteria, with an opening of an introduction of features that are used to discriminate effectors from non-effectors. The mechanism, contribution and deficiency of previous developed detection tools are presented, which are further benchmarked based on a curated testing data set. According to the results of benchmarking, potential improvements of the prediction performance are discussed, which include (1) more informative features for discriminating the effectors from non-effectors; (2) the construction of comprehensive training data set of the machine learning algorithms; (3) the advancement of reliable prediction methods and (4) a better interpretation of the mechanisms behind the molecular processes. The future of in silico identification of bacterial secreted effectors includes both opportunities and challenges.