In this work, we report what happens when two large language models respond to each other for many turns without any outside input in a multi-agent setup. The setup begins with a short seed sentence. After that, each model reads the other's output and generates a response. This continues for a fixed number of steps. We used Mistral Nemo Base 2407 and Llama 2 13B hf. We observed that most conversations start coherently but later fall into repetition. In many runs, a short phrase appears and repeats across turns. Once repetition begins, both models tend to produce similar output rather than introducing a new direction in the conversation. This leads to a loop where the same or similar text is produced repeatedly. We describe this behavior as a form of convergence. It occurs even though the models are large, trained separately, and not given any prompt instructions. To study this behavior, we apply lexical and embedding-based metrics to measure how far the conversation drifts from the initial seed and how similar the outputs of the two models becomes as the conversation progresses.

We thank all reviewers for their detailed constructive feedback and suggestions. Table B (below) demonstrates this empirically. Gumbel-Softmax has) with significantly less training time and resource consumption. These experiments show that when trained with Gumbel-CRF, the AR decoder outperforms REINFORCE. We will clarify this in the paper.

Ensuring reliable data collection in large-scale particle physics experiments demands Data Quality Monitoring (DQM) procedures to detect possible detector malfunctions and preserve data integrity. Traditionally, this resource-intensive task has been handled by human shifters that struggle with frequent changes in operational conditions. We present novel, interpretable, robust, and scalable DQM algorithms designed to automate anomaly detection in time-dependent settings. Our approach constructs evolving histogram templates with built-in uncertainties, featuring both a statistical variant - extending the classical Exponentially Weighted Moving Average (EWMA) - and a machine learning (ML)-enhanced version that leverages a transformer encoder for improved adaptability. Experimental validations on synthetic datasets demonstrate the high accuracy, adaptability, and interpretability of these methods, with the statistical variant being commissioned in the LHCb experiment at the Large Hadron Collider, underscoring its real-world impact. The code used in this study is available at https://github.com/ArseniiGav/DINAMO.

A subset of points in a metric space is said to resolve it if each point in the space is uniquely characterized by its distance to each point in the subset. In particular, resolving sets can be used to represent points in abstract metric spaces as Euclidean vectors. Importantly, due to the triangle inequality, points close by in the space are represented as vectors with similar coordinates, which may find applications in classification problems of symbolic objects under suitably chosen metrics. In this manuscript, we address the resolvability of Jaccard spaces, i.e., metric spaces of the form $(2^X,\text{Jac})$, where $2^X$ is the power set of a finite set $X$, and $\text{Jac}$ is the Jaccard distance between subsets of $X$. Specifically, for different $a,b\in 2^X$, $\text{Jac}(a,b)=|a\Delta b|/|a\cup b|$, where $|\cdot|$ denotes size (i.e., cardinality) and $\Delta$ denotes the symmetric difference of sets. We combine probabilistic and linear algebra arguments to construct highly likely but nearly optimal (i.e., of minimal size) resolving sets of $(2^X,\text{Jac})$. In particular, we show that the metric dimension of $(2^X,\text{Jac})$, i.e., the minimum size of a resolving set of this space, is $\Theta(|X|/\ln|X|)$. In addition, we show that a much smaller subset of $2^X$ suffices to resolve, with high probability, all different pairs of subsets of $X$ of cardinality at most $\sqrt{|X|}/\ln|X|$, up to a factor.

The remarkable capabilities of large-scale language models, such as ChatGPT, in text generation have impressed readers and spurred researchers to devise detectors to mitigate potential risks, including misinformation, phishing, and academic dishonesty. Despite this, most previous studies have been predominantly geared towards creating detectors that differentiate between purely ChatGPT-generated texts and human-authored texts. This approach, however, fails to work on discerning texts generated through human-machine collaboration, such as ChatGPT-polished texts. Addressing this gap, we introduce a novel dataset termed HPPT (ChatGPT-polished academic abstracts), facilitating the construction of more robust detectors. It diverges from extant corpora by comprising pairs of human-written and ChatGPT-polished abstracts instead of purely ChatGPT-generated texts. Additionally, we propose the "Polish Ratio" method, an innovative measure of the degree of modification made by ChatGPT compared to the original human-written text. It provides a mechanism to measure the degree of ChatGPT influence in the resulting text. Our experimental results show our proposed model has better robustness on the HPPT dataset and two existing datasets (HC3 and CDB). Furthermore, the "Polish Ratio" we proposed offers a more comprehensive explanation by quantifying the degree of ChatGPT involvement.

Safeguarding the Intellectual Property (IP) of data has become critically important as machine learning applications continue to proliferate, and their success heavily relies on the quality of training data. While various mechanisms exist to secure data during storage, transmission, and consumption, fewer studies have been developed to detect whether they are already leaked for model training without authorization. This issue is particularly challenging due to the absence of information and control over the training process conducted by potential attackers. In this paper, we concentrate on the domain of tabular data and introduce a novel methodology, Local Distribution Shifting Synthesis (\textsc{LDSS}), to detect leaked data that are used to train classification models. The core concept behind \textsc{LDSS} involves injecting a small volume of synthetic data--characterized by local shifts in class distribution--into the owner's dataset. This enables the effective identification of models trained on leaked data through model querying alone, as the synthetic data injection results in a pronounced disparity in the predictions of models trained on leaked and modified datasets. \textsc{LDSS} is \emph{model-oblivious} and hence compatible with a diverse range of classification models, such as Naive Bayes, Decision Tree, and Random Forest. We have conducted extensive experiments on seven types of classification models across five real-world datasets. The comprehensive results affirm the reliability, robustness, fidelity, security, and efficiency of \textsc{LDSS}.

Quantifying the dissimilarity of two texts is an important aspect of a number of natural language processing tasks, including semantic information retrieval, topic classification, and document clustering. In this paper, we compared the properties and performance of different dissimilarity measures $D$ using three different representations of texts -- vocabularies, word frequency distributions, and vector embeddings -- and three simple tasks -- clustering texts by author, subject, and time period. Using the Project Gutenberg database, we found that the generalised Jensen--Shannon divergence applied to word frequencies performed strongly across all tasks, that $D$'s based on vector embedding representations led to stronger performance for smaller texts, and that the optimal choice of approach was ultimately task-dependent. We also investigated, both analytically and numerically, the behaviour of the different $D$'s when the two texts varied in length by a factor $h$. We demonstrated that the (natural) estimator of the Jaccard distance between vocabularies was inconsistent and computed explicitly the $h$-dependency of the bias of the estimator of the generalised Jensen--Shannon divergence applied to word frequencies. We also found numerically that the Jensen--Shannon divergence and embedding-based approaches were robust to changes in $h$, while the Jaccard distance was not.