We introduce Ghostbuster, a state-of-the-art system for detecting AI-generated text. Our method works by passing documents through a series of weaker language models, running a structured search over possible combinations of their features, and then training a classifier on the selected features to predict whether documents are AI-generated. Crucially, Ghostbuster does not require access to token probabilities from the target model, making it useful for detecting text generated by black-box models or unknown model versions. In conjunction with our model, we release three new datasets of human-and AI-generated text as detection benchmarks in the domains of student essays, creative writing, and news articles. We compare Ghostbuster to a variety of existing detectors, including DetectGPT and GPTZero, as well as a new RoBERTa baseline. Ghostbuster achieves 99.0 F1 when evaluated across domains, which is 5.9 F1 higher than the best preexisting model. It also outperforms all previous approaches in generalization across writing domains (+7.5 F1), prompting strategies (+2.1 F1), and language models (+4.4 F1). We also analyze the robustness of our system to a variety of perturbations and paraphrasing attacks and evaluate its performance on documents written by non-native English speakers. Language models such as ChatGPT are capable of producing a wide range of fluent text that closely approximates human language use. However, the proliferation of these models has raised concerns about the authenticity and trustworthiness of text across a variety of domains. For example, fears that students are submitting assignments ghostwritten by language models have led many schools to adapt by restricting the use of ChatGPT in classrooms and homework assignments (Heaven, 2023). Meanwhile, because language models are prone to factual errors and hallucination, readers may desire to know if such tools have been used to ghostwrite news articles or other informative text when deciding whether or not to trust a source.

The strong performance (Chowdhery et al., 2022; Nostalgebraist, 2022; OpenAI, 2023; Google, 2023), and emergent abilities (Wei et al., 2022) of modern language models (LMs) depend on self-supervised pretraining on massive text datasets. All model developers implicitly or explicitly decide the composition of these datasets: what data sources to include, whether to filter for attributes such as quality and toxicity, and when to gather new documents. While many of the most prominent models do not document their curation procedures (OpenAI, 2023; Google, 2023), or only document which procedures they used (Brown et al., 2020; Nostalgebraist, 2022; Scao et al., 2022; Touvron et al., 2023), they rarely document why they chose those protocols or what effect they had. This documentation debt leaves practitioners to be guided by intuitions and precedents, neither thoroughly evaluated (Bandy and Vincent, 2021; Sambasivan et al., 2021). Given the outsized and fundamental role of pretraining data in modern LMs, we believe this neglectful practice has detracted from responsible data use and hampered effective model development (Rogers, 2021; Gebru et al., 2021; Bender and Friedman, 2018). Among the small number of general-purpose LMs dominating community use and discussion, the prevailing focus has been on the scale of pretraining data and number of optimization steps (Brown et al., 2020; Nostalgebraist, 2022; Google, 2023). In this work, we systematically test how common data design decisions affect model performance--specifically: the time of collection, content filtering strategy (toxicity/quality), and domain composition. We study the impacts in two ways.

Providing system-generated explanations for recommendations represents an important step towards transparent and trustworthy recommender systems. Explainable recommender systems provide a human-understandable rationale for their outputs. Over the last two decades, explainable recommendation has attracted much attention in the recommender systems research community. This paper aims to provide a comprehensive review of research efforts on visual explanation in recommender systems. More concretely, we systematically review the literature on explanations in recommender systems based on four dimensions, namely explanation goal, explanation scope, explanation style, and explanation format. Recognizing the importance of visualization, we approach the recommender system literature from the angle of explanatory visualizations, that is using visualizations as a display style of explanation. As a result, we derive a set of guidelines that might be constructive for designing explanatory visualizations in recommender systems and identify perspectives for future work in this field. The aim of this review is to help recommendation researchers and practitioners better understand the potential of visually explainable recommendation research and to support them in the systematic design of visual explanations in current and future recommender systems.

The increased deployment of LMs for real-world tasks involving knowledge and facts makes it important to understand model epistemology: what LMs think they know, and how their attitudes toward that knowledge are affected by language use in their inputs. Here, we study an aspect of model epistemology: how epistemic markers of certainty, uncertainty, or evidentiality like "I'm sure it's", "I think it's", or "Wikipedia says it's" affect models, and whether they contribute to model failures. We develop a typology of epistemic markers and inject 50 markers into prompts for question answering. We find that LMs are highly sensitive to epistemic markers in prompts, with accuracies varying more than 80%. Surprisingly, we find that expressions of high certainty result in a 7% decrease in accuracy as compared to low certainty expressions; similarly, factive verbs hurt performance, while evidentials benefit performance. Our analysis of a popular pretraining dataset shows that these markers of uncertainty are associated with answers on question-answering websites, while markers of certainty are associated with questions. These associations may suggest that the behavior of LMs is based on mimicking observed language use, rather than truly reflecting epistemic uncertainty.

Applying reinforcement learning (RL) to combinatorial optimization problems is attractive as it removes the need for expert knowledge or pre-solved instances. However, it is unrealistic to expect an agent to solve these (often NP-)hard problems in a single shot at inference due to their inherent complexity. Thus, leading approaches often implement additional search strategies, from stochastic sampling and beam-search to explicit fine-tuning. In this paper, we argue for the benefits of learning a population of complementary policies, which can be simultaneously rolled out at inference. To this end, we introduce Poppy, a simple training procedure for populations. Instead of relying on a predefined or hand-crafted notion of diversity, Poppy induces an unsupervised specialization targeted solely at maximizing the performance of the population. We show that Poppy produces a set of complementary policies, and obtains state-of-the-art RL results on four popular NP-hard problems: traveling salesman, capacitated vehicle routing, 0-1 knapsack, and job-shop scheduling.

Recent empirical studies found that arrivals in many real queueing systems exhibit a clustering or self-exciting behavior; that is, an arrival may increase the possibility of new arrivals. In some cases, such clustering behavior is intrinsic to the underlying system. For example, in the stock market, it is a common practice to split a large order into small child orders to reduce transaction cost. As a consequence, one observed arriving order may be followed by a sequence of other child orders (Abergel and Jedidi, 2015). As a natural extension of the classic Poisson process, Hawkes process has been used to model arrivals with self-excitement such as order flow in stock market (Abergel and Jedidi, 2015), infected patients during pandemic (Bertozzi et al., 2020), and the internet traffic in social media (Zhao et al., 2015). To understand the impact of self-excitement in the arrival process on the long-run performance of service systems, Koops et al. (2018) and Daw and Pender (2018) provided analytic solutions to steady-state moments on the number of people in system for different infinite-server systems with Hawkes arrivals.

Quadratic programming is a ubiquitous prototype in convex programming. Many combinatorial optimizations on graphs and machine learning problems can be formulated as quadratic programming; for example, Support Vector Machines (SVMs). Linear and kernel SVMs have been among the most popular models in machine learning over the past three decades, prior to the deep learning era. Generally, a quadratic program has an input size of $\Theta(n^2)$, where $n$ is the number of variables. Assuming the Strong Exponential Time Hypothesis ($\textsf{SETH}$), it is known that no $O(n^{2-o(1)})$ algorithm exists (Backurs, Indyk, and Schmidt, NIPS'17). However, problems such as SVMs usually feature much smaller input sizes: one is given $n$ data points, each of dimension $d$, with $d \ll n$. Furthermore, SVMs are variants with only $O(1)$ linear constraints. This suggests that faster algorithms are feasible, provided the program exhibits certain underlying structures. In this work, we design the first nearly-linear time algorithm for solving quadratic programs whenever the quadratic objective has small treewidth or admits a low-rank factorization, and the number of linear constraints is small. Consequently, we obtain a variety of results for SVMs: * For linear SVM, where the quadratic constraint matrix has treewidth $\tau$, we can solve the corresponding program in time $\widetilde O(n\tau^{(\omega+1)/2}\log(1/\epsilon))$; * For linear SVM, where the quadratic constraint matrix admits a low-rank factorization of rank-$k$, we can solve the corresponding program in time $\widetilde O(nk^{(\omega+1)/2}\log(1/\epsilon))$; * For Gaussian kernel SVM, where the data dimension $d = \Theta(\log n)$ and the squared dataset radius is small, we can solve it in time $O(n^{1+o(1)}\log(1/\epsilon))$. We also prove that when the squared dataset radius is large, then $\Omega(n^{2-o(1)})$ time is required.

Deep learning has shown impressive results in a variety of time series forecasting tasks, where modeling the conditional distribution of the future given the past is the essence. However, when this conditional distribution is non-stationary, it poses challenges for these models to learn consistently and to predict accurately. In this work, we propose a new method to model non-stationary conditional distributions over time by clearly decoupling stationary conditional distribution modeling from non-stationary dynamics modeling. Our method is based on a Bayesian dynamic model that can adapt to conditional distribution changes and a deep conditional distribution model that handles multivariate time series using a factorized output space. Our experimental results on synthetic and real-world datasets show that our model can adapt to non-stationary time series better than state-of-the-art deep learning solutions.

Machine learning (ML) may be oblivious to human bias but it is not immune to its perpetuation. Marginalisation and iniquitous group representation are often traceable in the very data used for training, and may be reflected or even enhanced by the learning models. In the present work, we aim at clarifying the role played by data geometry in the emergence of ML bias. We introduce an exactly solvable high-dimensional model of data imbalance, where parametric control over the many bias-inducing factors allows for an extensive exploration of the bias inheritance mechanism. Through the tools of statistical physics, we analytically characterise the typical properties of learning models trained in this synthetic framework and obtain exact predictions for the observables that are commonly employed for fairness assessment. Despite the simplicity of the data model, we retrace and unpack typical unfairness behaviour observed on real-world datasets. We also obtain a detailed analytical characterisation of a class of bias mitigation strategies. We first consider a basic loss-reweighing scheme, which allows for an implicit minimisation of different unfairness metrics, and quantify the incompatibilities between some existing fairness criteria. Then, we consider a novel mitigation strategy based on a matched inference approach, consisting in the introduction of coupled learning models. Our theoretical analysis of this approach shows that the coupled strategy can strike superior fairness-accuracy trade-offs.

Model selection in the context of bandit optimization is a challenging problem, as it requires balancing exploration and exploitation not only for action selection, but also for model selection. One natural approach is to rely on online learning algorithms that treat different models as experts. Existing methods, however, scale poorly ($\text{poly}M$) with the number of models $M$ in terms of their regret. Our key insight is that, for model selection in linear bandits, we can emulate full-information feedback to the online learner with a favorable bias-variance trade-off. This allows us to develop ALEXP, which has an exponentially improved ($\log M$) dependence on $M$ for its regret. ALEXP has anytime guarantees on its regret, and neither requires knowledge of the horizon $n$, nor relies on an initial purely exploratory stage. Our approach utilizes a novel time-uniform analysis of the Lasso, establishing a new connection between online learning and high-dimensional statistics.