A core challenge in the development of increasingly capable AI systems is to make them safe and reliable by ensuring their behaviour is consistent with human values. This challenge, known as the alignment problem, does not merely apply to hypothetical future AI systems that may pose catastrophic risks; it already applies to current systems, such as large language models, whose potential for harm is rapidly increasing. In this paper, I assess whether we are on track to solve the alignment problem for large language models, and what that means for the safety of future AI systems. I argue that existing strategies for alignment are insufficient, because large language models remain vulnerable to adversarial attacks that can reliably elicit unsafe behaviour. I offer an explanation of this lingering vulnerability on which it is not simply a contingent limitation of current language models, but has deep technical ties to a crucial aspect of what makes these models useful and versatile in the first place -- namely, their remarkable aptitude to learn "in context" directly from user instructions. It follows that the alignment problem is not only unsolved for current AI systems, but may be intrinsically difficult to solve without severely undermining their capabilities. Furthermore, this assessment raises concerns about the prospect of ensuring the safety of future and more capable AI systems.

Although autonomous functioning facilitates deployment of robotic systems in domains that admit limited human oversight on our planet and beyond, finding correspondence between task requirements and autonomous capability is still an open challenge. Consequently, a number of methods for quantifying autonomy have been proposed over the last three decades, but to our knowledge all these have no discernment of sub-mode features of variation of autonomy and some are based on metrics that violet the Goodhart's law. This paper focuses on the full autonomous mode and proposes a task-requirements based autonomy assessment framework. The framework starts by establishing robot task characteristics from which three autonomy metrics, namely requisite capability, reliability and responsiveness, and functions for determining autonomy as a two-part measure, namely of level of autonomy and degree of autonomy are derived. These characteristics are founded on the realization that robots ultimately replace human skilled workers, to find a mapping between human job and robot task characteristics. The distinction between level and degree of autonomy stemmed from the acknowledgment that autonomy is not just a question of existence, but also one of performance of requisite capability. When continuously monitored, the proposed metrics provide a means of monitoring the integrity of a system. The framework has been demonstrated on two case studies, namely autonomous vehicle at an on-road dynamic driving task and the DARPA subT challenge rules analysis. The framework provides not only a tool for quantifying autonomy, but also a regulatory interface and common language for autonomous systems developers and users.

Online forums encourage the exchange and discussion of different stances on many topics. Not only do they provide an opportunity to present one's own arguments, but may also gather a broad cross-section of others' arguments. However, the resulting long discussions are difficult to overview. This paper presents a novel unsupervised approach using large language models (LLMs) to generating indicative summaries for long discussions that basically serve as tables of contents. Our approach first clusters argument sentences, generates cluster labels as abstractive summaries, and classifies the generated cluster labels into argumentation frames resulting in a two-level summary. Based on an extensively optimized prompt engineering approach, we evaluate 19~LLMs for generative cluster labeling and frame classification. To evaluate the usefulness of our indicative summaries, we conduct a purpose-driven user study via a new visual interface called Discussion Explorer: It shows that our proposed indicative summaries serve as a convenient navigation tool to explore long discussions.

Neural text detectors are models trained to detect whether a given text was generated by a language model or written by a human. In this paper, we investigate three simple and resource-efficient strategies (parameter tweaking, prompt engineering, and character-level mutations) to alter texts generated by GPT-3.5 that are unsuspicious or unnoticeable for humans but cause misclassification by neural text detectors. The results show that especially parameter tweaking and character-level mutations are effective strategies.

We present Multi-EuP, a new multilingual benchmark dataset, comprising 22K multi-lingual documents collected from the European Parliament, spanning 24 languages. This dataset is designed to investigate fairness in a multilingual information retrieval (IR) context to analyze both language and demographic bias in a ranking context. It boasts an authentic multilingual corpus, featuring topics translated into all 24 languages, as well as cross-lingual relevance judgments. Furthermore, it offers rich demographic information associated with its documents, facilitating the study of demographic bias. We report the effectiveness of Multi-EuP for benchmarking both monolingual and multilingual IR. We also conduct a preliminary experiment on language bias caused by the choice of tokenization strategy.

Framing bias plays a significant role in exacerbating political polarization by distorting the perception of actual events. Media outlets with divergent political stances often use polarized language in their reporting of the same event. We propose a new loss function that encourages the model to minimize the polarity difference between the polarized input articles to reduce framing bias. Specifically, our loss is designed to jointly optimize the model to map polarity ends bidirectionally. Our experimental results demonstrate that incorporating the proposed polarity minimization loss leads to a substantial reduction in framing bias when compared to a BART-based multi-document summarization model. Notably, we find that the effectiveness of this approach is most pronounced when the model is trained to minimize the polarity loss associated with informational framing bias (i.e., skewed selection of information to report).

Hallucination detection is a critical step toward understanding the trustworthiness of modern language models (LMs). To achieve this goal, we re-examine existing detection approaches based on the self-consistency of LMs and uncover two types of hallucinations resulting from 1) question-level and 2) model-level, which cannot be effectively identified through self-consistency check alone. Building upon this discovery, we propose a novel sampling-based method, i.e., semantic-aware cross-check consistency (SAC$^3$) that expands on the principle of self-consistency checking. Our SAC$^3$ approach incorporates additional mechanisms to detect both question-level and model-level hallucinations by leveraging advances including semantically equivalent question perturbation and cross-model response consistency checking. Through extensive and systematic empirical analysis, we demonstrate that SAC$^3$ outperforms the state of the art in detecting both non-factual and factual statements across multiple question-answering and open-domain generation benchmarks.

Cooperative multi-agent multi-armed bandits (CMA2B) consider the collaborative efforts of multiple agents in a shared multi-armed bandit game. We study latent vulnerabilities exposed by this collaboration and consider adversarial attacks on a few agents with the goal of influencing the decisions of the rest. More specifically, we study adversarial attacks on CMA2B in both homogeneous settings, where agents operate with the same arm set, and heterogeneous settings, where agents have distinct arm sets. In the homogeneous setting, we propose attack strategies that, by targeting just one agent, convince all agents to select a particular target arm $T-o(T)$ times while incurring $o(T)$ attack costs in $T$ rounds. In the heterogeneous setting, we prove that a target arm attack requires linear attack costs and propose attack strategies that can force a maximum number of agents to suffer linear regrets while incurring sublinear costs and only manipulating the observations of a few target agents. Numerical experiments validate the effectiveness of our proposed attack strategies.

Although large language models (LLMs) are widely deployed, the data used to train them is rarely disclosed. Given the incredible scale of this data, up to trillions of tokens, it is all but certain that it includes potentially problematic text such as copyrighted materials, personally identifiable information, and test data for widely reported reference benchmarks. However, we currently have no way to know which data of these types is included or in what proportions. In this paper, we study the pretraining data detection problem: given a piece of text and black-box access to an LLM without knowing the pretraining data, can we determine if the model was trained on the provided text? To facilitate this study, we introduce a dynamic benchmark WIKIMIA that uses data created before and after model training to support gold truth detection. We also introduce a new detection method Min-K% Prob based on a simple hypothesis: an unseen example is likely to contain a few outlier words with low probabilities under the LLM, while a seen example is less likely to have words with such low probabilities. Min-K% Prob can be applied without any knowledge about the pretraining corpus or any additional training, departing from previous detection methods that require training a reference model on data that is similar to the pretraining data. Moreover, our experiments demonstrate that Min-K% Prob achieves a 7.4% improvement on WIKIMIA over these previous methods. We apply Min-K% Prob to three real-world scenarios, copyrighted book detection, contaminated downstream example detection and privacy auditing of machine unlearning, and find it a consistently effective solution.

Developing and discovering new drugs is a complex and resource-intensive endeavor that often involves substantial costs, time investment, and safety concerns. A key aspect of drug discovery involves identifying novel drug-target (DT) interactions. Existing computational methods for predicting DT interactions have primarily focused on binary classification tasks, aiming to determine whether a DT pair interacts or not. However, protein-ligand interactions exhibit a continuum of binding strengths, known as binding affinity, presenting a persistent challenge for accurate prediction. In this study, we investigate various techniques employed in Drug Target Interaction (DTI) prediction and propose novel enhancements to enhance their performance. Our approaches include the integration of Protein Language Models (PLMs) and the incorporation of Contact Map information as an inductive bias within current models. Through extensive experimentation, we demonstrate that our proposed approaches outperform the baseline models considered in this study, presenting a compelling case for further development in this direction. We anticipate that the insights gained from this work will significantly narrow the search space for potential drugs targeting specific proteins, thereby accelerating drug discovery.