Although Large Language Models (LLMs) have demonstrated significant capabilities in executing complex tasks in a zero-shot manner, they are susceptible to jailbreak attacks and can be manipulated to produce harmful outputs. Recently, a growing body of research has categorized jailbreak attacks into token-level and prompt-level attacks. However, previous work primarily overlooks the diverse key factors of jailbreak attacks, with most studies concentrating on LLM vulnerabilities and lacking exploration of defense-enhanced LLMs. To address these issues, we evaluate the impact of various attack settings on LLM performance and provide a baseline benchmark for jailbreak attacks, encouraging the adoption of a standardized evaluation framework. Specifically, we evaluate the eight key factors of implementing jailbreak attacks on LLMs from both target-level and attack-level perspectives. We further conduct seven representative jailbreak attacks on six defense methods across two widely used datasets, encompassing approximately 320 experiments with about 50,000 GPU hours on A800-80G. Our experimental results highlight the need for standardized benchmarking to evaluate these attacks on defense-enhanced LLMs.

Members of the Human-Robot Interaction (HRI) and Artificial Intelligence (AI) communities have proposed Large Language Models (LLMs) as a promising resource for robotics tasks such as natural language interactions, doing household and workplace tasks, approximating `common sense reasoning', and modeling humans. However, recent research has raised concerns about the potential for LLMs to produce discriminatory outcomes and unsafe behaviors in real-world robot experiments and applications. To address these concerns, we conduct an HRI-based evaluation of discrimination and safety criteria on several highly-rated LLMs. Our evaluation reveals that LLMs currently lack robustness when encountering people across a diverse range of protected identity characteristics (e.g., race, gender, disability status, nationality, religion, and their intersections), producing biased outputs consistent with directly discriminatory outcomes -- e.g. `gypsy' and `mute' people are labeled untrustworthy, but not `european' or `able-bodied' people. Furthermore, we test models in settings with unconstrained natural language (open vocabulary) inputs, and find they fail to act safely, generating responses that accept dangerous, violent, or unlawful instructions -- such as incident-causing misstatements, taking people's mobility aids, and sexual predation. Our results underscore the urgent need for systematic, routine, and comprehensive risk assessments and assurances to improve outcomes and ensure LLMs only operate on robots when it is safe, effective, and just to do so. Data and code will be made available.

Powered by remarkable advancements in Large Language Models (LLMs), Multimodal Large Language Models (MLLMs) demonstrate impressive capabilities in manifold tasks. However, the practical application scenarios of MLLMs are intricate, exposing them to potential malicious instructions and thereby posing safety risks. While current benchmarks do incorporate certain safety considerations, they often lack comprehensive coverage and fail to exhibit the necessary rigor and robustness. For instance, the common practice of employing GPT-4V as both the evaluator and a model to be evaluated lacks credibility, as it tends to exhibit a bias toward its own responses. In this paper, we present MLLMGuard, a multidimensional safety evaluation suite for MLLMs, including a bilingual image-text evaluation dataset, inference utilities, and a lightweight evaluator. MLLMGuard's assessment comprehensively covers two languages (English and Chinese) and five important safety dimensions (Privacy, Bias, Toxicity, Truthfulness, and Legality), each with corresponding rich subtasks. Focusing on these dimensions, our evaluation dataset is primarily sourced from platforms such as social media, and it integrates text-based and image-based red teaming techniques with meticulous annotation by human experts. This can prevent inaccurate evaluation caused by data leakage when using open-source datasets and ensures the quality and challenging nature of our benchmark. Additionally, a fully automated lightweight evaluator termed GuardRank is developed, which achieves significantly higher evaluation accuracy than GPT-4. Our evaluation results across 13 advanced models indicate that MLLMs still have a substantial journey ahead before they can be considered safe and responsible.

Multi-relational temporal graphs are powerful tools for modeling real-world data, capturing the evolving and interconnected nature of entities over time. Recently, many novel models are proposed for ML on such graphs intensifying the need for robust evaluation and standardized benchmark datasets. However, the availability of such resources remains scarce and evaluation faces added complexity due to reproducibility issues in experimental protocols. To address these challenges, we introduce Temporal Graph Benchmark 2.0 (TGB 2.0), a novel benchmarking framework tailored for evaluating methods for predicting future links on Temporal Knowledge Graphs and Temporal Heterogeneous Graphs with a focus on large-scale datasets, extending the Temporal Graph Benchmark. TGB 2.0 facilitates comprehensive evaluations by presenting eight novel datasets spanning five domains with up to 53 million edges. TGB 2.0 datasets are significantly larger than existing datasets in terms of number of nodes, edges, or timestamps. In addition, TGB 2.0 provides a reproducible and realistic evaluation pipeline for multi-relational temporal graphs. Through extensive experimentation, we observe that 1) leveraging edge-type information is crucial to obtain high performance, 2) simple heuristic baselines are often competitive with more complex methods, 3) most methods fail to run on our largest datasets, highlighting the need for research on more scalable methods.

The compelling goal of eradicating undesirable data behaviors, while preserving usual model functioning, underscores the significance of machine unlearning within the domain of large language models (LLMs). Recent research has begun to approach LLM unlearning via gradient ascent (GA) -- increasing the prediction risk for those training strings targeted to be unlearned, thereby erasing their parameterized responses. Despite their simplicity and efficiency, we suggest that GA-based methods face the propensity towards excessive unlearning, resulting in various undesirable model behaviors, such as catastrophic forgetting, that diminish their practical utility. In this paper, we suggest a set of metrics that can capture multiple facets of real-world utility and propose several controlling methods that can regulate the extent of excessive unlearning. Accordingly, we suggest a general framework to better reflect the practical efficacy of various unlearning methods -- we begin by controlling the unlearning procedures/unlearned models such that no excessive unlearning occurs and follow by the evaluation for unlearning efficacy. Our experimental analysis on established benchmarks revealed that GA-based methods are far from perfect in practice, as strong unlearning is at the high cost of hindering the model utility. We conclude that there is still a long way towards practical and effective LLM unlearning, and more efforts are required in this field.

The evolution of AI is set to profoundly reshape the future. The European Union, recognizing this impending prominence, has enacted the AI Act, regulating market access for AI-based systems. A salient feature of the Act is to guard democratic and humanistic values by focusing regulation on transparency, explainability, and the human ability to understand and control AI systems. Hereby, the EU AI Act does not merely specify technological requirements for AI systems. The EU issues a democratic call for human-centered AI systems and, in turn, an interdisciplinary research agenda for human-centered innovation in AI development. Without robust methods to assess AI systems and their effect on individuals and society, the EU AI Act may lead to repeating the mistakes of the General Data Protection Regulation of the EU and to rushed, chaotic, ad-hoc, and ambiguous implementation, causing more confusion than lending guidance. Moreover, determined research activities in Human-AI interaction will be pivotal for both regulatory compliance and the advancement of AI in a manner that is both ethical and effective. Such an approach will ensure that AI development aligns with human values and needs, fostering a technology landscape that is innovative, responsible, and an integral part of our society.

Large Language Models have taken the cognitive science world by storm. It is perhaps timely now to take stock of the various research paradigms that have been used to make scientific inferences about ``cognition" in these models or about human cognition. We review several emerging research paradigms -- GPT-ology, LLMs-as-computational-models, and ``silicon sampling" -- and review recent papers that have used LLMs under these paradigms. In doing so, we discuss their claims as well as challenges to scientific inference under these various paradigms. We highlight several outstanding issues about LLMs that have to be addressed to push our science forward: closed-source vs open-sourced models; (the lack of visibility of) training data; and reproducibility in LLM research, including forming conventions on new task ``hyperparameters" like instructions and prompts.

As soon as Apple announced its plans to inject generative AI into the iPhone, it was as good as official: The technology is now all but unavoidable. AI has already colonized web search, appearing in Google and Bing. OpenAI, the 80 billion start-up that has partnered with Apple and Microsoft, feels ubiquitous; the auto-generated products of its ChatGPTs and DALL-Es are everywhere. Rarely has a technology risen--or been forced--into prominence amid such controversy and consumer anxiety. Certainly, some Americans are excited about AI, though a majority said in a recent survey, for instance, that they are concerned AI will increase unemployment; in another, three out of four said they believe it will be abused to interfere with the upcoming presidential election.

Predators active on the dark web are increasingly using artificial intelligence to create sexually explicit images of children, fixating especially on "star" victims, child safety experts warn. Child safety groups tracking the activity of predators chatting in dark web forums say they are increasingly finding conversations about creating new images based on older child sexual abuse material (CSAM). Many of these predators using AI obsess over child victims referred to as "stars" in predator communities for the popularity of their images. "The communities of people who trade this material get infatuated with individual children," said Sarah Gardner, chief executive officer of the Heat Initiative, a Los Angeles non-profit focused on child protection. "They want more content of those children, which AI has now allowed them to do."

Elon Musk has withdrawn his lawsuit against OpenAI, a day before a judge was set to hear a request for dismissal. Musk sued OpenAI, saying its founders had violated its nonprofit status, to become a de-facto part of Microsoft. OpenAI said there was no such violation, and the lawsuit was likely a way for Musk to gain access to its secrets. Despite ending the suit, Musk might be nursing this grudge, tweeting if Apple integrates OpenAI's tools into its software, he'll ban iPhones from his companies. You can't mirror your iPhone while mirroring your Mac on Apple Vision Pro Netflix drops a proper trailer for Arcane's second (and last) season Apple Intelligence: What devices and features will actually be supported?