Related papers: Decoding Safety Feedback from Diverse Raters: A Da…
Addressing public safety effectively requires incorporating diverse stakeholder perspectives, particularly those of the community, which are often underrepresented compared to other stakeholders. This study presents a comprehensive analysis…
We evaluate the robustness of several large language models on multiple datasets. Robustness here refers to the relative insensitivity of the model's answers to meaning-preserving variants of their input. Benchmark datasets are constructed…
Incorporating every annotator's perspective is crucial for unbiased data modeling. Annotator fatigue and changing opinions over time can distort dataset annotations. To combat this, we propose to learn a more accurate representation of…
With the rapid evolution of large language models (LLMs), new and hard-to-predict harmful capabilities are emerging. This requires developers to be able to identify risks through the evaluation of "dangerous capabilities" in order to…
Text-based safety classifiers are widely used for content moderation and increasingly to tune generative language model behavior - a topic of growing concern for the safety of digital assistants and chatbots. However, different policies…
Recent work introduced the model of learning from discriminative feature feedback, in which a human annotator not only provides labels of instances, but also identifies discriminative features that highlight important differences between…
The prevalence and impact of toxic discussions online have made content moderation crucial.Automated systems can play a vital role in identifying toxicity, and reducing the reliance on human moderation.Nevertheless, identifying toxic…
Understanding human driving behavior is crucial to develop autonomous vehicles' algorithms. However, most low level automation, such as the one in advanced driving assistance systems (ADAS), is based on objective safety measures, which are…
Safety and responsibility evaluations of advanced AI models are a critical but developing field of research and practice. In the development of Google DeepMind's advanced AI models, we innovated on and applied a broad set of approaches to…
Ensuring safety in human-robot interaction (HRI) is essential to foster user trust and enable the broader adoption of robotic systems. Traditional safety models primarily rely on sensor-based measures, such as relative distance and…
AI Safety Moderation (ASM) classifiers are designed to moderate content on social media platforms and to serve as guardrails that prevent Large Language Models (LLMs) from being fine-tuned on unsafe inputs. Owing to their potential for…
As large language models (LLMs) become increasingly integrated into real-world applications, scalable and rigorous safety evaluation is essential. This paper introduces Aymara AI, a programmatic platform for generating and administering…
We measure the association between generative AI (GAI) tool adoption and four metrics spanning security operations, information protection, and endpoint management: 1) number of security alerts per incident, 2) probability of security…
AI incident reporting requirements are emerging in regulation and policy, yet no operational criteria exist for determining when a detected AI incident warrants escalation beyond national handling to international coordination. This paper…
Modeling and calibrating the fidelity of synthetic data is paramount in shaping the future of safe and reliable self-driving technology by offering a cost-effective and scalable alternative to real-world data collection. We focus on its…
Artificial intelligence (AI) systems are deployed as collaborators in human decision-making. Yet, evaluation practices focus primarily on model accuracy rather than whether human-AI teams are prepared to collaborate safely and effectively.…
Natural Language Inference (NLI) is foundational for evaluating language understanding in AI. However, progress has plateaued, with models failing on ambiguous examples and exhibiting poor generalization. We argue that this stems from…
In this work, we introduce a novel metric for auditing group fairness in ranked lists. Our approach offers two benefits compared to the state of the art. First, we offer a blueprint for modeling of user attention. Rather than assuming a…
Security evaluations inherently depend on stable identifiers. Any finding, audit, or regulatory decision must remain attached to the specific artifact it pertains to. Continuously updated artificial intelligence systems violate this core…
In commonsense generation, given a set of input concepts, a model must generate a response that is not only commonsense bearing, but also capturing multiple diverse viewpoints. Numerous evaluation metrics based on form- and content-level…