Artificial intelligence (AI) has matured as a technology, necessitating the development of responsibility frameworks that are fair, inclusive, trustworthy, safe and secure, transparent, and accountable. By establishing such frameworks, we can harness the full potential of AI while mitigating its risks, particularly in high-risk scenarios. This requires the design of responsible AI systems based on trustworthy AI technologies and ethical principles, with the aim of ensuring auditability and accountability throughout their design, development, and deployment, adhering to domain-specific regulations and standards. This paper explores the concept of a responsible AI system from a holistic perspective, which encompasses four key dimensions: 1) regulatory context; 2) trustworthy AI technology along with standardization and assessments; 3) auditability and accountability; and 4) AI governance. The aim of this paper is double. First, we analyze and understand these four dimensions and their interconnections in the form of an analysis and overview. Second, the final goal of the paper is to propose a roadmap in the design of responsible AI systems, ensuring that they can gain society's trust. To achieve this trustworthiness, this paper also fosters interdisciplinary discussions on the ethical, legal, social, economic, and cultural aspects of AI from a global governance perspective. Last but not least, we also reflect on the current state and those aspects that need to be developed in the near future, as ten lessons learned.

As large language models (LLMs) are becoming more capable and widespread, the study of their failure cases is becoming increasingly important. Recent advances in standardizing, measuring, and scaling test-time compute suggest new methodologies for optimizing models to achieve high performance on hard tasks. In this paper, we apply these advances to the task of model jailbreaking: eliciting harmful responses from aligned LLMs. We develop an adversarial reasoning approach to automatic jailbreaking via test-time computation that achieves SOTA attack success rates (ASR) against many aligned LLMs, even the ones that aim to trade inference-time compute for adversarial robustness. Our approach introduces a new paradigm in understanding LLM vulnerabilities, laying the foundation for the development of more robust and trustworthy AI systems.

Machine learning models are routinely trained on a mixture of different data domains. Different domain weights yield very different downstream performances. We propose the Soup-of-Experts, a novel architecture that can instantiate a model at test time for any domain weights with minimal computational cost and without re-training the model. Our architecture consists of a bank of expert parameters, which are linearly combined to instantiate one model. We learn the linear combination coefficients as a function of the input domain weights. To train this architecture, we sample random domain weights, instantiate the corresponding model, and backprop through one batch of data sampled with these domain weights. We demonstrate how our approach obtains small specialized models on several language modeling tasks quickly. Soup-of-Experts are particularly appealing when one needs to ship many different specialist models quickly under a model size constraint.

Navigating densely vegetated environments poses significant challenges for autonomous ground vehicles. Learning-based systems typically use prior and in-situ data to predict terrain traversability but often degrade in performance when encountering out-of-distribution elements caused by rapid environmental changes or novel conditions. This paper presents a novel, lidar-only, online adaptive traversability estimation (TE) method that trains a model directly on the robot using self-supervised data collected through robot-environment interaction. The proposed approach utilises a probabilistic 3D voxel representation to integrate lidar measurements and robot experience, creating a salient environmental model. To ensure computational efficiency, a sparse graph-based representation is employed to update temporarily evolving voxel distributions. Extensive experiments with an unmanned ground vehicle in natural terrain demonstrate that the system adapts to complex environments with as little as 8 minutes of operational data, achieving a Matthews Correlation Coefficient (MCC) score of 0.63 and enabling safe navigation in densely vegetated environments. This work examines different training strategies for voxel-based TE methods and offers recommendations for training strategies to improve adaptability. The proposed method is validated on a robotic platform with limited computational resources (25W GPU), achieving accuracy comparable to offline-trained models while maintaining reliable performance across varied environments.

Interpreting and controlling the internal mechanisms of large language models (LLMs) is crucial for improving their trustworthiness and utility. Recent efforts have primarily focused on identifying and manipulating neurons by establishing discrete mappings between neurons and semantic concepts. However, such mappings struggle to handle the inherent polysemanticity in LLMs, where individual neurons encode multiple, distinct concepts. This makes precise control challenging and complicates downstream interventions. Through an in-depth analysis of both encoder and decoder-based LLMs across multiple text classification datasets, we uncover that while individual neurons encode multiple concepts, their activation magnitudes vary across concepts in distinct, Gaussian-like patterns. Building on this insight, we introduce NeuronLens, a novel range-based interpretation and manipulation framework that provides a finer view of neuron activation distributions to localize concept attribution within a neuron. Extensive empirical evaluations demonstrate that NeuronLens significantly reduces unintended interference, while maintaining precise control for manipulation of targeted concepts, outperforming existing methods.

The prevalence of drawing causal conclusions from observational studies has raised concerns about potential exaggeration in science communication. While some believe causal language should only apply to randomized controlled trials, others argue that rigorous methods can justify causal claims in observational studies. Ideally, causal language should align with the strength of the evidence. However, through the analysis of over 80,000 observational study abstracts using computational linguistic and regression methods, we found that causal language is more frequently used by less experienced authors, smaller research teams, male last authors, and authors from countries with higher uncertainty avoidance indices. These findings suggest that the use of causal language may be influenced by external factors such as the sociocultural backgrounds of authors and the dynamics of research collaboration. This newly identified link deepens our understanding of how such factors help shape scientific conclusions in causal inference and science communication.

Evaluations of large language model (LLM) risks and capabilities are increasingly being incorporated into AI risk management and governance frameworks. Currently, most risk evaluations are conducted by designing inputs that elicit harmful behaviors from the system. However, a fundamental limitation of this approach is that the harmfulness of the behaviors identified during any particular evaluation can only lower bound the model's worst-possible-case behavior. As a complementary method for eliciting harmful behaviors, we propose evaluating LLMs with model tampering attacks which allow for modifications to latent activations or weights. We pit state-of-the-art techniques for removing harmful LLM capabilities against a suite of 5 input-space and 6 model tampering attacks. In addition to benchmarking these methods against each other, we show that (1) model resilience to capability elicitation attacks lies on a low-dimensional robustness subspace; (2) the attack success rate of model tampering attacks can empirically predict and offer conservative estimates for the success of held-out input-space attacks; and (3) state-of-the-art unlearning methods can easily be undone within 16 steps of fine-tuning. Together these results highlight the difficulty of removing harmful LLM capabilities and show that model tampering attacks enable substantially more rigorous evaluations than input-space attacks alone. We release models at https://huggingface.co/LLM-GAT

Inferring physical laws from data is a central challenge in science and engineering, including but not limited to healthcare, physical sciences, biosciences, social sciences, sustainability, climate, and robotics. Deep networks offer high-accuracy results but lack interpretability, prompting interest in models built from simple components. The Sparse Identification of Nonlinear Dynamics (SINDy) method has become the go-to approach for building such modular and interpretable models. SINDy leverages sparse regression with L1 regularization to identify key terms from a library of candidate functions. However, SINDy's choice of candidate library and optimization method requires significant technical expertise, limiting its widespread applicability. This work introduces Al-Khwarizmi, a novel agentic framework for physical law discovery from data, which integrates foundational models with SINDy. Leveraging LLMs, VLMs, and Retrieval-Augmented Generation (RAG), our approach automates physical law discovery, incorporating prior knowledge and iteratively refining candidate solutions via reflection. Al-Khwarizmi operates in two steps: it summarizes system observations-comprising textual descriptions, raw data, and plots-followed by a secondary step that generates candidate feature libraries and optimizer configurations to identify hidden physics laws correctly. Evaluating our algorithm on over 198 models, we demonstrate state-of-the-art performance compared to alternatives, reaching a 20 percent increase against the best-performing alternative.

Multimodal fusion leverages information across modalities to learn better feature representations with the goal of improving performance in fusion-based tasks. However, multimodal datasets, especially in medical settings, are typically smaller than their unimodal counterparts, which can impede the performance of multimodal models. Additionally, the increase in the number of modalities is often associated with an overall increase in the size of the multimodal network, which may be undesirable in medical use cases. Utilizing smaller unimodal encoders may lead to sub-optimal performance, particularly when dealing with high-dimensional clinical data. In this paper, we propose the Modality-INformed knowledge Distillation (MIND) framework, a multimodal model compression approach based on knowledge distillation that transfers knowledge from ensembles of pre-trained deep neural networks of varying sizes into a smaller multimodal student. The teacher models consist of unimodal networks, allowing the student to learn from diverse representations. MIND employs multi-head joint fusion models, as opposed to single-head models, enabling the use of unimodal encoders in the case of unimodal samples without requiring imputation or masking of absent modalities. As a result, MIND generates an optimized multimodal model, enhancing both multimodal and unimodal representations. It can also be leveraged to balance multimodal learning during training. We evaluate MIND on binary and multilabel clinical prediction tasks using time series data and chest X-ray images. Additionally, we assess the generalizability of the MIND framework on three non-medical multimodal multiclass datasets. Experimental results demonstrate that MIND enhances the performance of the smaller multimodal network across all five tasks, as well as various fusion methods and multimodal architectures, compared to state-of-the-art baselines.

The rise of Agent AI and Large Language Model-powered Multi-Agent Systems (LLM-MAS) has underscored the need for responsible and dependable system operation. Tools like LangChain and Retrieval-Augmented Generation have expanded LLM capabilities, enabling deeper integration into MAS through enhanced knowledge retrieval and reasoning. However, these advancements introduce critical challenges: LLM agents exhibit inherent unpredictability, and uncertainties in their outputs can compound across interactions, threatening system stability. To address these risks, a human-centered design approach with active dynamic moderation is essential. Such an approach enhances traditional passive oversight by facilitating coherent inter-agent communication and effective system governance, allowing MAS to achieve desired outcomes more efficiently.