The rapid advancement of Large Vision-Language Models (LVLMs) has enhanced capabilities offering potential applications from content creation to productivity enhancement. Despite their innovative potential, LVLMs exhibit vulnerabilities, especially in generating potentially toxic or unsafe responses. Malicious actors can exploit these vulnerabilities to propagate toxic content in an automated (or semi-) manner, leveraging the susceptibility of LVLMs to deception via strategically crafted prompts without fine-tuning or compute-intensive procedures. Despite the red-teaming efforts and inherent potential risks associated with the LVLMs, exploring vulnerabilities of LVLMs remains nascent and yet to be fully addressed in a systematic manner. This study systematically examines the vulnerabilities of open-source LVLMs, including LLaVA, InstructBLIP, Fuyu, and Qwen, using adversarial prompt strategies that simulate real-world social manipulation tactics informed by social theories. Our findings show that (i) toxicity and insulting are the most prevalent behaviors, with the mean rates of 16.13% and 9.75%, respectively; (ii) Qwen-VL-Chat, LLaVA-v1.6-Vicuna-7b, and InstructBLIP-Vicuna-7b are the most vulnerable models, exhibiting toxic response rates of 21.50%, 18.30% and 17.90%, and insulting responses of 13.40%, 11.70% and 10.10%, respectively; (iii) prompting strategies incorporating dark humor and multimodal toxic prompt completion significantly elevated these vulnerabilities. Despite being fine-tuned for safety, these models still generate content with varying degrees of toxicity when prompted with adversarial inputs, highlighting the urgent need for enhanced safety mechanisms and robust guardrails in LVLM development.

Wang et al., 2024b; Wu et al., 2024; Zhang et al., 2024b), as Recent studies have revealed significant shown in Figure 1. Initial successes in LLM-based human gaps between LLM-based human simulations and simulations have been demonstrated across diverse fields, real-world observations, highlighting these dual including society, economics, policy, and psychology (Chen challenges. To address these gaps, we present a et al., 2024a; Li et al., 2024b;f; Lin et al., 2023; Park et al., comprehensive analysis of LLM limitations and 2023b; Yang et al., 2024b). Moreover, reliable LLM simulations our design issues, proposing targeted solutions can generate high-quality data for LLM training for both aspects. Furthermore, we explore future (Tang et al., 2024; Zhang et al., 2024a) and evaluate data directions that address both challenges simultaneously, quality (Chiang et al., 2024; Moniri et al., 2024; Xu et al., particularly in data collection, LLM generation, 2023b; Zheng et al., 2023b), serving as a data generator and and evaluation. To support further research evaluator (Gu et al., 2024; Li et al., 2024c; Son et al., 2024) in this field, we provide a curated collection of to enhance LLM pre-training and simulation abilities.

In human-robot teams, human situational awareness is the operator's conscious knowledge of the team's states, actions, plans and their environment. Appropriate human situational awareness is critical to successful human-robot collaboration. In human-robot teaming, it is often assumed that the best and required level of situational awareness is knowing everything at all times. This view is problematic, because what a human needs to know for optimal team performance varies given the dynamic environmental conditions, task context and roles and capabilities of team members. We explore this topic by interviewing 16 participants with active and repeated experience in diverse human-robot teaming applications. Based on analysis of these interviews, we derive a framework explaining the dynamic nature of required situational awareness in human-robot teaming. In addition, we identify a range of factors affecting the dynamic nature of required and actual levels of situational awareness (i.e., dynamic situational awareness), types of situational awareness inefficiencies resulting from gaps between actual and required situational awareness, and their main consequences. We also reveal various strategies, initiated by humans and robots, that assist in maintaining the required situational awareness. Our findings inform the implementation of accurate estimates of dynamic situational awareness and the design of user-adaptive human-robot interfaces. Therefore, this work contributes to the future design of more collaborative and effective human-robot teams.

Neural Architecture Search (NAS) is a powerful approach of automating the design of efficient neural architectures. In contrast to traditional NAS methods, recently proposed one-shot NAS methods prove to be more efficient in performing NAS. One-shot NAS works by generating a singular weight-sharing supernetwork that acts as a search space (container) of subnetworks. Despite its achievements, designing the one-shot search space remains a major challenge. In this work we propose a search space design strategy for Vision Transformer (ViT)-based architectures. In particular, we convert the Segment Anything Model (SAM) into a weight-sharing supernetwork called SuperSAM. Our approach involves automating the search space design via layer-wise structured pruning and parameter prioritization. While the structured pruning applies probabilistic removal of certain transformer layers, parameter prioritization performs weight reordering and slicing of MLP-blocks in the remaining layers. We train supernetworks on several datasets using the sandwich rule. For deployment, we enhance subnetwork discovery by utilizing a program autotuner to identify efficient subnetworks within the search space. The resulting subnetworks are 30-70% smaller in size compared to the original pre-trained SAM ViT-B, yet outperform the pretrained model. Our work introduces a new and effective method for ViT NAS search-space design.

We describe a method for modeling spatial context to enable video anomaly detection. The main idea is to discover regions that share similar object-level activities by clustering joint object attributes using Gaussian mixture models. We demonstrate that this straightforward approach, using orders of magnitude fewer parameters than competing models, achieves state-of-the-art performance in the challenging spatial-context-dependent Street Scene dataset. As a side benefit, the high-resolution discovered regions learned by the model also provide explainable normalcy maps for human operators without the need for any pre-trained segmentation model.

Recent studies have revealed that modern image and video quality assessment (IQA/VQA) metrics are vulnerable to adversarial attacks. An attacker can manipulate a video through preprocessing to artificially increase its quality score according to a certain metric, despite no actual improvement in visual quality. Most of the attacks studied in the literature are white-box attacks, while black-box attacks in the context of VQA have received less attention. Moreover, some research indicates a lack of transferability of adversarial examples generated for one model to another when applied to VQA. In this paper, we propose a cross-modal attack method, IC2VQA, aimed at exploring the vulnerabilities of modern VQA models. This approach is motivated by the observation that the low-level feature spaces of images and videos are similar. We investigate the transferability of adversarial perturbations across different modalities; specifically, we analyze how adversarial perturbations generated on a white-box IQA model with an additional CLIP module can effectively target a VQA model. The addition of the CLIP module serves as a valuable aid in increasing transferability, as the CLIP model is known for its effective capture of low-level semantics. Extensive experiments demonstrate that IC2VQA achieves a high success rate in attacking three black-box VQA models. We compare our method with existing black-box attack strategies, highlighting its superiority in terms of attack success within the same number of iterations and levels of attack strength. We believe that the proposed method will contribute to the deeper analysis of robust VQA metrics.

Wireless communications rely on path loss modeling, which is most effective when it includes the physical details of the propagation environment. Acquiring this data has historically been challenging, but geographic information system data is becoming increasingly available with higher resolution and accuracy. Access to such details enables propagation models to more accurately predict coverage and minimize interference in wireless deployments. Machine learning-based modeling can significantly support this effort, with feature-based approaches allowing for accurate, efficient, and scalable propagation modeling. Building on previous work, we introduce an extended set of features that improves prediction accuracy while, most importantly, maintaining model generalization across a broad range of environments.

Recent advancements in long-context language models (LCLMs) promise to transform Retrieval-Augmented Generation (RAG) by simplifying pipelines. With their expanded context windows, LCLMs can process entire knowledge bases and perform retrieval and reasoning directly -- a capability we define as In-Context Retrieval and Reasoning (ICR^2). However, existing benchmarks like LOFT often overestimate LCLM performance by providing overly simplified contexts. To address this, we introduce ICR^2, a benchmark that evaluates LCLMs in more realistic scenarios by including confounding passages retrieved with strong retrievers. We then propose three methods to enhance LCLM performance: (1) retrieve-then-generate fine-tuning, (2) retrieval-attention-probing, which uses attention heads to filter and de-noise long contexts during decoding, and (3) joint retrieval head training alongside the generation head. Our evaluation of five well-known LCLMs on LOFT and ICR^2 demonstrates significant gains with our best approach applied to Mistral-7B: +17 and +15 points by Exact Match on LOFT, and +13 and +2 points on ICR^2, compared to vanilla RAG and supervised fine-tuning, respectively. It even outperforms GPT-4-Turbo on most tasks despite being a much smaller model.

Models trained specifically to generate long Chains of Thought (CoTs) have recently achieved impressive results. We refer to these models as Inference-Time-Compute (ITC) models. Are the CoTs of ITC models more faithful compared to traditional non-ITC models? We evaluate two ITC models (based on Qwen-2.5 and Gemini-2) on an existing test of faithful CoT To measure faithfulness, we test if models articulate cues in their prompt that influence their answers to MMLU questions. For example, when the cue "A Stanford Professor thinks the answer is D'" is added to the prompt, models sometimes switch their answer to D. In such cases, the Gemini ITC model articulates the cue 54% of the time, compared to 14% for the non-ITC Gemini. We evaluate 7 types of cue, such as misleading few-shot examples and anchoring on past responses. ITC models articulate cues that influence them much more reliably than all the 6 non-ITC models tested, such as Claude-3.5-Sonnet and GPT-4o, which often articulate close to 0% of the time. However, our study has important limitations. We evaluate only two ITC models -- we cannot evaluate OpenAI's SOTA o1 model. We also lack details about the training of these ITC models, making it hard to attribute our findings to specific processes. We think faithfulness of CoT is an important property for AI Safety. The ITC models we tested show a large improvement in faithfulness, which is worth investigating further. To speed up this investigation, we release these early results as a research note.

Artificial Intelligence (AI) spreads quickly as new technologies and services take over modern society. The need to regulate AI design, development, and use is strictly necessary to avoid unethical and potentially dangerous consequences to humans. The European Union (EU) has released a new legal framework, the AI Act, to regulate AI by undertaking a risk-based approach to safeguard humans during interaction. At the same time, researchers offer a new perspective on AI systems, commonly known as Human-Centred AI (HCAI), highlighting the need for a human-centred approach to their design. In this context, Symbiotic AI (a subtype of HCAI) promises to enhance human capabilities through a deeper and continuous collaboration between human intelligence and AI. This article presents the results of a Systematic Literature Review (SLR) that aims to identify principles that characterise the design and development of Symbiotic AI systems while considering humans as the core of the process. Through content analysis, four principles emerged from the review that must be applied to create Human-Centred AI systems that can establish a symbiotic relationship with humans. In addition, current trends and challenges were defined to indicate open questions that may guide future research for the development of SAI systems that comply with the AI Act.