Deepfakes have recently raised significant trust issues and security concerns among the public. Compared to CNN face forgery detectors, ViT-based methods take advantage of the expressivity of transformers, achieving superior detection performance. However, these approaches still exhibit the following limitations: (1). Fully fine-tuning ViT-based models from ImageNet weights demands substantial computational and storage resources; (2). ViT-based methods struggle to capture local forgery clues, leading to model bias and limited generalizability. To tackle these challenges, this work introduces Mixture-of-Experts modules for Face Forgery Detection (MoE-FFD), a generalized yet parameter-efficient ViT-based approach. MoE-FFD only updates lightweight Low-Rank Adaptation (LoRA) and Adapter layers while keeping the ViT backbone frozen, thereby achieving parameter-efficient training. Moreover, MoE-FFD leverages the expressivity of transformers and local priors of CNNs to simultaneously extract global and local forgery clues. Additionally, novel MoE modules are designed to scale the model's capacity and select optimal forgery experts, further enhancing forgery detection performance. The proposed MoE learning scheme can be seamlessly adapted to various transformer backbones in a plug-and-play manner. Extensive experimental results demonstrate that the proposed method achieves state-of-the-art face forgery detection performance with reduced parameter overhead. The code will be released upon acceptance.
相關內容
Recent advancements in Chain-of-Thought prompting have facilitated significant breakthroughs for Large Language Models (LLMs) in complex reasoning tasks. Current research enhances the reasoning performance of LLMs by sampling multiple reasoning chains and ensembling based on the answer frequency. However, this approach fails in scenarios where the correct answers are in the minority. We identify this as a primary factor constraining the reasoning capabilities of LLMs, a limitation that cannot be resolved solely based on the predicted answers. To address this shortcoming, we introduce a hierarchical reasoning aggregation framework AoR (Aggregation of Reasoning), which selects answers based on the evaluation of reasoning chains. Additionally, AoR incorporates dynamic sampling, adjusting the number of reasoning chains in accordance with the complexity of the task. Experimental results on a series of complex reasoning tasks show that AoR outperforms prominent ensemble methods. Further analysis reveals that AoR not only adapts various LLMs but also achieves a superior performance ceiling when compared to current methods.
Large Language Models (LLMs) have demonstrated remarkable abilities in general scenarios. Instruction finetuning empowers them to align with humans in various tasks. Nevertheless, the Diversity and Quality of the instruction data remain two main challenges for instruction finetuning. With regard to this, in this paper, we propose a novel gradient-based method to automatically select high-quality and diverse instruction finetuning data for machine translation. Our key innovation centers around analyzing how individual training examples influence the model during training. Specifically, we select training examples that exert beneficial influences on the model as high-quality ones by means of Influence Function plus a small high-quality seed dataset. Moreover, to enhance the diversity of the training data we maximize the variety of influences they have on the model by clustering on their gradients and resampling. Extensive experiments on WMT22 and FLORES translation tasks demonstrate the superiority of our methods, and in-depth analysis further validates their effectiveness and generalization.
The transformative potential of AI presents remarkable opportunities, but also significant risks, underscoring the importance of responsible AI development and deployment. Despite a growing emphasis on this area, there is limited understanding of industry's engagement in responsible AI research, i.e., the critical examination of AI's ethical, social, and legal dimensions. To address this gap, we analyzed over 6 million peer-reviewed articles and 32 million patent citations using multiple methods across five distinct datasets to quantify industry's engagement. Our findings reveal that the majority of AI firms show limited or no engagement in this critical subfield of AI. We show a stark disparity between industry's dominant presence in conventional AI research and its limited engagement in responsible AI. Leading AI firms exhibit significantly lower output in responsible AI research compared to their conventional AI research and the contributions of leading academic institutions. Our linguistic analysis documents a narrower scope of responsible AI research within industry, with a lack of diversity in key topics addressed. Our large-scale patent citation analysis uncovers a pronounced disconnect between responsible AI research and the commercialization of AI technologies, suggesting that industry patents rarely build upon insights generated by the responsible AI literature. This gap highlights the potential for AI development to diverge from a socially optimal path, risking unintended consequences due to insufficient consideration of ethical and societal implications. Our results highlight the urgent need for industry to publicly engage in responsible AI research to absorb academic knowledge, cultivate public trust, and proactively mitigate AI-induced societal harms.
This paper revisits the Brewer-Nash security policy model inspired by ethical Chinese Wall policies. We draw attention to the fact that write access can be revoked in the Brewer-Nash model. The semantics of write access were underspecified originally, leading to multiple interpretations for which we provide a modern operational semantics. We go on to modernise the analysis of information flow in the Brewer-Nash model, by adopting a more precise definition adapted from Kessler. For our modernised reformulation, we provide full mechanised coverage for all theorems proposed by Brewer & Nash. Most theorems are established automatically using the tool {log} with the exception of a theorem regarding information flow, which combines a lemma in {log} with a theorem mechanised in Coq. Having covered all theorems originally posed by Brewer-Nash, achieving modern precision and mechanisation, we propose this work as a step towards a methodology for automated checking of more complex security policy models.
Ensuring that AI systems reliably and robustly avoid harmful or dangerous behaviours is a crucial challenge, especially for AI systems with a high degree of autonomy and general intelligence, or systems used in safety-critical contexts. In this paper, we will introduce and define a family of approaches to AI safety, which we will refer to as guaranteed safe (GS) AI. The core feature of these approaches is that they aim to produce AI systems which are equipped with high-assurance quantitative safety guarantees. This is achieved by the interplay of three core components: a world model (which provides a mathematical description of how the AI system affects the outside world), a safety specification (which is a mathematical description of what effects are acceptable), and a verifier (which provides an auditable proof certificate that the AI satisfies the safety specification relative to the world model). We outline a number of approaches for creating each of these three core components, describe the main technical challenges, and suggest a number of potential solutions to them. We also argue for the necessity of this approach to AI safety, and for the inadequacy of the main alternative approaches.
There is a significant overlap between people who are supported by income-related social welfare benefits, often in precarious situations, and those who experience greater digital exclusion. We report on a study of claimants using the UK's Universal Credit online welfare benefit system designed as, and still, "digital by default". Through data collection involving remote interviews (n=11) and online surveys (n=66), we expose claimants' own lived experiences interacting with this system. The claimants explain how digital channels can contribute to an imbalance of power and agency, at a time when their own circumstances mean they have reduced abilities, resources and capacities, and where design choices can adversely affect people's utility to leverage help from their own wider socio-technical ecosystems. We contribute eight recommendations from these accounts to inform the future design and development of digital welfare benefit systems for this population, to reduce digital barriers and harms.
Stickers are increasingly used in social media to express sentiment and intent. When finding typing troublesome, people often use a sticker instead. Despite the significant impact of stickers on sentiment analysis and intent recognition, little research has been conducted. To address this gap, we propose a new task: Multimodal chat Sentiment Analysis and Intent Recognition involving Stickers (MSAIRS). Additionally, we introduce a novel multimodal dataset containing Chinese chat records and stickers excerpted from several mainstream social media platforms. Our dataset includes paired data with the same text but different stickers, and various stickers consisting of the same images with different texts, allowing us to better understand the impact of stickers on chat sentiment and intent. We also propose an effective multimodal joint model, MMSAIR, for our task, which is validated on our datasets and indicates that visual information of stickers counts. Our dataset and code will be publicly available.
A growing literature on speech interruptions describes how people interrupt one another with speech, but these behaviours have not yet been implemented in the design of artificial agents which interrupt. Perceptions of a prototype proactive speech agent which adapts its speech to both urgency and to the difficulty of the ongoing task it interrupts are compared against perceptions of a static proactive agent which does not. The study hypothesises that adaptive proactive speech modelled on human speech interruptions will lead to partner models which consider the proactive agent as a stronger conversational partner than a static agent, and that interruptions initiated by an adaptive agent will be judged as better timed and more appropriately asked. These hypotheses are all rejected however, as quantitative analysis reveals that participants view the adaptive agent as a poorer dialogue partner than the static agent and as less appropriate in the style it interrupts. Qualitative analysis sheds light on the source of this surprising finding, as participants see the adaptive agent as less socially appropriate and as less consistent in its interactions than the static agent.
Although Federated Learning (FL) is promising in knowledge sharing for heterogeneous Artificial Intelligence of Thing (AIoT) devices, their training performance and energy efficacy are severely restricted in practical battery-driven scenarios due to the ``wooden barrel effect'' caused by the mismatch between homogeneous model paradigms and heterogeneous device capability. As a result, due to various kinds of differences among devices, it is hard for existing FL methods to conduct training effectively in energy-constrained scenarios, such as the battery constraints of devices. To tackle the above issues, we propose an energy-aware FL framework named DR-FL, which considers the energy constraints in both clients and heterogeneous deep learning models to enable energy-efficient FL. Unlike Vanilla FL, DR-FL adopts our proposed Muti-Agents Reinforcement Learning (MARL)-based dual-selection method, which allows participated devices to make contributions to the global model effectively and adaptively based on their computing capabilities and energy capacities in a MARL-based manner. Experiments on various well-known datasets show that DR-FL can not only maximise knowledge sharing among heterogeneous models under the energy constraint of large-scale AIoT systems but also improve the model performance of each involved heterogeneous device.
Mapping agencies are increasingly adopting Aerial Lidar Scanning (ALS) as a new tool to monitor territory and support public policies. Processing ALS data at scale requires efficient point classification methods that perform well over highly diverse territories. To evaluate them, researchers need large annotated Lidar datasets, however, current Lidar benchmark datasets have restricted scope and often cover a single urban area. To bridge this data gap, we present the FRench ALS Clouds from TArgeted Landscapes (FRACTAL) dataset: an ultra-large-scale aerial Lidar dataset made of 100,000 dense point clouds with high-quality labels for 7 semantic classes and spanning 250 km$^2$. FRACTAL is built upon France's nationwide open Lidar data. It achieves spatial and semantic diversity via a sampling scheme that explicitly concentrates rare classes and challenging landscapes from five French regions. It should support the development of 3D deep learning approaches for large-scale land monitoring. We describe the nature of the source data, the sampling workflow, the content of the resulting dataset, and provide an initial evaluation of segmentation performance using a performant 3D neural architecture.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191