Open-source large language models (LLMs) have become increasingly popular among both the general public and industry, as they can be customized, fine-tuned, and freely used. However, some open-source LLMs require approval before usage, which has led to third parties publishing their own easily accessible versions. Similarly, third parties have been publishing fine-tuned or quantized variants of these LLMs. These versions are particularly appealing to users because of their ease of access and reduced computational resource demands. This trend has increased the risk of training time attacks, compromising the integrity and security of LLMs. In this work, we present a new training time attack, SOS, which is designed to be low in computational demand and does not require clean data or modification of the model weights, thereby maintaining the model's utility intact. The attack addresses security issues in various scenarios, including the backdoor attack, jailbreak attack, and prompt stealing attack. Our experimental findings demonstrate that the proposed attack is effective across all evaluated targets. Furthermore, we present the other side of our SOS technique, namely the copyright token -- a novel technique that enables users to mark their copyrighted content and prevent models from using it.

Three years ago, Google set an ambitious plan to address climate change by going "net zero", meaning it would release no more climate-changing gases into the air than it removes, by 2030. But a report from the company on Tuesday showed it is nowhere near meeting that goal. Rather than declining, its emissions grew 13 percent in 2023 over the year before. Compared with its baseline year of 2019, emissions have soared 48 percent. Google cited artificial intelligence and the demand it puts on data centres, which require massive amounts of electricity, for last year's growth.

Google's goal of reducing its climate footprint is in jeopardy as it relies on more and more energy-hungry data centres to power its new artificial intelligence products. The tech giant revealed Tuesday that its greenhouse gas emissions have climbed 48% over the past five years. Google said electricity consumption by data centres and supply chain emissions were the primary cause of the increase. It also revealed in its annual environmental report that its emissions had risen by 13% year-on-year. The tech company, which has invested substantially in AI, said its "extremely ambitious" goal of reaching net zero emissions by 2030 "won't be easy".

Most modern reinforcement learning algorithms optimize a cumulative single-step cost along a trajectory. The optimized motions are often 'unnatural', representing, for example, behaviors with sudden accelerations that waste energy and lack predictability. In this work, we present a novel paradigm of controlling nonlinear systems via the minimization of the Koopman spectrum cost: a cost over the Koopman operator of the controlled dynamics. This induces a broader class of dynamical behaviors that evolve over stable manifolds such as nonlinear oscillators, closed loops, and smooth movements. We demonstrate that some dynamics characterizations that are not possible with a cumulative cost are feasible in this paradigm, which generalizes the classical eigenstructure and pole assignments to nonlinear decision making. Moreover, we present a sample efficient online learning algorithm for our problem that enjoys a sub-linear regret bound under some structural assumptions.

Miniature-legged robots are constrained by their onboard computation and control, thus motivating the need for simple, first-principles-based geometric models that connect \emph{periodic actuation or gaits} (a universal robot control paradigm) to the induced average locomotion. In this paper, we develop a \emph{modulable two-beat gait design framework} for sprawled planar quadrupedal systems under the no-slip using tools from geometric mechanics. We reduce standard two-beat gaits into unique subgaits in mutually exclusive shape subspaces. Subgaits are characterized by a locomotive stance phase when limbs are in ground contact and a non-locomotive, instantaneous swing phase where the limbs are reset without contact. During the stance phase, the contacting limbs form a four-bar mechanism. To analyze the ensuing locomotion, we develop the following tools: (a) a vector field to generate nonslip actuation, (b) the kinematics of a four-bar mechanism as a local connection, and (c) stratified panels that combine the kinematics and constrained actuation to encode the net change in the system's position generated by a stance-swing subgait cycle. Decoupled subgaits are then designed independently using flows on the shape-change basis and are combined with appropriate phasing to produce a two-beat gait. Further, we introduce ``scaling" and ``sliding" control inputs to continuously modulate the global trajectories of the quadrupedal system in gait time through which we demonstrate cycle-average speed, direction, and steering control using the control inputs. Thus, this framework has the potential to create uncomplicated open-loop gait plans or gain schedules for robots with limited resources, bringing them closer to achieving autonomous control.

Classification systems are evaluated in a countless number of papers. However, we find that evaluation practice is often nebulous. Frequently, metrics are selected without arguments, and blurry terminology invites misconceptions. For instance, many works use so-called 'macro' metrics to rank systems (e.g., 'macro F1') but do not clearly specify what they would expect from such a `macro' metric. This is problematic, since picking a metric can affect research findings, and thus any clarity in the process should be maximized. Starting from the intuitive concepts of bias and prevalence, we perform an analysis of common evaluation metrics. The analysis helps us understand the metrics' underlying properties, and how they align with expectations as found expressed in papers. Then we reflect on the practical situation in the field, and survey evaluation practice in recent shared tasks. We find that metric selection is often not supported with convincing arguments, an issue that can make a system ranking seem arbitrary. Our work aims at providing overview and guidance for more informed and transparent metric selection, fostering meaningful evaluation.

Building on our previous work, this paper investigates the effectiveness of interpolating control (IC) for real-time trajectory tracking. Unlike prior studies that focused on trajectory tracking itself or UAV stabilization control in simulation, we evaluate the performance of a modified extended IC (eIC) controller compared to Model Predictive Control (MPC) through both simulated and laboratory experiments with a remotely controlled UAV. The evaluation focuses on the computational efficiency and control quality of real-time UAV trajectory tracking compared to previous IC applications. The results demonstrate that the eIC controller achieves competitive performance compared to MPC while significantly reducing computational complexity, making it a promising alternative for resource-constrained platforms.

Documents are visually rich structures that convey information through text, as well as tables, figures, page layouts, or fonts. While modern document retrieval systems exhibit strong performance on query-to-text matching, they struggle to exploit visual cues efficiently, hindering their performance on practical document retrieval applications such as Retrieval Augmented Generation. To benchmark current systems on visually rich document retrieval, we introduce the Visual Document Retrieval Benchmark ViDoRe, composed of various page-level retrieving tasks spanning multiple domains, languages, and settings. The inherent shortcomings Figure 1: For each term in a user query, ColPali identifies of modern systems motivate the introduction the most relevant document image patches (highlighted of a new retrieval model architecture, zones) and computes a query-to-page matching ColPali, which leverages the document score. We can then swiftly retrieve the most relevant understanding capabilities of recent Vision Language documents from a large pre-indexed corpus. Models to produce high-quality contextualized embeddings solely from images of document pages.

Central to the digital transformation of the process industry are Digital Twins (DTs), virtual replicas of physical manufacturing systems that combine sensor data with sophisticated data-based or physics-based models, or a combination thereof, to tackle a variety of industrial-relevant tasks like process monitoring, predictive control or decision support. The backbone of a DT, i.e. the concrete modelling methodologies and architectural frameworks supporting these models, are complex, diverse and evolve fast, necessitating a thorough understanding of the latest state-of-the-art methods and trends to stay on top of a highly competitive market. From a research perspective, despite the high research interest in reviewing various aspects of DTs, structured literature reports specifically focusing on unravelling the utilized learning paradigms (e.g. self-supervised learning) for DT-creation in the process industry are a novel contribution in this field. This study aims to address these gaps by (1) systematically analyzing the modelling methodologies (e.g. Convolutional Neural Network, Encoder-Decoder, Hidden Markov Model) and paradigms (e.g. data-driven, physics-based, hybrid) used for DT-creation; (2) assessing the utilized learning strategies (e.g. supervised, unsupervised, self-supervised); (3) analyzing the type of modelling task (e.g. regression, classification, clustering); and (4) identifying the challenges and research gaps, as well as, discuss potential resolutions provided.

STRIDE aims to propel bipedal robotics research and education by providing a cost-effective implementation with step-by-step instructions for building a bipedal robotic platform while providing flexible customizations via a modular and durable design. Moreover, a versatile terrain setup and a quantitative disturbance injection system are augmented to the robot platform to replicate natural terrains and push forces that can be used to evaluate legged locomotion in practical and adversarial scenarios. We demonstrate the functionalities of this platform by realizing an adaptive step-to-step dynamics based walking controller to achieve dynamic walking. Our work with the open-soured implementation shows that STRIDE is a highly versatile and durable platform that can be used in research and education to evaluate locomotion algorithms, mechanical designs, and robust and adaptative controls.