Search

Training on high-quality synthetic data from strong language models (LMs) is a common strategy to improve the reasoning performance of LMs. In this work, we revisit whether this strategy is compute-optimal under a fixed inference budget (e.g., FLOPs). To do so, we investigate the trade-offs between generating synthetic data using a stronger but more expensive (SE) model versus a weaker but cheaper (WC) model. We evaluate the generated data across three key metrics: coverage, diversity, and false positive rate, and show that the data from WC models may have higher coverage and diversity, but also exhibit higher false positive rates. We then finetune LMs on data from SE and WC models in different settings: knowledge distillation, self-improvement, and a novel weak-to-strong improvement setup where a weaker LM teaches reasoning to a stronger LM. Our findings reveal that models finetuned on WC-generated data consistently outperform those trained on SE-generated data across multiple benchmarks and multiple choices of WC and SE models. These results challenge the prevailing practice of relying on SE models for synthetic data generation, suggesting that WC may be the compute-optimal approach for training advanced LM reasoners.

Verifiers or reward models are often used to enhance the reasoning performance of large language models (LLMs). A common approach is the Best-of-N method, where N candidate solutions generated by the LLM are ranked by a verifier, and the best one is selected. While LLM-based verifiers are typically trained as discriminative classifiers to score solutions, they do not utilize the text generation capabilities of pretrained LLMs. To overcome this limitation, we instead propose training verifiers using the ubiquitous next-token prediction objective, jointly on verification and solution generation. Compared to standard verifiers, such generative verifiers (GenRM) can benefit from several advantages of LLMs: they integrate seamlessly with instruction tuning, enable chain-of-thought reasoning, and can utilize additional inference-time compute via majority voting for better verification. We demonstrate that when using Gemma-based verifiers on algorithmic and grade-school math reasoning tasks, GenRM outperforms discriminative verifiers and LLM-as-a-Judge, showing a 16-64% improvement in the percentage of problems solved with Best-of-N. Furthermore, we show that GenRM scales favorably across dataset size, model capacity, and inference-time compute.

This paper considers an unsourced random access (URA) set-up equipped with a passive reconfigurable intelligent surface (RIS), where a massive number of unidentified users (only a small fraction of them being active at any given time) are connected to the base station (BS). We introduce a slotted coding scheme for which each active user chooses a slot at random for transmitting its signal, consisting of a pilot part and a randomly spread polar codeword. The proposed decoder operates in two phases. In the first phase, called the RIS configuration phase, the BS detects the transmitted pilots. The detected pilots are then utilized to estimate the corresponding users' channel state information, using which the BS suitably selects RIS phase shift employing the proposed RIS design algorithms. The proposed channel estimator offers the capability to obtain the channel coefficients of the users whose pilots interfere with each other without prior access to the list of transmitted pilots or the number of active users. In the second phase, called the data phase, transmitted messages of active users are decoded. Moreover, we establish an approximate achievability bound for the RIS-based URA scheme, providing a valuable benchmark. Computer simulations show that the proposed scheme outperforms the state-of-the-art for RIS-aided URA.

This work investigates the application of Machine Unlearning (MU) for mitigating the impact of trojans embedded in conventional large language models of natural language (Text-LLMs) and large language models of code (Code-LLMs) We propose a novel unlearning approach, LYA, that leverages both gradient ascent and elastic weight consolidation, a Fisher Information Matrix (FIM) based regularization technique, to unlearn trojans from poisoned models. We compare the effectiveness of LYA against conventional techniques like fine-tuning, retraining, and vanilla gradient ascent. The subject models we investigate are BERT and CodeBERT, for sentiment analysis and code defect detection tasks, respectively. Our findings demonstrate that the combination of gradient ascent and FIM-based regularization, as done in LYA, outperforms existing methods in removing the trojan's influence from the poisoned model, while preserving its original functionality. To the best of our knowledge, this is the first work that compares and contrasts MU of trojans in LLMs, in the NL and Coding domain.

The layering approach used in fused filament fabrication (FFF) enables creation of complex designs generated by topology optimization. Defects associated with the layer-by-layer process, introduce considerable random variability to the local elastic modulus of the print. The elastic modulus along the fusion layers connecting bulk materials differs from that of the bulk areas. Accurate quantitative measurements of variations in both areas are essential to achieve robust optimized designs. This study aims to quantify the parameters of the random distributions given the surface strain field of the print measured by digital image correlation (DIC). Two statistical properties, mean and standard deviation, are sufficient to characterize the stochastic elastic modulus fields in each region. An efficient neural network model is developed to estimate spatial variations in the local elastic modulus within both bulk and fusion layers. This model is trained on a dataset of synthetic strain fields with known distributions in the elastic modulus fields. It performs well in correlating the elastic modulus with the input strain, as long as the standard deviation is below 60% of the mean of the random field. The predictive accuracy of the model on testing data, measured by the R2 score, is 0.99 and 0.95 for mean and standard deviation in the fusion material. The scores for the bulk material are 0.97 each. The trained model is implemented to predict the elastic modulus distribution of an FFF-printed material at a print speed of 30 mm/s and an extrusion temperature of 493.15 K, based on its DIC-measured surface strain data. The model predicts a mean and standard deviation of 1.2 GPa and 1 GPa for the bulk material and 400 MPa and 430 MPa for the fusion region. Validation of predictions confirms the reliability of this approach in measuring uncertainty in the local properties of the prints.

As humankind prepares to establish outposts and infrastructure on the Moon, the ability to manufacture parts and buildings on-site is crucial. While transporting raw materials from Earth can be costly and time-consuming, in-situ resource utilization (ISRU) presents an attractive alternative. This review paper aims to provide a thorough examination of the current state and future potential of Lunar-based manufacturing and construction (LBMC), with a particular focus on the prospect of utilizing in-situ resources and additive manufacturing. The paper analyzes existing research on LBMC from various perspectives, including different manufacturing techniques and compositions, the potential of ISRU for LBMC, characterization of built parts and structures, the role of energy sources and efficiency, the impact of low-gravity and vacuum conditions, and the feasibility of using artificial intelligence, automation, and robotics. By synthesizing these findings, this review offers valuable insights into the challenges and opportunities that lie ahead for LBMC.

In this work, we introduce Gemma 2, a new addition to the Gemma family of lightweight, state-of-the-art open models, ranging in scale from 2 billion to 27 billion parameters. In this new version, we apply several known technical modifications to the Transformer architecture, such as interleaving local-global attentions (Beltagy et al., 2020a) and group-query attention (Ainslie et al., 2023). We also train the 2B and 9B models with knowledge distillation (Hinton et al., 2015) instead of next token prediction. The resulting models deliver the best performance for their size, and even offer competitive alternatives to models that are 2-3 times bigger. We release all our models to the community.

In this paper, we investigate the entanglement dynamics of a two qutrits system interacting with a spin environment. Using negativity as the entanglement measure, we study the entanglement dynamics of the system. The calculations show that in cases where the entanglement decays quickly, the environment will have a quantum phase transition., Comment: 10 pages, 9 figures

This paper introduces a novel zero-force control method for upper-limb exoskeleton robots, which are used in a variety of applications including rehabilitation, assistance, and human physical capability enhancement. The proposed control method employs an Adaptive Integral Terminal Sliding Mode (AITSM) controller, combined with an exponential reaching law and Proximal Policy Optimization (PPO), a type of Deep Reinforcement Learning (DRL). The PPO system incorporates an attention mechanism and Long Short-Term Memory (LSTM) neural networks, enabling the controller to selectively focus on relevant system states, adapt to changing behavior, and capture long-term dependencies. This controller is designed to manage a 5-DOF upper-limb exoskeleton robot with zero force, even amidst system uncertainties. The controller uses an integral terminal sliding surface to ensure finite-time convergence to the desired state, a crucial feature for applications requiring quick responses. It also includes an exponential switching control term to reduce chattering and improve system accuracy. The controller's adaptability, facilitated by the PPO system, allows real-time parameter adjustments based on system feedback, making the controller robust and capable of dealing with uncertainties and disturbances that could affect the performance of the exoskeleton. The proposed control method's effectiveness and superiority are confirmed through numerical simulations and comparisons with existing control methods.

Web authors frequently embed social media to support and enrich their content, creating the potential to derive web-based, cross-platform social media representations that can enable more effective social media retrieval systems and richer scientific analyses. As step toward such capabilities, we introduce a novel language modeling framework that enables automatic annotation of roles that social media entities play in their embedded web context. Using related communication theory, we liken social media embeddings to quotes, formalize the page context as structured natural language signals, and identify a taxonomy of roles for quotes within the page context. We release SocialQuotes, a new data set built from the Common Crawl of over 32 million social quotes, 8.3k of them with crowdsourced quote annotations. Using SocialQuotes and the accompanying annotations, we provide a role classification case study, showing reasonable performance with modern-day LLMs, and exposing explainable aspects of our framework via page content ablations. We also classify a large batch of un-annotated quotes, revealing interesting cross-domain, cross-platform role distributions on the web.

This investigation delves into the intricate entanglement dynamics exhibited by qubit-qutrit systems when subjected to the intricate dynamics of both the XXX isotropic Heisenberg model and the anisotropic XYZ model. These models are enriched by the presence of Dzyaloshinskii-Moriya (DM) interactions and external magnetic fields, creating a complex interplay of quantum phenomena. The exploration commences by initializing the system in a superposition of spin-coherent states, setting the stage for a nuanced analysis of entanglement evolution over time. By employing the negativity metric as the pivotal criterion for assessing entanglement, we gain insights into the subtle relationships between DM interactions, external magnetic fields, and the entanglement dynamics within this composite system. Through this comprehensive investigation, we aim to unravel the intricate dynamics and behaviors that emerge from the interplay of these fundamental elements, shedding light on the underlying principles governing quantum entanglement in this context., Comment: 10 pages, 3 figures, The original file is in LaTeX format

In this paper, the quantum entanglement dynamics of a qubit qubit compound system in the isotropic XXX Heisenberg and anisotropic XYZ models with DM interaction under magnetic fields is investigated. The system's initial state is considered as a spin coherence state, and the entanglement dynamics of this compound system is analyzed using the negativity criterion as an entanglement measure to assess the impact of DM interaction and magnetic fields., Comment: 16 pages,5 figures, The original file is in LaTeX format

Context: Machine learning (ML) and deep learning (DL) analyze raw data to extract valuable insights in specific phases. The rise of continuous practices in software projects emphasizes automating Continuous Integration (CI) with these learning-based methods, while the growing adoption of such approaches underscores the need for systematizing knowledge. Objective: Our objective is to comprehensively review and analyze existing literature concerning learning-based methods within the CI domain. We endeavour to identify and analyse various techniques documented in the literature, emphasizing the fundamental attributes of training phases within learning-based solutions in the context of CI. Method: We conducted a Systematic Literature Review (SLR) involving 52 primary studies. Through statistical and thematic analyses, we explored the correlations between CI tasks and the training phases of learning-based methodologies across the selected studies, encompassing a spectrum from data engineering techniques to evaluation metrics. Results: This paper presents an analysis of the automation of CI tasks utilizing learning-based methods. We identify and analyze nine types of data sources, four steps in data preparation, four feature types, nine subsets of data features, five approaches for hyperparameter selection and tuning, and fifteen evaluation metrics. Furthermore, we discuss the latest techniques employed, existing gaps in CI task automation, and the characteristics of the utilized learning-based techniques. Conclusion: This study provides a comprehensive overview of learning-based methods in CI, offering valuable insights for researchers and practitioners developing CI task automation. It also highlights the need for further research to advance these methods in CI., Comment: This paper has been accepted to be published in IEEE Access

Large language models can memorize and repeat their training data, causing privacy and copyright risks. To mitigate memorization, we introduce a subtle modification to the next-token training objective that we call the goldfish loss. During training, a randomly sampled subset of tokens are excluded from the loss computation. These dropped tokens are not memorized by the model, which prevents verbatim reproduction of a complete chain of tokens from the training set. We run extensive experiments training billion-scale Llama-2 models, both pre-trained and trained from scratch, and demonstrate significant reductions in extractable memorization with little to no impact on downstream benchmarks., Comment: 9.5 pages, 8 figures, and 1 table in the main body. Code available at https://github.com/ahans30/goldfish-loss

With the continuous advancement of large language models (LLMs), it is essential to create new benchmarks to effectively evaluate their expanding capabilities and identify areas for improvement. This work focuses on multi-image reasoning, an emerging capability in state-of-the-art LLMs. We introduce ReMI, a dataset designed to assess LLMs' ability to Reason with Multiple Images. This dataset encompasses a diverse range of tasks, spanning various reasoning domains such as math, physics, logic, code, table/chart understanding, and spatial and temporal reasoning. It also covers a broad spectrum of characteristics found in multi-image reasoning scenarios. We have benchmarked several cutting-edge LLMs using ReMI and found a substantial gap between their performance and human-level proficiency. This highlights the challenges in multi-image reasoning and the need for further research. Our analysis also reveals the strengths and weaknesses of different models, shedding light on the types of reasoning that are currently attainable and areas where future models require improvement. To foster further research in this area, we are releasing ReMI publicly: https://huggingface.co/datasets/mehrankazemi/ReMI.

Large language models (LLMs) have showcased remarkable reasoning capabilities, yet they remain susceptible to errors, particularly in temporal reasoning tasks involving complex temporal logic. Existing research has explored LLM performance on temporal reasoning using diverse datasets and benchmarks. However, these studies often rely on real-world data that LLMs may have encountered during pre-training or employ anonymization techniques that can inadvertently introduce factual inconsistencies. In this work, we address these limitations by introducing novel synthetic datasets specifically designed to assess LLM temporal reasoning abilities in various scenarios. The diversity of question types across these datasets enables systematic investigation into the impact of the problem structure, size, question type, fact order, and other factors on LLM performance. Our findings provide valuable insights into the strengths and weaknesses of current LLMs in temporal reasoning tasks. To foster further research in this area, we are open-sourcing the datasets and evaluation framework used in our experiments: https://huggingface.co/datasets/baharef/ToT.

We investigate the unsourced random access scheme assuming that the base station is equipped with multiple antennas, and propose a high-performing solution utilizing on-off-division multiple access. We assume that each user spreads its pilot sequence and polar codeword to the pilot and data parts of the transmission frame, respectively, based on a transmission pattern. The iterative receiver operation consists of pilot and pattern detection followed by channel vector and symbol estimation, polar decoding, and successive interference cancellation. Numerical findings demonstrate that the proposed scheme has superior performance compared to the state-of-the-art in various antenna settings.

Which transformer scaling regimes are able to perfectly solve different classes of algorithmic problems? While tremendous empirical advances have been attained by transformer-based neural networks, a theoretical understanding of their algorithmic reasoning capabilities in realistic parameter regimes is lacking. We investigate this question in terms of the network's depth, width, and number of extra tokens for algorithm execution. Our novel representational hierarchy separates 9 algorithmic reasoning problems into classes solvable by transformers in different realistic parameter scaling regimes. We prove that logarithmic depth is necessary and sufficient for tasks like graph connectivity, while single-layer transformers with small embedding dimensions can solve contextual retrieval tasks. We also support our theoretical analysis with ample empirical evidence using the GraphQA benchmark. These results show that transformers excel at many graph reasoning tasks, even outperforming specialized graph neural networks., Comment: 43 pages, 8 figures

In this study, we explore the dynamics of quantum entanglement using the negativity criterion for the W_zeta quantum state. We investigate changes in negativity in terms of anisotropy parameters, gamma, the strength of the external magnetic field applied to the spin chain, eta, the triple interaction strength, alpha. We examine how these parameters affect the entanglement properties of the system and discuss the implications for quantum information processing and quantum communication protocols. By analyzing the negativity of the W_zeta state under different conditions, we gain insights into the behaviour of entanglement in complex quantum systems. Our results shed light on the intricate interplay between various factors that influence quantum entanglement and provide a foundation for further investigations in this field of research., Comment: 14 pages, 9 figures

Although Large Language Models (LLMs) achieve remarkable performance across various tasks, they often struggle with complex reasoning tasks, such as answering mathematical questions. Recent efforts to address this issue have primarily focused on leveraging mathematical datasets through supervised fine-tuning or self-improvement techniques. However, these methods often depend on high-quality datasets that are difficult to prepare, or they require substantial computational resources for fine-tuning. Inspired by findings that LLMs know how to produce the right answer but struggle to select the correct reasoning path, we propose a purely inference-based searching method -- MindStar (M*). This method formulates reasoning tasks as searching problems and proposes two search ideas to identify the optimal reasoning paths. We evaluate the M* framework on both the GSM8K and MATH datasets, comparing its performance with existing open and closed-source LLMs. Our results demonstrate that M* significantly enhances the reasoning abilities of open-source models, such as Llama-2-13B and Mistral-7B, and achieves comparable performance to GPT-3.5 and Grok-1, but with substantially reduced model size and computational costs.

In this article, we investigate the entanglement evolution of three-qubit states in the presence of a spin chain environment. Utilizing negativity as a metric for entanglement assessment, we focus on the GHZ, W, and W_zeta quantum states as the initial system states. We explore and analyze the entanglement dynamics of these states based on various parameters., Comment: 18 pages, 10 figures

Diaspora communities are disproportionately impacted by off-the-radar misinformation and often neglected by mainstream fact-checking efforts, creating a critical need to scale-up efforts of nascent fact-checking initiatives. In this paper we present SynDy, a framework for Synthetic Dynamic Dataset Generation to leverage the capabilities of the largest frontier Large Language Models (LLMs) to train local, specialized language models. To the best of our knowledge, SynDy is the first paper utilizing LLMs to create fine-grained synthetic labels for tasks of direct relevance to misinformation mitigation, namely Claim Matching, Topical Clustering, and Claim Relationship Classification. SynDy utilizes LLMs and social media queries to automatically generate distantly-supervised, topically-focused datasets with synthetic labels on these three tasks, providing essential tools to scale up human-led fact-checking at a fraction of the cost of human-annotated data. Training on SynDy's generated labels shows improvement over a standard baseline and is not significantly worse compared to training on human labels (which may be infeasible to acquire). SynDy is being integrated into Meedan's chatbot tiplines that are used by over 50 organizations, serve over 230K users annually, and automatically distribute human-written fact-checks via messaging apps such as WhatsApp. SynDy will also be integrated into our deployed Co-Insights toolkit, enabling low-resource organizations to launch tiplines for their communities. Finally, we envision SynDy enabling additional fact-checking tools such as matching new misinformation claims to high-quality explainers on common misinformation topics.

The silicon T centre's narrow, telecommunications-band optical emission, long spin coherence, and direct photonic integration have spurred interest in this emitter as a spin-photon interface for distributed quantum computing and networking. However, key parameters of the T centre's spin-selective optical transitions remain undetermined or ambiguous in literature. In this paper we present a Hamiltonian of the T centre TX state and determine key parameters of the optical transition from T$_0$ to TX$_0$ from a combined analysis of published results, density functional theory, and new spectroscopy. We resolve ambiguous values of the internal defect potential in the literature, and we present the first measurements of electrically tuned T centre emission. As a result, we provide a model of the T centre's optical and spin properties under strain, electric, and magnetic fields that can be utilized for realizing quantum technologies., Comment: 9 pages and 6 figures in the main manuscript. 10 pages and 6 figures in the supplementary information

Many developers rely on Large Language Models (LLMs) to facilitate software development. Nevertheless, these models have exhibited limited capabilities in the security domain. We introduce LLMSecGuard, a framework to offer enhanced code security through the synergy between static code analyzers and LLMs. LLMSecGuard is open source and aims to equip developers with code solutions that are more secure than the code initially generated by LLMs. This framework also has a benchmarking feature, aimed at providing insights into the evolving security attributes of these models., Comment: SECUTE, EASE 2024

Social media platforms actively moderate content glorifying harmful behaviors like eating disorders, which include anorexia and bulimia. However, users have adapted to evade moderation by using coded hashtags. Our study investigates the prevalence of moderation evaders on the popular social media platform TikTok and contrasts their use and emotional valence with mainstream hashtags. We notice that moderation evaders and mainstream hashtags appear together, indicating that vulnerable users might inadvertently encounter harmful content even when searching for mainstream terms. Additionally, through an analysis of emotional expressions in video descriptions and comments, we find that mainstream hashtags generally promote positive engagement, while moderation evaders evoke a wider range of emotions, including heightened negativity. These findings provide valuable insights for content creators, platform moderation efforts, and interventions aimed at cultivating a supportive online environment for discussions on mental health and eating disorders., Comment: 10 pages, 5 figures, 2 tables

Large Vision Language Models (VLMs), such as CLIP, have significantly contributed to various computer vision tasks, including object recognition and object detection. Their open vocabulary feature enhances their value. However, their black-box nature and lack of explainability in predictions make them less trustworthy in critical domains. Recently, some work has been done to force VLMs to provide reasonable rationales for object recognition, but this often comes at the expense of classification accuracy. In this paper, we first propose a mathematical definition of explainability in the object recognition task based on the joint probability distribution of categories and rationales, then leverage this definition to fine-tune CLIP in an explainable manner. Through evaluations of different datasets, our method demonstrates state-of-the-art performance in explainable classification. Notably, it excels in zero-shot settings, showcasing its adaptability. This advancement improves explainable object recognition, enhancing trust across diverse applications. The code will be made available online upon publication.

In this paper, we put forward a proof of concept for sixth generation (6G) Terabit infrared (IR) laser-based indoor optical wireless networks. We propose a novel double-tier access point (AP) architecture based on an array of arrays of vertical cavity surface emitting lasers (VCSELs) to provide a seamless grid-of-beam coverage with multi-Gb/s per beam. We present systematic design and thorough analytical modeling of the AP architecture, which are then applied to downlink system modeling using non-imaging angle diversity receivers (ADRs). We propose static beam clustering with coordinated multi-beam joint transmission (CoMB-JT) for network interference management and devise various clustering strategies to address inter-beam interference (IBI) and inter-cluster interference (ICI). Non-orthogonal multiple access (NOMA) and orthogonal frequency division multiple access (OFDMA) schemes are also adopted to handle intra-cluster interference, and the resulting signal-to-interference-plus-noise ratio (SINR) and achievable data rate are derived. The network performance is studied in terms of spatial distributions and statistics of the downlink SINR and data rate through extensive computer simulations. The results demonstrate that data rates up to 15 Gb/s are achieved within the coverage area and a properly devised clustering strikes a balance between the sum rate and fairness depending on the number of users., Comment: 15 pages, 13 figures, 1 table

We report the observation of a coalescing compact binary with component masses $2.5\text{-}4.5~M_\odot$ and $1.2\text{-}2.0~M_\odot$ (all measurements quoted at the 90% credible level). The gravitational-wave signal GW230529_181500 was observed during the fourth observing run of the LIGO-Virgo-KAGRA detector network on 2023 May 29 by the LIGO Livingston Observatory. The primary component of the source has a mass less than $5~M_\odot$ at 99% credibility. We cannot definitively determine from gravitational-wave data alone whether either component of the source is a neutron star or a black hole. However, given existing estimates of the maximum neutron star mass, we find the most probable interpretation of the source to be the coalescence of a neutron star with a black hole that has a mass between the most massive neutron stars and the least massive black holes observed in the Galaxy. We provisionally estimate a merger rate density of $55^{+127}_{-47}~\text{Gpc}^{-3}\,\text{yr}^{-1}$ for compact binary coalescences with properties similar to the source of GW230529_181500; assuming that the source is a neutron star-black hole merger, GW230529_181500-like sources constitute about 60% of the total merger rate inferred for neutron star-black hole coalescences. The discovery of this system implies an increase in the expected rate of neutron star-black hole mergers with electromagnetic counterparts and provides further evidence for compact objects existing within the purported lower mass gap., Comment: 45 pages (10 pages author list, 13 pages main text, 1 page acknowledgements, 13 pages appendices, 8 pages bibliography), 17 figures, 16 tables. Update to match version published in The Astrophysical Journal Letters. Data products available from https://zenodo.org/records/10845779

Fabrication of low-RF loss GaN-on-Si HEMT stacks is critical to enable competitive front-end-modules for 5G and 6G applications. The main contribution to RF losses is the interface between the III-N layer and the HR Si wafer, more specifically the AlN/Si interface. At this interface, a parasitic surface conduction layer exists in Si, which decreases the substrate effective resistivity sensed by overlying circuitry below the nominal Si resistivity. However, a clear understanding of this interface with control of the parasitic channel is lacking. In this letter, a detailed physical and electrical description of MOCVD-grown AlN/Si structures is presented. The presence of a $\text{SiC}_\text{x}\text{N}_\text{y}$ interfacial layer is revealed and its importance for RF losses is shown. Through C-V and I-V characterisation, an increase in the C concentration of this interfacial layer is linked to the formation of negative charge at the AlN/Si interface, which counteracts the positive charge present in the 0-predose limit. The variation of TMAl predose is shown to allow precise tuning of the C composition and, consequently, the resulting interface charge. Notably, a linear relationship between predose and net interface charge is observed and confirmed by the fabrication of an AlN/Si sample with close to zero net charge. In addition, a higher $D_{it}$ ($\sim 2\times 10^{12}$ cm$^\text{-2}$) for such compensated samples is observed and can contribute to low RF loss. An exceptionally high effective resistivity of above 8 k$\Omega\cdot$cm is achieved, corresponding to an RF loss below 0.3 dB/mm at 10 GHz., Comment: The following article has been accepted for publication in Applied Physics Letters. After it is published, it will be found at https://pubs.aip.org/aip/apl

Dialog Structure Induction (DSI) is the task of inferring the latent dialog structure (i.e., a set of dialog states and their temporal transitions) of a given goal-oriented dialog. It is a critical component for modern dialog system design and discourse analysis. Existing DSI approaches are often purely data-driven, deploy models that infer latent states without access to domain knowledge, underperform when the training corpus is limited/noisy, or have difficulty when test dialogs exhibit distributional shifts from the training domain. This work explores a neural-symbolic approach as a potential solution to these problems. We introduce Neural Probabilistic Soft Logic Dialogue Structure Induction (NEUPSL DSI), a principled approach that injects symbolic knowledge into the latent space of a generative neural model. We conduct a thorough empirical investigation on the effect of NEUPSL DSI learning on hidden representation quality, few-shot learning, and out-of-domain generalization performance. Over three dialog structure induction datasets and across unsupervised and semi-supervised settings for standard and cross-domain generalization, the injection of symbolic knowledge using NEUPSL DSI provides a consistent boost in performance over the canonical baselines.

Off-Policy Prediction (OPP), i.e., predicting the outcomes of a target policy using only data collected under a nominal (behavioural) policy, is a paramount problem in data-driven analysis of safety-critical systems where the deployment of a new policy may be unsafe. To achieve dependable off-policy predictions, recent work on Conformal Off-Policy Prediction (COPP) leverage the conformal prediction framework to derive prediction regions with probabilistic guarantees under the target process. Existing COPP methods can account for the distribution shifts induced by policy switching, but are limited to single-agent systems and scalar outcomes (e.g., rewards). In this work, we introduce MA-COPP, the first conformal prediction method to solve OPP problems involving multi-agent systems, deriving joint prediction regions for all agents' trajectories when one or more "ego" agents change their policies. Unlike the single-agent scenario, this setting introduces higher complexity as the distribution shifts affect predictions for all agents, not just the ego agents, and the prediction task involves full multi-dimensional trajectories, not just reward values. A key contribution of MA-COPP is to avoid enumeration or exhaustive search of the output space of agent trajectories, which is instead required by existing COPP methods to construct the prediction region. We achieve this by showing that an over-approximation of the true JPR can be constructed, without enumeration, from the maximum density ratio of the JPR trajectories. We evaluate the effectiveness of MA-COPP in multi-agent systems from the PettingZoo library and the F1TENTH autonomous racing environment, achieving nominal coverage in higher dimensions and various shift settings., Comment: Submitted to the 63rd IEEE Conference on Decision and Control (CDC)

Modern neural networks are often trained on massive datasets that are web scraped with minimal human inspection. As a result of this insecure curation pipeline, an adversary can poison or backdoor the resulting model by uploading malicious data to the internet and waiting for a victim to scrape and train on it. Existing approaches for creating poisons and backdoors start with randomly sampled clean data, called base samples, and then modify those samples to craft poisons. However, some base samples may be significantly more amenable to poisoning than others. As a result, we may be able to craft more potent poisons by carefully choosing the base samples. In this work, we use guided diffusion to synthesize base samples from scratch that lead to significantly more potent poisons and backdoors than previous state-of-the-art attacks. Our Guided Diffusion Poisoning (GDP) base samples can be combined with any downstream poisoning or backdoor attack to boost its effectiveness. Our implementation code is publicly available at: https://github.com/hsouri/GDP .

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

We employ an inversion-based approach to examine CLIP models. Our examination reveals that inverting CLIP models results in the generation of images that exhibit semantic alignment with the specified target prompts. We leverage these inverted images to gain insights into various aspects of CLIP models, such as their ability to blend concepts and inclusion of gender biases. We notably observe instances of NSFW (Not Safe For Work) images during model inversion. This phenomenon occurs even for semantically innocuous prompts, like "a beautiful landscape," as well as for prompts involving the names of celebrities., Comment: Warning: This paper contains sexually explicit images and language, offensive visuals and terminology, discussions on pornography, gender bias, and other potentially unsettling, distressing, and/or offensive content for certain readers

The potential for manipulating characteristics of skyrmions in a CrI$_3$ monolayer using circularly polarised light is explored. The effective skyrmion-light interaction is mediated by bright excitons whose magnetization is selectively influenced by the polarization of photons. The light-induced skyrmion dynamics is illustrated by the dependencies of the skyrmion size and the skyrmion lifetime on the intensity and polarization of the incident light pulse. Two-dimensional magnets hosting excitons thus represent a promising platform for the control of topological magnetic structures by light.

The concept of Quality of Life (QoL) refers to a holistic measurement of an individual's well-being, incorporating psychological and social aspects. Pregnant women, especially those with obesity and stress, often experience lower QoL. Physical activity (PA) has shown the potential to enhance the QoL. However, pregnant women who are overweight and obese rarely meet the recommended level of PA. Studies have investigated the relationship between PA and QoL during pregnancy using correlation-based approaches. These methods aim to discover spurious correlations between variables rather than causal relationships. Besides, the existing methods mainly rely on physical activity parameters and neglect the use of different factors such as maternal (medical) history and context data, leading to biased estimates. Furthermore, the estimations lack an understanding of mediators and counterfactual scenarios that might affect them. In this paper, we investigate the causal relationship between being physically active (treatment variable) and the QoL (outcome) during pregnancy and postpartum. To estimate the causal effect, we develop a Causal Machine Learning method, integrating causal discovery and causal inference components. The data for our investigation is derived from a long-term wearable-based health monitoring study focusing on overweight and obese pregnant women. The machine learning (meta-learner) estimation technique is used to estimate the causal effect. Our result shows that performing adequate physical activity during pregnancy and postpartum improves the QoL by units of 7.3 and 3.4 on average in physical health and psychological domains, respectively. In the final step, four refutation analysis techniques are employed to validate our estimation.

Relaxing the postulates of an axiomatic theory is a natural way to find more general theories, and historically, the discovery of non-Euclidean geometry is a famous example of this procedure. Here, we use this way to extend quantum mechanics by ignoring the heart of Heisenberg's quantum mechanics -- We do not assume the existence of a position operator that satisfies the Heisenberg commutation relation, $[\hat x,\hat p]=i\hbar$. The remaining axioms of quantum theory, besides Galilean symmetry, lead to a more general quantum theory with a free parameter $l_0$ of length dimension, such that as $l_0 \to 0$ the theory reduces to standard quantum theory. Perhaps surprisingly, this non-Heisenberg quantum theory, without a priori assumption of the non-commutation relation, leads to a modified Heisenberg uncertainty relation, $\Delta x \Delta p\geq \sqrt{\hbar^2/4+l_0^2(\Delta p)^2}$, which ensures the existence of a minimal position uncertainty, $l_0$, as expected from various quantum gravity studies. By comparing the results of this framework with some observed data, which includes the first longitudinal normal modes of the bar gravitational wave detector AURIGA and the $1S-2S$ transition in the hydrogen atom, we obtain upper bounds on the $l_0$.

Our work addresses a fundamental problem in the context of counterfactual inference for Markov Decision Processes (MDPs). Given an MDP path $\tau$, this kind of inference allows us to derive counterfactual paths $\tau'$ describing what-if versions of $\tau$ obtained under different action sequences than those observed in $\tau$. However, as the counterfactual states and actions deviate from the observed ones over time, the observation $\tau$ may no longer influence the counterfactual world, meaning that the analysis is no longer tailored to the individual observation, resulting in interventional outcomes rather than counterfactual ones. Even though this issue specifically affects the popular Gumbel-max structural causal model used for MDP counterfactuals, it has remained overlooked until now. In this work, we introduce a formal characterisation of influence based on comparing counterfactual and interventional distributions. We devise an algorithm to construct counterfactual models that automatically satisfy influence constraints. Leveraging such models, we derive counterfactual policies that are not just optimal for a given reward structure but also remain tailored to the observed path. Even though there is an unavoidable trade-off between policy optimality and strength of influence constraints, our experiments demonstrate that it is possible to derive (near-)optimal policies while remaining under the influence of the observation., Comment: 12 pages, 6 figures

How can we best encode structured data into sequential form for use in large language models (LLMs)? In this work, we introduce a parameter-efficient method to explicitly represent structured data for LLMs. Our method, GraphToken, learns an encoding function to extend prompts with explicit structured information. Unlike other work which focuses on limited domains (e.g. knowledge graph representation), our work is the first effort focused on the general encoding of structured data to be used for various reasoning tasks. We show that explicitly representing the graph structure allows significant improvements to graph reasoning tasks. Specifically, we see across the board improvements - up to 73% points - on node, edge and, graph-level tasks from the GraphQA benchmark.

Video quality can suffer from limited internet speed while being streamed by users. Compression artifacts start to appear when the bitrate decreases to match the available bandwidth. Existing algorithms either focus on removing the compression artifacts at the same video resolution, or on upscaling the video resolution but not removing the artifacts. Super resolution-only approaches will amplify the artifacts along with the details by default. We propose a lightweight convolutional neural network (CNN)-based algorithm which simultaneously performs artifacts reduction and super resolution (ARSR) by enhancing the feature extraction layers and designing a custom training dataset. The output of this neural network is evaluated for test streams compressed at low bitrates using variable bitrate (VBR) encoding. The output video quality shows a 4-6 increase in video multi-method assessment fusion (VMAF) score compared to traditional interpolation upscaling approaches such as Lanczos or Bicubic.

Kibble-Zurek mechanism relates the domain of non-equilibrium dynamics with the critical properties at equilibrium. It establishes a power law connection between non-equilibrium defects quenched through a continuous phase transition and the quench rate via the scaling exponent. We present a novel numerical scheme to estimate the scaling exponent wherein the notion of defects is mapped to errors, previously introduced to quantify a variety of gapped quantum phases. To demonstrate the versatility of our method we conduct numerical experiments across a broad spectrum of spin-half models hosting local and symmetry protected topological order. Furthermore, an implementation of the quench dynamics featuring a topological phase transition on a digital quantum computer is proposed to quantify the associated criticality., Comment: 20 pages, 18 figures. Expanded to include new analysis. Comments are welcome

Detecting text generated by modern large language models is thought to be hard, as both LLMs and humans can exhibit a wide range of complex behaviors. However, we find that a score based on contrasting two closely related language models is highly accurate at separating human-generated and machine-generated text. Based on this mechanism, we propose a novel LLM detector that only requires simple calculations using a pair of pre-trained LLMs. The method, called Binoculars, achieves state-of-the-art accuracy without any training data. It is capable of spotting machine text from a range of modern LLMs without any model-specific modifications. We comprehensively evaluate Binoculars on a number of text sources and in varied situations. Over a wide range of document types, Binoculars detects over 90% of generated samples from ChatGPT (and other LLMs) at a false positive rate of 0.01%, despite not being trained on any ChatGPT data., Comment: 20 pages, code available at https://github.com/ahans30/Binoculars

Language models, especially pre-trained large language models, have showcased remarkable abilities as few-shot in-context learners (ICL), adept at adapting to new tasks with just a few demonstrations in the input context. However, the model's ability to perform ICL is sensitive to the choice of the few-shot demonstrations. Instead of using a fixed set of demonstrations, one recent development is to retrieve demonstrations tailored to each input query. The implementation of demonstration retrieval is relatively straightforward, leveraging existing databases and retrieval systems. This not only improves the efficiency and scalability of the learning process but also has been shown to reduce biases inherent in manual example selection. In light of the encouraging results and growing research in ICL with retrieved demonstrations, we conduct an extensive review of studies in this area. In this survey, we discuss and compare different design choices for retrieval models, retrieval training procedures, and inference algorithms.

As automated driving technology advances, the role of the driver to resume control of the vehicle in conditionally automated vehicles becomes increasingly critical. In the SAE Level 3 or partly automated vehicles, the driver needs to be available and ready to intervene when necessary. This makes it essential to evaluate their readiness accurately. This article presents a comprehensive analysis of driver readiness assessment by combining head pose features and eye-tracking data. The study explores the effectiveness of predictive models in evaluating driver readiness, addressing the challenges of dataset limitations and limited ground truth labels. Machine learning techniques, including LSTM architectures, are utilised to model driver readiness based on the Spatio-temporal status of the driver's head pose and eye gaze. The experiments in this article revealed that a Bidirectional LSTM architecture, combining both feature sets, achieves a mean absolute error of 0.363 on the DMD dataset, demonstrating superior performance in assessing driver readiness. The modular architecture of the proposed model also allows the integration of additional driver-specific features, such as steering wheel activity, enhancing its adaptability and real-world applicability.

Respiratory rate (RR) serves as an indicator of various medical conditions, such as cardiovascular diseases and sleep disorders. These RR estimation methods were mostly designed for finger-based PPG collected from subjects in stationary situations (e.g., in hospitals). In contrast to finger-based PPG signals, wrist-based PPG are more susceptible to noise, particularly in their low frequency range, which includes respiratory information. Therefore, the existing methods struggle to accurately extract RR when PPG data are collected from wrist area under free-living conditions. The increasing popularity of smartwatches, equipped with various sensors including PPG, has prompted the need for a robust RR estimation method. In this paper, we propose a convolutional neural network-based approach to extract RR from PPG, accelerometer, and gyroscope signals captured via smartwatches. Our method, including a dilated residual inception module and 1D convolutions, extract the temporal information from the signals, enabling RR estimation. Our method is trained and tested using data collected from 36 subjects under free-living conditions for one day using Samsung Gear Sport watches. For evaluation, we compare the proposed method with four state-of-the-art RR estimation methods. The RR estimates are compared with RR references obtained from a chest-band device. The results show that our method outperforms the existing methods with the Mean-Absolute-Error and Root-Mean-Square-Error of 1.85 and 2.34, while the best results obtained by the other methods are 2.41 and 3.29, respectively. Moreover, compared to the other methods, the absolute error distribution of our method was narrow (with the lowest median), indicating a higher level of agreement between the estimated and reference RR values.

Scientists have demonstrated that quantum computing has presented novel approaches to address computational challenges, each varying in complexity. Adapting problem-solving strategies is crucial to harness the full potential of quantum computing. Nonetheless, there are defined boundaries to the capabilities of quantum computing. This paper concentrates on aggregating prior research efforts dedicated to solving intricate classical computational problems through quantum computing. The objective is to systematically compile an exhaustive inventory of these solutions and categorize a collection of demanding open problems that await further exploration.hese solutions and categorize a collection of demanding open problems that await further exploration., Comment: arXiv admin note: text overlap with arXiv:1612.07366 by other authors