Search

Background: Software modelling is a creative yet challenging task. Modellers often find themselves lost in the process, from understanding the modelling problem to solving it with proper modelling strategies and modelling tools. Students learning modelling often get overwhelmed with the notations and tools. To teach students systematic modelling, we must investigate students' practical modelling knowledge and the challenges they face while modelling. Aim: We aim to explore students' modelling knowledge and modelling actions. Further, we want to investigate students' challenges while solving a modelling task on specific modelling tools. Method: We conducted an empirical study by observing 16 pairs of students from two universities and countries solving modelling tasks for one hour. Results: We find distinct patterns of modelling of class and sequence diagrams based on individual modelling styles, the tools' interface and modelling knowledge. We observed how modelling tools influence students' modelling styles and how they can be used to foster students' confidence and creativity. Based on these observations, we developed a set of guidelines aimed at enhancing modelling education and helping students acquire practical modelling skills. Conclusions: The guidance for modelling in education needs to be structured and systematic. Our findings reveal that different modelling styles exist, which should be properly studied. It is essential to nurture the creative aspect of a modeller, particularly while they are still students. Therefore, selecting the right tool is important, and students should understand how a tool can influence their modelling style., Comment: Total 31 pages, including 4 pages of references and 2 pages of appendix (supporting materials)

Dialog systems, such as voice assistants, are expected to engage with users in complex, evolving conversations. Unfortunately, traditional automatic speech recognition (ASR) systems deployed in such applications are usually trained to recognize each turn independently and lack the ability to adapt to the conversational context or incorporate user feedback. In this work, we introduce a general framework for ASR in dialog systems that can go beyond learning from single-turn utterances and learn over time how to adapt to both explicit supervision and implicit user feedback present in multi-turn conversations. We accomplish that by leveraging advances in student-teacher learning and context-aware dialog processing, and designing contrastive self-supervision approaches with Ohm, a new online hard-negative mining approach. We show that leveraging our new framework compared to traditional training leads to relative WER reductions of close to 10% in real-world dialog systems, and up to 26% on public synthetic data., Comment: Presented at ICML 2024

Nucleic acids such as mRNA have emerged as a promising therapeutic modality with the capability of addressing a wide range of diseases. Lipid nanoparticles (LNPs) as a delivery platform for nucleic acids were used in the COVID-19 vaccines and have received much attention. While modern manufacturing processes which involve rapidly mixing an organic stream containing the lipids with an aqueous stream containing the nucleic acids are conceptually straightforward, detailed understanding of LNP formation and structure is still limited and scale-up can be challenging. Mathematical and computational methods are a promising avenue for deepening scientific understanding of the LNP formation process and facilitating improved process development and control. This article describes strategies for the mechanistic modeling of LNP formation, starting with strategies to estimate and predict important physicochemical properties of the various species such as diffusivities and solubilities. Subsequently, a framework is outlined for constructing mechanistic models of reactor- and particle-scale processes. Insights gained from the various models are mapped back to product quality attributes and process insights. Lastly, the use of the models to guide development of advanced process control and optimization strategies is discussed., Comment: 67 pages, 10 figures

The Virtual Experimental Research Assistant (VERA) is an inquiry-based learning environment that empowers a learner to build conceptual models of complex ecological systems and experiment with agent-based simulations of the models. This study investigates the convergence of cognitive AI and generative AI for self-explanation in interactive AI agents such as VERA. From a cognitive AI viewpoint, we endow VERA with a functional model of its own design, knowledge, and reasoning represented in the Task--Method--Knowledge (TMK) language. From the perspective of generative AI, we use ChatGPT, LangChain, and Chain-of-Thought to answer user questions based on the VERA TMK model. Thus, we combine cognitive and generative AI to generate explanations about how VERA works and produces its answers. The preliminary evaluation of the generation of explanations in VERA on a bank of 66 questions derived from earlier work appears promising., Comment: 10 pages, 2 figures, 2 tables, 1 appendix, HEXED Workshop @EDM July 2024

In this paper, we address the problem of joint allocation of transmit and jamming power at the source and destination, respectively, to enhance the long-term cumulative secrecy performance of an energy-harvesting wireless communication system until it stops functioning in the presence of an eavesdropper. The source and destination have energy-harvesting devices with limited battery capacities. The destination also has a full-duplex transceiver to transmit jamming signals for secrecy. We frame the problem as an infinite-horizon Markov decision process (MDP) problem and propose a reinforcement learning-based optimal joint power allocation (OJPA) algorithm that employs a policy iteration (PI) algorithm. Since the optimal algorithm is computationally expensive, we develop a low-complexity sub-optimal joint power allocation (SJPA) algorithm, namely, reduced state joint power allocation (RSJPA). Two other SJPA algorithms, the greedy algorithm (GA) and the naive algorithm (NA), are implemented as benchmarks. In addition, the OJPA algorithm outperforms the individual power allocation (IPA) algorithms termed individual transmit power allocation (ITPA) and individual jamming power allocation (IJPA), where the transmit and jamming powers, respectively, are optimized individually. The results show that the OJPA algorithm is also more energy efficient. Simulation results show that the OJPA algorithm significantly improves the secrecy performance compared to all SJPA algorithms. The proposed RSJPA algorithm achieves nearly optimal performance with significantly less computational complexity marking it the balanced choice between the complexity and the performance. We find that the computational time for the RSJPA algorithm is around 75 percent less than the OJPA algorithm.

Active metamaterials promise to enable arbitrary, temporal control over the propagation of wavefronts of light for applications such as beam steering, optical communication modulators, and holograms. This has been done in the past using patterned silicon photonics to locally control the phase of light such that the metasurface acts as a large number of wavelets. Although phase modulation only requires refractive index modulation when the interaction length is on the order of the wavelength, this is not enough to significantly modulate the phase of light in flatland. Instead, phase modulation is achieved using a resonant mode such as a plasmon or high-Q cavity mode that enable light to accumulate a large amount of phase over a short distance and coupling it to an active material that modulates the light-matter interactions. Here, we report that electrostatic doping can modulate the light-matter interaction strength of a two-dimensional WS2 based multi quantum well (MQW) structure going from strongly-coupled, phase-accumulating exciton-polaritons to weakly-coupled exciton-trion-polaritons. As a result of this transition, 2.02{\pi} radians of phase modulation is observed using spectroscopic ellipsometry. This result demonstrates the potential of the MQW structure as a compact, lightweight electro-optical modulators for LiDAR and optical communications in the red region of visible spectrum.

Broadband absorption in the visible spectrum is essential in optoelectronic applications that involve power conversion such as photovoltaics and photocatalysis. Most ultrathin broadband absorbers use parasitic plasmonic structures that maximize absorption using surface plasmons and/or Fabry-Perot cavities, which limits the weight efficiency of the device. Here, we show the theoretical and experimental realization of an unpatterned/planar semiconductor thin-film absorber based on monolayer transition metal dichalcogenides (TMDCs). We experimentally demonstrate an average total absorption in the visible range (450 nm - 700 nm) of > 70% using > 4 nm of semiconductor absorbing materials scalable over large areas with vapor phase growth techniques. Our analysis suggests that a power conversion efficiency (PCE) of 15.54% and a specific power > 300 W g^-1 may be achieved in a photovoltaic cell based on this metamaterial absorber.

Recent breakthroughs in single-image 3D portrait reconstruction have enabled telepresence systems to stream 3D portrait videos from a single camera in real-time, potentially democratizing telepresence. However, per-frame 3D reconstruction exhibits temporal inconsistency and forgets the user's appearance. On the other hand, self-reenactment methods can render coherent 3D portraits by driving a personalized 3D prior, but fail to faithfully reconstruct the user's per-frame appearance (e.g., facial expressions and lighting). In this work, we recognize the need to maintain both coherent identity and dynamic per-frame appearance to enable the best possible realism. To this end, we propose a new fusion-based method that fuses a personalized 3D subject prior with per-frame information, producing temporally stable 3D videos with faithful reconstruction of the user's per-frame appearances. Trained only using synthetic data produced by an expression-conditioned 3D GAN, our encoder-based method achieves both state-of-the-art 3D reconstruction accuracy and temporal consistency on in-studio and in-the-wild datasets.

Table detection within document images is a crucial task in document processing, involving the identification and localization of tables. Recent strides in deep learning have substantially improved the accuracy of this task, but it still heavily relies on large labeled datasets for effective training. Several semi-supervised approaches have emerged to overcome this challenge, often employing CNN-based detectors with anchor proposals and post-processing techniques like non-maximal suppression (NMS). However, recent advancements in the field have shifted the focus towards transformer-based techniques, eliminating the need for NMS and emphasizing object queries and attention mechanisms. Previous research has focused on two key areas to improve transformer-based detectors: refining the quality of object queries and optimizing attention mechanisms. However, increasing object queries can introduce redundancy, while adjustments to the attention mechanism can increase complexity. To address these challenges, we introduce a semi-supervised approach employing SAM-DETR, a novel approach for precise alignment between object queries and target features. Our approach demonstrates remarkable reductions in false positives and substantial enhancements in table detection performance, particularly in complex documents characterized by diverse table structures. This work provides more efficient and accurate table detection in semi-supervised settings., Comment: ICDAR 2024

Purpose: Software modelling and Model-Driven Engineering (MDE) is traditionally studied from a technical perspective. However, one of the core motivations behind the use of software models is inherently human-centred. Models aim to enable practitioners to communicate about software designs, make software understandable, or make software easier to write through domain-specific modelling languages. Several recent studies challenge the idea that these aims can always be reached and indicate that human factors play a role in the success of MDE. However, there is an under-representation of research focusing on human factors in modelling. Methods: During a GI-Dagstuhl seminar, topics related to human factors in modelling were discussed by 26 expert participants from research and industry. Results: In breakout groups, five topics were covered in depth, namely modelling human aspects, factors of modeller experience, diversity and inclusion in MDE, collaboration and MDE, and teaching human-aware MDE. Conclusion: We summarise our insights gained during the discussions on the five topics. We formulate research goals, questions, and propositions that support directing future initiatives towards an MDE community that is aware of and supportive of human factors and values.

FemTech, a rising trend in mobile apps, empowers women to digitally manage their health and family planning. However, privacy and security vulnerabilities in period-tracking and fertility-monitoring apps present significant risks, such as unintended pregnancies and legal consequences. Our approach involves manual observations of privacy policies and app permissions, along with dynamic and static analysis using multiple evaluation frameworks. Our research reveals that many of these apps gather personally identifiable information (PII) and sensitive healthcare data. Furthermore, our analysis identifies that 61% of the code vulnerabilities found in the apps are classified under the top-ten Open Web Application Security Project (OWASP) vulnerabilities. Our research emphasizes the significance of tackling the privacy and security vulnerabilities present in period-tracking and fertility-monitoring mobile apps. By highlighting these crucial risks, we aim to initiate a vital discussion and advocate for increased accountability and transparency of digital tools for women's health. We encourage the industry to prioritize user privacy and security, ultimately promoting a safer and more secure environment for women's health management., Comment: 16 pages, 1 figure, 7 tables

Recent advances in machine learning have demonstrated that multi-modal pre-training can improve automatic speech recognition (ASR) performance compared to randomly initialized models, even when models are fine-tuned on uni-modal tasks. Existing multi-modal pre-training methods for the ASR task have primarily focused on single-stage pre-training where a single unsupervised task is used for pre-training followed by fine-tuning on the downstream task. In this work, we introduce a novel method combining multi-modal and multi-task unsupervised pre-training with a translation-based supervised mid-training approach. We empirically demonstrate that such a multi-stage approach leads to relative word error rate (WER) improvements of up to 38.45% over baselines on both Librispeech and SUPERB. Additionally, we share several important findings for choosing pre-training methods and datasets., Comment: Accepted in LREC-COLING 2024 - The 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation

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.

Vision language models (VLMs) have experienced rapid advancements through the integration of large language models (LLMs) with image-text pairs, yet they struggle with detailed regional visual understanding due to limited spatial awareness of the vision encoder, and the use of coarse-grained training data that lacks detailed, region-specific captions. To address this, we introduce RegionGPT (short as RGPT), a novel framework designed for complex region-level captioning and understanding. RGPT enhances the spatial awareness of regional representation with simple yet effective modifications to existing visual encoders in VLMs. We further improve performance on tasks requiring a specific output scope by integrating task-guided instruction prompts during both training and inference phases, while maintaining the model's versatility for general-purpose tasks. Additionally, we develop an automated region caption data generation pipeline, enriching the training set with detailed region-level captions. We demonstrate that a universal RGPT model can be effectively applied and significantly enhancing performance across a range of region-level tasks, including but not limited to complex region descriptions, reasoning, object classification, and referring expressions comprehension., Comment: Accepted by CVPR 2024

Field evaporation from ionic or covalently bonded materials often leads to the emission of molecular ions. The metastability of these molecular ions, particularly under the influence of the intense electrostatic field (1010 Vm-1), makes them prone to dissociation with or without an exchange of energy amongst them. These processes can affect the analytical performance of atom probe tomography (APT). For instance, neutral species formed through dissociation may not be detected at all or with a time of flight no longer related to their mass, causing their loss from the analysis. Here, we evaluated the changes in the measured composition of FeO, Fe2O3 and Fe3O4 across a wide range of analysis conditions. Possible dissociation reactions are predicted by density-functional theory (DFT) calculations considering the spin states of the molecules. The energetically favoured reactions are traced on to the multi-hit ion correlation histograms, to confirm their existence within experiments, using an automated Python-based routine. The detected reactions are carefully analysed to reflect upon the influence of these neutrals from dissociation reactions on the performance of APT for analysing iron oxides.

Low-dimensional frustrated magnetic square networks feature a variety of unconventional phases with novel emergent excitations. Often these excitations are intertwined and manifest into intriguing phenomena, an area that has remained largely unexplored in square-lattice systems, especially, double perovskites (A2BB'O6). In this study, we explore these interactions between the fundamental excitations such as phonons and magnons in square-lattice Sr2CuTeO6, Sr2CuWO6, and Ba2CuWO6 isostructural double perovskites that exhibit both short-ranged (TS) as well as long-ranged Neel antiferromagnetic (TN) transitions. Our Raman measurements at variable temperatures reveal an intriguing broad peak (identified as 2-magnon (2M)) surviving beyond TS for W-based compositions contrary to the Te-based system, suggesting a key role of diamagnetic B'-site cation on their magnetism. The thermal response of 2M intriguingly shows signatures of correlation with phonons and control over their anharmonicity, depicting magnon-phonon interaction. Further, a few phonons exhibit anomalies across the magnetic transitions implying the presence of spin-phonon coupling. In particular, the phonon modes at ~ 194 cm-1 of Sr2CuTeO6 and ~ 168 cm-1 of Sr2CuWO6, that show a strong correlation with the 2M, exhibit the strongest spin-phonon coupling suggesting their roles in mediating magnon-phonon interactions in these systems.

In the realm of spoken language understanding (SLU), numerous natural language understanding (NLU) methodologies have been adapted by supplying large language models (LLMs) with transcribed speech instead of conventional written text. In real-world scenarios, prior to input into an LLM, an automated speech recognition (ASR) system generates an output transcript hypothesis, where inherent errors can degrade subsequent SLU tasks. Here we introduce a method that utilizes the ASR system's lattice output instead of relying solely on the top hypothesis, aiming to encapsulate speech ambiguities and enhance SLU outcomes. Our in-context learning experiments, covering spoken question answering and intent classification, underline the LLM's resilience to noisy speech transcripts with the help of word confusion networks from lattices, bridging the SLU performance gap between using the top ASR hypothesis and an oracle upper bound. Additionally, we delve into the LLM's robustness to varying ASR performance conditions and scrutinize the aspects of in-context learning which prove the most influential., Comment: Accepted to ICASSP 2024

While word error rates of automatic speech recognition (ASR) systems have consistently fallen, natural language understanding (NLU) applications built on top of ASR systems still attribute significant numbers of failures to low-quality speech recognition results. Existing assistant systems collect large numbers of these unsuccessful interactions, but these systems usually fail to learn from these interactions, even in an offline fashion. In this work, we introduce CLC: Contrastive Learning for Conversations, a family of methods for contrastive fine-tuning of models in a self-supervised fashion, making use of easily detectable artifacts in unsuccessful conversations with assistants. We demonstrate that our CLC family of approaches can improve the performance of ASR models on OD3, a new public large-scale semi-synthetic meta-dataset of audio task-oriented dialogues, by up to 19.2%. These gains transfer to real-world systems as well, where we show that CLC can help to improve performance by up to 6.7% over baselines. We make OD3 publicly available at https://github.com/amazon-science/amazon-od3 ., Comment: To appear in ICASSP 2024

3D-aware Generative Adversarial Networks (GANs) have shown remarkable progress in learning to generate multi-view-consistent images and 3D geometries of scenes from collections of 2D images via neural volume rendering. Yet, the significant memory and computational costs of dense sampling in volume rendering have forced 3D GANs to adopt patch-based training or employ low-resolution rendering with post-processing 2D super resolution, which sacrifices multiview consistency and the quality of resolved geometry. Consequently, 3D GANs have not yet been able to fully resolve the rich 3D geometry present in 2D images. In this work, we propose techniques to scale neural volume rendering to the much higher resolution of native 2D images, thereby resolving fine-grained 3D geometry with unprecedented detail. Our approach employs learning-based samplers for accelerating neural rendering for 3D GAN training using up to 5 times fewer depth samples. This enables us to explicitly "render every pixel" of the full-resolution image during training and inference without post-processing superresolution in 2D. Together with our strategy to learn high-quality surface geometry, our method synthesizes high-resolution 3D geometry and strictly view-consistent images while maintaining image quality on par with baselines relying on post-processing super resolution. We demonstrate state-of-the-art 3D gemetric quality on FFHQ and AFHQ, setting a new standard for unsupervised learning of 3D shapes in 3D GANs., Comment: See our project page: https://research.nvidia.com/labs/nxp/wysiwyg/