Search

In this paper virtual synchronous generation (VSG) approach is investigated in application to low- and zero-inertia grids operated by grid-forming (GFM) inverters. The key idea here is to introduce dynamic inertia and damping constants in order to keep power gird stable during different types of faults, islanding or large power balance oscillations. In order to achieve such robustness, we introduce frequency and phase angle shift functions to VSG along with dynamics virtual generator parameters. The stability of such approach is theoretically proven and theoretical results are supported by detailed case studies in RTDS (Real-Time Digital Simulator) NovaCor 1.0 with GFM inverters dynamics simulated with 1-3 microseconds timestep using two-level universal inverter model. Case studies include all aforementioned types of faults and demonstrate increased power grid robustness and survivability in comparison with traditional synchronous generation of comparable size.

Text-to-image (T2I) diffusion models are popular for introducing image manipulation methods, such as editing, image fusion, inpainting, etc. At the same time, image-to-video (I2V) and text-to-video (T2V) models are also built on top of T2I models. We present Kandinsky 3, a novel T2I model based on latent diffusion, achieving a high level of quality and photorealism. The key feature of the new architecture is the simplicity and efficiency of its adaptation for many types of generation tasks. We extend the base T2I model for various applications and create a multifunctional generation system that includes text-guided inpainting/outpainting, image fusion, text-image fusion, image variations generation, I2V and T2V generation. We also present a distilled version of the T2I model, evaluating inference in 4 steps of the reverse process without reducing image quality and 3 times faster than the base model. We deployed a user-friendly demo system in which all the features can be tested in the public domain. Additionally, we released the source code and checkpoints for the Kandinsky 3 and extended models. Human evaluations show that Kandinsky 3 demonstrates one of the highest quality scores among open source generation systems., Comment: Accepted for EMNLP 2024 (Demo track)

Direct observation of capture cross-section is challenging due to the need of extremely short filling pulses in the two-gate Deep-Level Transient Spectroscopy (DLTS). Simple estimation of cross-section can be done from DLTS and Admittance Spectroscopy (AS) data, but it is not feasible to distinguish temperature dependence of pre-exponential and exponential parts of the emission rate equation with sufficient precision conducting a single experiment. This paper presents experimental data of deep-levels in $\beta$-Ga2O3 that has been gathered by our group since 2017. Based on the gathered data we propose a derivation of apparent activation energy ($E_a^m$) and capture cross-section ($\sigma_n^m$) assuming temperature dependent capture via multiphonon emission model, which resulted in strong correlation between $E_a^m$ and $\sigma_n^m$ according to Meyer-Neldel rule, which allowed us to estimate low- and high-temperature capture coefficients $C_0$ and $C_1$ as well as capture barrier $E_b$. It also has been shown that without considering the temperature dependence of capture cross-section, the experimental values of $\sigma_n$ are overestimated by 1-3 orders of magnitude. A careful consideration of the data also allows to be more certain identifying deep-levels by their "fingerprints" ($E_a$ and $\sigma_n$) considering two additional parameters ($E_{MN}$ and $\sigma_{00}$) and to verify the density functional theory (DFT) computation of deep-level recombination properties.

The growing reproducibility crisis in machine learning has brought forward a need for careful examination of research findings. This paper investigates the claims made by Lei et al. (2023) regarding their proposed method, LICO, for enhancing post-hoc interpretability techniques and improving image classification performance. LICO leverages natural language supervision from a vision-language model to enrich feature representations and guide the learning process. We conduct a comprehensive reproducibility study, employing (Wide) ResNets and established interpretability methods like Grad-CAM and RISE. We were mostly unable to reproduce the authors' results. In particular, we did not find that LICO consistently led to improved classification performance or improvements in quantitative and qualitative measures of interpretability. Thus, our findings highlight the importance of rigorous evaluation and transparent reporting in interpretability research., Comment: 15 pages, 2 figures, Machine Learning Reproducibility Challenge 2024

In the study of reading, eye-tracking technology offers unique insights into the time-course of how individuals extract information from text. A significant hurdle in using multi-line paragraph stimuli is the need to align eye gaze position with the correct line. This is made more difficult by positional noise in the eye-tracking data, primarily due to vertical drift, and often necessitates manual intervention. Such manual correction is labor-intensive, subjective, and limits the scalability of research efforts. As a result, automated solutions are desirable, especially those that do not require extensive technical skills and still allow close control over the outcome. To address this, we introduce GazeGenie: a comprehensive software solution designed specifically for researchers in eye-tracking studies on multi-line reading. Accessible via an intuitive web browser-based user interface and easily installed using Docker, GazeGenie streamlines the entire data processing pipeline from parsing fixations from raw data to calculation of word and sentence-based measures based on cleaned and drift-corrected fixations. The software's core features include the recently introduced Dual Input Stream Transformer (DIST) model and various classical algorithms all of which can be combined within a Wisdom of the Crowds (WOC) approach to enhance accuracy in fixation line-assignment. By providing an all-in-one solution for researchers, we hope to make automated fixation alignment more accessible, reducing researchers' reliance on manual intervention in vertical fixation alignment. This should lead to more accurate, efficient, and reproducible analyses of multi-line eye-movement data and pave the way to enabling larger scale studies to be carried out., Comment: This work has been submitted to the Springer Behavior Research Methods for possible publication

We develop and evaluate neural architectures to model the user behavior in recommender systems (RS) inspired by click models for Web search but going beyond standard click models. Proposed architectures include recurrent networks, Transformer-based models that alleviate the quadratic complexity of self-attention, adversarial and hierarchical architectures. Our models outperform baselines on the ContentWise and RL4RS datasets and can be used in RS simulators to model user response for RS evaluation and pretraining.

The goal of modern sequential recommender systems is often formulated in terms of next-item prediction. In this paper, we explore the applicability of generative transformer-based models for the Top-K sequential recommendation task, where the goal is to predict items a user is likely to interact with in the "near future". We explore commonly used autoregressive generation strategies, including greedy decoding, beam search, and temperature sampling, to evaluate their performance for the Top-K sequential recommendation task. In addition, we propose novel Reciprocal Rank Aggregation (RRA) and Relevance Aggregation (RA) generation strategies based on multi-sequence generation with temperature sampling and subsequent aggregation. Experiments on diverse datasets give valuable insights regarding commonly used strategies' applicability and show that suggested approaches improve performance on longer time horizons compared to widely-used Top-K prediction approach and single-sequence autoregressive generation strategies.

Data sparsity has been one of the long-standing problems for recommender systems. One of the solutions to mitigate this issue is to exploit knowledge available in other source domains. However, many cross-domain recommender systems introduce a complex architecture that makes them less scalable in practice. On the other hand, matrix factorization methods are still considered to be strong baselines for single-domain recommendations. In this paper, we introduce the CDIMF, a model that extends the standard implicit matrix factorization with ALS to cross-domain scenarios. We apply the Alternating Direction Method of Multipliers to learn shared latent factors for overlapped users while factorizing the interaction matrix. In a dual-domain setting, experiments on industrial datasets demonstrate a competing performance of CDIMF for both cold-start and warm-start. The proposed model can outperform most other recent cross-domain and single-domain models. We also provide the code to reproduce experiments on GitHub.

Machine learning (ML) models trained on datasets owned by different organizations and physically located in remote databases offer benefits in many real-world use cases. State regulations or business requirements often prevent data transfer to a central location, making it difficult to utilize standard machine learning algorithms. Federated Learning (FL) is a technique that enables models to learn from distributed datasets without revealing the original data. Vertical Federated learning (VFL) is a type of FL where data samples are divided by features across several data owners. For instance, in a recommendation task, a user can interact with various sets of items, and the logs of these interactions are stored by different organizations. In this demo paper, we present \emph{Stalactite} - an open-source framework for VFL that provides the necessary functionality for building prototypes of VFL systems. It has several advantages over the existing frameworks. In particular, it allows researchers to focus on the algorithmic side rather than engineering and to easily deploy learning in a distributed environment. It implements several VFL algorithms and has a built-in homomorphic encryption layer. We demonstrate its use on a real-world recommendation datasets.

Using a single tool to build and compare recommender systems significantly reduces the time to market for new models. In addition, the comparison results when using such tools look more consistent. This is why many different tools and libraries for researchers in the field of recommendations have recently appeared. Unfortunately, most of these frameworks are aimed primarily at researchers and require modification for use in production due to the inability to work on large datasets or an inappropriate architecture. In this demo, we present our open-source toolkit RePlay - a framework containing an end-to-end pipeline for building recommender systems, which is ready for production use. RePlay also allows you to use a suitable stack for the pipeline on each stage: Pandas, Polars, or Spark. This allows the library to scale computations and deploy to a cluster. Thus, RePlay allows data scientists to easily move from research mode to production mode using the same interfaces.

Sequential recommender systems are an important and demanded area of research. Such systems aim to use the order of interactions in a user's history to predict future interactions. The premise is that the order of interactions and sequential patterns play an essential role. Therefore, it is crucial to use datasets that exhibit a sequential structure to evaluate sequential recommenders properly. We apply several methods based on the random shuffling of the user's sequence of interactions to assess the strength of sequential structure across 15 datasets, frequently used for sequential recommender systems evaluation in recent research papers presented at top-tier conferences. As shuffling explicitly breaks sequential dependencies inherent in datasets, we estimate the strength of sequential patterns by comparing metrics for shuffled and original versions of the dataset. Our findings show that several popular datasets have a rather weak sequential structure.

Hybrid systems of two-dimensional (2D) materials (such as graphene-family materials and 2D transition metal dichalcogenides) are attracting much attention due to their distinctive optoelectronic, thermal, mechanical, and chemical properties. The application perspectives of these materials in various fields further expand when enriching those with liquid crystals (LCs) primarily due to their enhanced tunability and functionality. In this study, we report on the hydrothermal synthesis of hybrid nanocomposites composed of MoS$_2$ and rGO and discuss tuning possibilities of their electro-optical properties by incorporating thermotropic LCs. In particular, we demonstrate that the incorporation of 5CB LC increases the sensitivity and charge storage efficiency of the hybrid nanocomposites. In addition, we also present the responsivity, detectivity, and response time properties of the hybrid nanocomposites of MoS$_2$/rGO, both with and without the inclusion of nematic LCs. Furthermore, we demonstrate that the system exhibits a 5CB-induced photocurrent switching effect. We believe the findings will open new doors for applications of these materials in optoelectronics and photonics., Comment: 3 figures, 113036 page, research paper

Traps and structural defects at the hole and electron transport interfaces of the microcrystalline absorber limits the efficiency and long-term stability of perovskite solar cells (PSCs) due to accumulation of the ionic clusters, non-radiative recombination and electrochemical corrosion. Surface engineering using self-assembled monolayers (SAM) was considered as an effective strategy for modification of charge-collection junctions. In this work, we demonstrate the first report about complex integration of a SAM for double-side passivation in p-i-n PSCs. Integrating the novel 5-(4-[bis(4-fluorophenyl)amino]phenyl)thiophene-2-carboxylic acid (FTPATC) as a fluorinated SAM at the hole-transport interface reduced potential barriers and lattice stresses in the absorber. At the electron-transport side, FTPATC interacted with the A-site cations of the perovskite molecule (Cs, formamidinium), inducing a dipole for defect compensation. Using the passivation approach with fluorinated SAM demonstrated benefits in the gain of the output performance up to 22.2%. The key-advantage of double-side passivation was confirmed by the enhanced stability under continuous light-soaking (1-sun equivalent, 65 C, ISOS-L-2), maintaining 88% of the initial performance over 1680 hours and thermal stabilization under harsh heating at 90 C.

Event sequences (ESs) arise in many practical domains including finance, retail, social networks, and healthcare. In the context of machine learning, event sequences can be seen as a special type of tabular data with annotated timestamps. Despite the importance of ESs modeling and analysis, little effort was made in adapting large language models (LLMs) to the ESs domain. In this paper, we highlight the common difficulties of ESs processing and propose a novel solution capable of solving multiple downstream tasks with little or no finetuning. In particular, we solve the problem of working with long sequences and improve time and numeric features processing. The resulting method, called ESQA, effectively utilizes the power of LLMs and, according to extensive experiments, achieves state-of-the-art results in the ESs domain., Comment: 25 pages, 3 figures

We report on microlasers based on high-quality micropillars with lasing on whispering-gallery modes. Usage of low-absorbing Al$\scriptsize 0.2$Ga$\scriptsize 0.8$As\Al$\scriptsize 0.9$Ga$\scriptsize 0.1$As distributed Bragg reflectors as well as the smooth pillar sidewalls allows us to realize whispering-gallery modes lasing by excitation and collection of emission in the pillar axis direction. As a result, simultaneous whispering gallery modes lasing (comb-like structure) in the wavelength range of 930-970 nm is observed for 3-7 $\mu$m pillar diameters. Increase the temperature up to 130 K results single-mode lasing for 5 $\mu$m pillars with about 8000 cold cavity quality-factor and 240 $\mu$W estimated threshold excitation power.

Uncertainty quantification (UQ) is a critical component of machine learning (ML) applications. The rapid proliferation of large language models (LLMs) has stimulated researchers to seek efficient and effective approaches to UQ for text generation. As with other ML models, LLMs are prone to making incorrect predictions, in the form of ``hallucinations'' whereby claims are fabricated or low-quality outputs are generated for a given input. UQ is a key element in dealing with these challenges. However, research to date on UQ methods for LLMs has been fragmented, in terms of the literature on UQ techniques and evaluation methods. In this work, we tackle this issue by introducing a novel benchmark that implements a collection of state-of-the-art UQ baselines, and provides an environment for controllable and consistent evaluation of novel UQ techniques over various text generation tasks. Our benchmark also supports the assessment of confidence normalization methods in terms of their ability to provide interpretable scores. Using our benchmark, we conduct a large-scale empirical investigation of UQ and normalization techniques across nine tasks, and identify the most promising approaches. Code: https://github.com/IINemo/lm-polygraph, Comment: Roman Vashurin, Ekaterina Fadeeva, Artem Vazhentsev contributed equally

We discuss the possibility of a scalar field being the fundamental description of dark energy. We focus on shift-symmetric scalar-tensor theories since this symmetry potentially avoids some fine-tuning problems. We also restrict attention to theories satisfying that the propagation speed of gravitational waves is equal to the speed of light. These considerations lead us to investigate shift-symmetric Kinetic Gravity Braiding theories. Analysing the stability of scalar linear perturbations, we discuss the conditions that seems to be necessary to describe (super) accelerated cosmic expansion without introducing instabilities. However, it has been previously established that the linearised analysis does not guarantee the stability of this non-canonical scalar theory, as potentially dangerous interactions between dark energy fluctuations and tensor perturbations (essentially gravitational waves) appear at a higher order in perturbation theory. Indeed, although we shall point out that the standard proof of absence of dark energy stable braiding models due to this interaction has a possible way-out, we find general arguments suggesting that there are no dark energy stable solutions that can exploit this loophole. Thus, we discuss future research directions for finding viable fundamental descriptions of dark energy. We also provide a dictionary between the covariant version of the scalar field theory and that of the Effective Field Theory approach, explicitly computing the parameters in the latter formalism in terms of the functions appearing in the covariant version, and its derivatives. To the best of our knowledge, this is the first time these expressions are explicitly obtained up-to arbitrary order in perturbation theory., Comment: 21 pages, 6 figures. v2: 22 pages, references added, typos corrected, minor changes in the text, no physics changes. This version accepted for publication in Physics of the Dark Universe

The use of self-assembled monolayers (SAMs) with anchoring groups was considered as an effective approach for interface engineering in perovskite solar cells with metal oxide charge transporting layers. However, the coating of SAM layers in PSMs by means of a slot-die is a challenging process due to the low viscosity of the solutions and the low wettability of the films. In this study, we integrate a triphenylamine-based polymer, pTPA-TDP, blended with SAM based on 5-[4-[4-(diphenylamino) phenyl] thiophene-2-carboxylic acid (TPATC), to address the challenges of uniform slot-die coating and interface passivation in large-area modules. We fabricated p-i-n oriented PSMs on 50x50 mm2 substrates (12-sub-cells) with NiO hole transport layer (HTL) and organic interlayers for surface modification. Wetting angle mapping demonstrated that ununiform regions of the slot-die coated SAM have hydrophobicity with contact angle values up to 90{\deg}, causing fluctuations in absorber thickness and the presence of macro-defects at buried interfaces. The incorporation of the blended interlayer to NiO/perovskite junction homogenized the surface wettability (contact angle=40{\deg}) and mitigated lattice strain in the absorber. This enabled the effective use of SAM properties on a large-area surface, improving energy level alignment and enhancing the power conversion efficiency (PCE) of the modules from 13.98% to 15.83% and stability (ISOS-L-2, T80 period) from 500-1000 hours to 1630 hours. Investigation of PSMs upon cooling till -5 {\deg}C showed that the PCE increased by +0.19%/{\deg}C for samples with NiO HTL, while using SAM and blended interlayers raised the coefficient to ~0.40%/{\deg}C due to changes in activation energy and trap contributions to device performance across a wide temperature range.

We use the Hamiltonian reduction method to construct the Ruijsenaars dual systems to generalized Toda chains associated with the classical Lie algebras of types $B, C, D$. The dual systems turn out to be the $B, C$ and $D$ analogues of the rational Goldfish model, which is, as in the type $A$ case, the strong coupling limit of rational Ruijsenaars systems. We explain how both types of systems emerge in the reduction of the cotangent bundle of a Lie group and provide the formulae for dual Hamiltonians. We compute explicitly the higher Hamiltonians of Goldfish models using the Cauchy--Binet theorem., Comment: 28 pages, Corrections added

In the rapidly evolving domain of Recommender Systems (RecSys), new algorithms frequently claim state-of-the-art performance based on evaluations over a limited set of arbitrarily selected datasets. However, this approach may fail to holistically reflect their effectiveness due to the significant impact of dataset characteristics on algorithm performance. Addressing this deficiency, this paper introduces a novel benchmarking methodology to facilitate a fair and robust comparison of RecSys algorithms, thereby advancing evaluation practices. By utilizing a diverse set of $30$ open datasets, including two introduced in this work, and evaluating $11$ collaborative filtering algorithms across $9$ metrics, we critically examine the influence of dataset characteristics on algorithm performance. We further investigate the feasibility of aggregating outcomes from multiple datasets into a unified ranking. Through rigorous experimental analysis, we validate the reliability of our methodology under the variability of datasets, offering a benchmarking strategy that balances quality and computational demands. This methodology enables a fair yet effective means of evaluating RecSys algorithms, providing valuable guidance for future research endeavors., Comment: 8 pages with 11 figures

The evolution of dynamic processes in graphene-family materials are of great interest for both scientific purposes and technical applications. Scanning electron microscopy and transmission electron microscopy outstand among the techniques that allow both observing and controlling such dynamic processes in real time. On the other hand, functionalized graphene oxide emerges as a favorable candidate from graphene-family materials for such an investigation due to its distinctive properties, that encompass a large surface area, robust thermal stability, and noteworthy electrical and mechanical properties after its reduction. Here, we report on studies of surface structure and adsorption dynamics of L-Cysteine on electrochemically exfoliated graphene oxides basal plane. We show that electron beam irradiation prompts an amorphization of functionalized graphene oxide along with the formation of micropatterns of controlled geometry composed of L-Cysteine-Graphene oxide nanostructures. The controlled growth and predetermined arrangement of micropatterns as well as controlled structure disorder induced by e beam amorphization, in its turn potentially offering tailored properties and functionalities paving the way for potential applications in nanotechnology, sensor development, and surface engineering. Our findings demonstrate that graphene oxide can cover L-Cysteine in such a way to provide a control on the positioning of emerging microstructures about 10-20 um in diameter. Besides, Raman and SAED measurement analyses yield above 50% amorphization in a material. The results of our studies demonstrate that such a technique enables the direct creation of micropatterns of L-Cysteine-Graphene oxide eliminating the need for complicated mask patterning procedures.

In this study, we present a complex investigation for miniaturizing of perovskite photodiodes (PPDs) in various geometries with use of ultraviolet laser scribing (UV-LS). Employing a 355 nm (3.5 eV) pulsed laser at 30 kHz, we successfully manufactured PPDs with pixel configurations of 70x130 um2, 520x580 um2, and 2000x2000 um2. The utilization of UV-LS has a proven efficiency in achieving relevant diode characteristics, such as low dark currents and high shunt resistance, as well as ultrafast response. The multi-step scribing cycle provided precise patterning of perovskite photodiodes (PPDs) in a string design. The dark current densities demonstrated exceptional uniformity, ranging from 10-10 A/cm2 for 2000x2000 um2 pixelated PPDs to 10-9 A/cm2 for the 70x130 um2 configuration. The string PPDs, consisting of 10 pixels per string, displayed homogenous dark current values, ensuring effective isolation between devices. Under green light illumination (540 nm), all PPD types exhibited a broad Linear Dynamic Range (LDR). Specifically, LDR values reached 110 dB, 117 dB, and 136 dB for 70x130, 520x580, and 2000x2000 devices, respectively, spanning an illumination intensity range from 2*10-3 mW/cm2 to 2 mW/cm2. High responsivity values up to 0.38 A/W, depending on the PPDs geometry, highlight the potential of laser scribing devices for sensing in the visible range. The calculated specific detectivity performance (from 1011 to 1013 Jones) surpasses commercial analogs, while the sub-microsecond response of 70x130 um2 and 520x580 um2 miniaturized devices underscores their suitability for precise time resolution detection systems.

We present Kandinsky 3.0, a large-scale text-to-image generation model based on latent diffusion, continuing the series of text-to-image Kandinsky models and reflecting our progress to achieve higher quality and realism of image generation. In this report we describe the architecture of the model, the data collection procedure, the training technique, and the production system for user interaction. We focus on the key components that, as we have identified as a result of a large number of experiments, had the most significant impact on improving the quality of our model compared to the others. We also describe extensions and applications of our model, including super resolution, inpainting, image editing, image-to-video generation, and a distilled version of Kandinsky 3.0 - Kandinsky 3.1, which does inference in 4 steps of the reverse process and 20 times faster without visual quality decrease. By side-by-side human preferences comparison, Kandinsky becomes better in text understanding and works better on specific domains. The code is available at https://github.com/ai-forever/Kandinsky-3, Comment: Project page: https://ai-forever.github.io/Kandinsky-3

A novel triphenylamine-based hole transport material (HTM) with a carboxyl anchoring group (TPATC) was developed for tuning the interface between nanocrystalline NiO and double cation CsCH3(NH2)2PbI3-xClx absorber in p-i-n device architectures. We present a unique comprehensive confinement study of PSCs with TPATC as the self-assembled HTM, including analysis of numerical defect parameters, phase composition evolution, and up-scaling capabilities. Our investigation shows that the ultrathin TPATC interlayer effectively passivates traps, increases work-function of HTL by about ~0.2 eV, and enhances the charge carrier extraction efficiency. Advanced transient spectroscopy measurements revealed that modification of the NiO surface with TPATC in perovskite solar cells (PSCs) reduces the concentration of ionic defects by an order of magnitude. Interface engineering with TPATC allowed to reach power conversion efficiency of 20.58% for small area devices (0.15 cm2) under standard AM 1.5 G conditions. Using TPATC interlayer also provided stabilized performance of PSCs under operation conditions and improved sustainability of the perovskite absorber to decomposition. After continuous light-soaking (1000 h, ISOS-L-2 protocol), NiO/TPATC devices showed a slight decrease of 2% in maximum power. In contrast, NiO PSCs demonstrated decrease in power output (>20%) after 400 h. We explored the potential of TPATC to modify interfaces in large-area perovskite solar modules (PSM, active area-64.8 cm2, 12 sub-cells). By applying slot-die-coated TPATC, the PCE at AM 1.5 G conditions increased from 13.22% for NiO PSM to 15.64% for NiO/TPATC ones. This study provides new insights into the interface engineering for p-i-n perovskite solar cells, behavior of the ionic defects and their contribution to the long-term stability.

Multimedia generation approaches occupy a prominent place in artificial intelligence research. Text-to-image models achieved high-quality results over the last few years. However, video synthesis methods recently started to develop. This paper presents a new two-stage latent diffusion text-to-video generation architecture based on the text-to-image diffusion model. The first stage concerns keyframes synthesis to figure the storyline of a video, while the second one is devoted to interpolation frames generation to make movements of the scene and objects smooth. We compare several temporal conditioning approaches for keyframes generation. The results show the advantage of using separate temporal blocks over temporal layers in terms of metrics reflecting video generation quality aspects and human preference. The design of our interpolation model significantly reduces computational costs compared to other masked frame interpolation approaches. Furthermore, we evaluate different configurations of MoVQ-based video decoding scheme to improve consistency and achieve higher PSNR, SSIM, MSE, and LPIPS scores. Finally, we compare our pipeline with existing solutions and achieve top-2 scores overall and top-1 among open-source solutions: CLIPSIM = 0.2976 and FVD = 433.054. Project page: https://ai-forever.github.io/kandinsky-video/, Comment: Project page: https://ai-forever.github.io/kandinsky-video/

Recent advancements in the capabilities of large language models (LLMs) have paved the way for a myriad of groundbreaking applications in various fields. However, a significant challenge arises as these models often "hallucinate", i.e., fabricate facts without providing users an apparent means to discern the veracity of their statements. Uncertainty estimation (UE) methods are one path to safer, more responsible, and more effective use of LLMs. However, to date, research on UE methods for LLMs has been focused primarily on theoretical rather than engineering contributions. In this work, we tackle this issue by introducing LM-Polygraph, a framework with implementations of a battery of state-of-the-art UE methods for LLMs in text generation tasks, with unified program interfaces in Python. Additionally, it introduces an extendable benchmark for consistent evaluation of UE techniques by researchers, and a demo web application that enriches the standard chat dialog with confidence scores, empowering end-users to discern unreliable responses. LM-Polygraph is compatible with the most recent LLMs, including BLOOMz, LLaMA-2, ChatGPT, and GPT-4, and is designed to support future releases of similarly-styled LMs., Comment: Accepted at EMNLP-2023

We introduce a novel Dual Input Stream Transformer (DIST) for the challenging problem of assigning fixation points from eye-tracking data collected during passage reading to the line of text that the reader was actually focused on. This post-processing step is crucial for analysis of the reading data due to the presence of noise in the form of vertical drift. We evaluate DIST against eleven classical approaches on a comprehensive suite of nine diverse datasets. We demonstrate that combining multiple instances of the DIST model in an ensemble achieves high accuracy across all datasets. Further combining the DIST ensemble with the best classical approach yields an average accuracy of 98.17 %. Our approach presents a significant step towards addressing the bottleneck of manual line assignment in reading research. Through extensive analysis and ablation studies, we identify key factors that contribute to DIST's success, including the incorporation of line overlap features and the use of a second input stream. Via rigorous evaluation, we demonstrate that DIST is robust to various experimental setups, making it a safe first choice for practitioners in the field., Comment: This work has been submitted to the IEEE Transactions on pattern analysis and machine intelligence for possible publication

Background and objective: CTA is a gold standard of preoperative diagnosis of abdominal aorta and typically used for geometric-only characteristic extraction. We assume that a model describing the dynamic behavior of the contrast agent in the vessel can be developed from the data of routine CTA studies, allowing the procedure to be investigated and optimized without the need for additional perfusion CT studies. Obtained spatial distribution of CA can be valuable for both increasing the diagnostic value of a particular study and improving the CT data processing tools. Methods: In accordance with the Beer-Lambert law and the absence of chemical interaction between blood and CA, we postulated the existence of a deterministic CA-induced component in the CT signal density. The proposed model, having a double-sigmoid structure, contains six coefficients relevant to the properties of hemodynamics. To validate the model, expert segmentation was performed using the 3D Slicer application for the CTA data obtained from publicly available source. The model was fitted to the data using the non-linear least square method with Levenberg-Marquardt optimization. Results: We analyzed 594 CTA images (4 studies with median size of 144 slices, IQR [134; 158.5]; 1:1 normal:pathology balance). Goodness-of-fit was proved by Wilcox test (p-value > 0.05 for all cases). The proposed model correctly simulated normal blood flow and hemodynamics disturbances caused by local abnormalities (aneurysm, thrombus and arterial branching). Conclusions: Proposed approach can be useful for personalized CA modeling of vessels, improvement of CTA image processing and preparation of synthetic CT training data for artificial intelligence.

Recently sequential recommendations and next-item prediction task has become increasingly popular in the field of recommender systems. Currently, two state-of-the-art baselines are Transformer-based models SASRec and BERT4Rec. Over the past few years, there have been quite a few publications comparing these two algorithms and proposing new state-of-the-art models. In most of the publications, BERT4Rec achieves better performance than SASRec. But BERT4Rec uses cross-entropy over softmax for all items, while SASRec uses negative sampling and calculates binary cross-entropy loss for one positive and one negative item. In our work, we show that if both models are trained with the same loss, which is used by BERT4Rec, then SASRec will significantly outperform BERT4Rec both in terms of quality and training speed. In addition, we show that SASRec could be effectively trained with negative sampling and still outperform BERT4Rec, but the number of negative examples should be much larger than one.