Vladimir Ivanov

Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri et al. · allen-ai, berkeley

Modern artificial intelligence (AI) systems are powered by foundation models. This paper presents a new set of foundation models, called Llama 3. It is a herd of language models that natively support multilinguality, coding, reasoning, and tool usage. Our largest model is a dense Transformer with 405B parameters and a context window of up to 128K tokens. This paper presents an extensive empirical evaluation of Llama 3. We find that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development.

Natalia Loukachevitch, Suresh Manandhar, Elina Baral et al.

This paper describes NEREL-BIO -- an annotation scheme and corpus of PubMed abstracts in Russian and smaller number of abstracts in English. NEREL-BIO extends the general domain dataset NEREL by introducing domain-specific entity types. NEREL-BIO annotation scheme covers both general and biomedical domains making it suitable for domain transfer experiments. NEREL-BIO provides annotation for nested named entities as an extension of the scheme employed for NEREL. Nested named entities may cross entity boundaries to connect to shorter entities nested within longer entities, making them harder to detect. NEREL-BIO contains annotations for 700+ Russian and 100+ English abstracts. All English PubMed annotations have corresponding Russian counterparts. Thus, NEREL-BIO comprises the following specific features: annotation of nested named entities, it can be used as a benchmark for cross-domain (NEREL -> NEREL-BIO) and cross-language (English -> Russian) transfer. We experiment with both transformer-based sequence models and machine reading comprehension (MRC) models and report their results. The dataset is freely available at https://github.com/nerel-ds/NEREL-BIO.

Ekaterina Artemova, Maxim Zmeev, Natalia Loukachevitch et al.

The RuNNE Shared Task approaches the problem of nested named entity recognition. The annotation schema is designed in such a way, that an entity may partially overlap or even be nested into another entity. This way, the named entity "The Yermolova Theatre" of type "organization" houses another entity "Yermolova" of type "person". We adopt the Russian NEREL dataset for the RuNNE Shared Task. NEREL comprises news texts written in the Russian language and collected from the Wikinews portal. The annotation schema includes 29 entity types. The nestedness of named entities in NEREL reaches up to six levels. The RuNNE Shared Task explores two setups. (i) In the general setup all entities occur more or less with the same frequency. (ii) In the few-shot setup the majority of entity types occur often in the training set. However, some of the entity types are have lower frequency, being thus challenging to recognize. In the test set the frequency of all entity types is even. This paper reports on the results of the RuNNE Shared Task. Overall the shared task has received 156 submissions from nine teams. Half of the submissions outperform a straightforward BERT-based baseline in both setups. This paper overviews the shared task setup and discusses the submitted systems, discovering meaning insights for the problem of nested NER. The links to the evaluation platform and the data from the shared task are available in our github repository: https://github.com/dialogue-evaluation/RuNNE.

Omkar Salpekar, Rohan Varma, Kenny Yu et al.

Large-scale training systems typically use synchronous training, requiring all GPUs to be healthy simultaneously. In our experience training on O(100K) GPUs, synchronous training results in a low efficiency due to frequent failures and long recovery time. To address this problem, we propose a novel training paradigm, Fault Tolerant Hybrid-Shared Data Parallelism (FT-HSDP). FT-HSDP uses data parallel replicas as units of fault tolerance. When failures occur, only a single data-parallel replica containing the failed GPU or server is taken offline and restarted, while the other replicas continue training. To realize this idea at scale, FT-HSDP incorporates several techniques: 1) We introduce a Fault Tolerant All Reduce (FTAR) protocol for gradient exchange across data parallel replicas. FTAR relies on the CPU to drive the complex control logic for tasks like adding or removing participants dynamically, and relies on GPU to perform data transfer for best performance. 2) We introduce a non-blocking catch-up protocol, allowing a recovering replica to join training with minimal stall. Compared with fully synchronous training at O(100K) GPUs, FT-HSDP can reduce the stall time due to failure recovery from 10 minutes to 3 minutes, increasing effective training time from 44\% to 80\%. We further demonstrate that FT-HSDP's asynchronous recovery does not bring any meaning degradation to the accuracy of the result model.

Ariana Azarbal, Victor Gillioz, Vladimir Ivanov et al.

Developers often struggle to specify correct training labels and rewards. Perhaps they don't need to. We propose recontextualization, which reduces how often language models "game" training signals, performing misbehaviors those signals mistakenly reinforce. We show recontextualization prevents models from learning to 1) prioritize evaluation metrics over chat response quality; 2) special-case code to pass incorrect tests; 3) lie to users; and 4) become sycophantic. Our method works by generating completions from prompts discouraging misbehavior and then recontextualizing them as though they were in response to prompts permitting misbehavior. Recontextualization trains language models to resist misbehavior even when instructions permit it. This mitigates the reinforcement of misbehavior from misspecified training signals, reducing specification gaming without improving the supervision signal.

Danis Alukaev, Semen Kiselev, Ilya Pershin et al.

Concept Bottleneck Models (CBMs) assume that training examples (e.g., x-ray images) are annotated with high-level concepts (e.g., types of abnormalities), and perform classification by first predicting the concepts, followed by predicting the label relying on these concepts. The main difficulty in using CBMs comes from having to choose concepts that are predictive of the label and then having to label training examples with these concepts. In our approach, we adopt a more moderate assumption and instead use text descriptions (e.g., radiology reports), accompanying the images in training, to guide the induction of concepts. Our cross-modal approach treats concepts as discrete latent variables and promotes concepts that (1) are predictive of the label, and (2) can be predicted reliably from both the image and text. Through experiments conducted on datasets ranging from synthetic datasets (e.g., synthetic images with generated descriptions) to realistic medical imaging datasets, we demonstrate that cross-modal learning encourages the induction of interpretable concepts while also facilitating disentanglement. Our results also suggest that this guidance leads to increased robustness by suppressing the reliance on shortcut features.

Andrei Semenov, Vladimir Ivanov, Aleksandr Beznosikov et al.

We propose a novel architecture and method of explainable classification with Concept Bottleneck Models (CBMs). While SOTA approaches to Image Classification task work as a black box, there is a growing demand for models that would provide interpreted results. Such a models often learn to predict the distribution over class labels using additional description of this target instances, called concepts. However, existing Bottleneck methods have a number of limitations: their accuracy is lower than that of a standard model and CBMs require an additional set of concepts to leverage. We provide a framework for creating Concept Bottleneck Model from pre-trained multi-modal encoder and new CLIP-like architectures. By introducing a new type of layers known as Concept Bottleneck Layers, we outline three methods for training them: with $\ell_1$-loss, contrastive loss and loss function based on Gumbel-Softmax distribution (Sparse-CBM), while final FC layer is still trained with Cross-Entropy. We show a significant increase in accuracy using sparse hidden layers in CLIP-based bottleneck models. Which means that sparse representation of concepts activation vector is meaningful in Concept Bottleneck Models. Moreover, with our Concept Matrix Search algorithm we can improve CLIP predictions on complex datasets without any additional training or fine-tuning. The code is available at: https://github.com/Andron00e/SparseCBM.

Vladimir Ivanov, Valery Solovyev

Concrete/abstract words are used in a growing number of psychological and neurophysiological research. For a few languages, large dictionaries have been created manually. This is a very time-consuming and costly process. To generate large high-quality dictionaries of concrete/abstract words automatically one needs extrapolating the expert assessments obtained on smaller samples. The research question that arises is how small such samples should be to do a good enough extrapolation. In this paper, we present a method for automatic ranking concreteness of words and propose an approach to significantly decrease amount of expert assessment. The method has been evaluated on a large test set for English. The quality of the constructed dictionaries is comparable to the expert ones. The correlation between predicted and expert ratings is higher comparing to the state-of-the-art methods.

Vladimir Makharev, Vladimir Ivanov

Code summarization is a critical task in natural language processing and software engineering, which aims to generate concise descriptions of source code. Recent advancements have improved the quality of these summaries, enhancing code readability and maintainability. However, the content of a repository or a class has not been considered in function code summarization. This study investigated the effectiveness of code summarization models beyond the function level, exploring the impact of class and repository contexts on the summary quality. The study involved revising benchmarks for evaluating models at class and repository levels, assessing baseline models, and evaluating LLMs with in-context learning to determine the enhancement of summary quality with additional context. The findings revealed that the fine-tuned state-of-the-art CodeT5+ base model excelled in code summarization, while incorporating few-shot learning and retrieved code chunks from RAG significantly enhanced the performance of LLMs in this task. Notably, the Deepseek Coder 1.3B and Starcoder2 15B models demonstrated substantial improvements in metrics such as BLEURT, METEOR, and BLEU-4 at both class and repository levels. Repository-level summarization exhibited promising potential but necessitates significant computational resources and gains from the inclusion of structured context. Lastly, we employed the recent SIDE code summarization metric in our evaluation. This study contributes to refining strategies for prompt engineering, few-shot learning, and RAG, addressing gaps in benchmarks for code summarization at various levels. Finally, we publish all study details, code, datasets, and results of evaluation in the GitHub repository available at https://github.com/kilimanj4r0/code-summarization-beyond-function-level.

Georgy Andryushchenko, Vladimir Ivanov, Vladimir Makharev et al.

Question answering over source code provides software engineers and project managers with helpful information about the implemented features of a software product. This paper presents a work devoted to using large language models for question answering over source code in Python. The proposed method for implementing a source code question answering system involves fine-tuning a large language model on a unified dataset of questions and answers for Python code. To achieve the highest quality answers, we tested various models trained on datasets preprocessed in different ways: a dataset without grammar correction, a dataset with grammar correction, and a dataset augmented with the generated summaries. The model answers were also analyzed for errors manually. We report BLEU-4, BERTScore F1, BLEURT, and Exact Match metric values, along with the conclusions from the manual error analysis. The obtained experimental results highlight the current problems of the research area, such as poor quality of the public genuine question-answering datasets. In addition, the findings include the positive effect of the grammar correction of the training data on the testing metric values. The addressed findings and issues could be important for other researchers who attempt to improve the quality of source code question answering solutions. The training and evaluation code is publicly available at https://github.com/IU-AES-AI4Code/CodeQuestionAnswering.

Artūrs Kanepajs, Vladimir Ivanov, Richard Moulange

Linguistically inclusive LLMs -- which maintain good performance regardless of the language with which they are prompted -- are necessary for the diffusion of AI benefits around the world. Multilingual jailbreaks that rely on language translation to evade safety measures undermine the safe and inclusive deployment of AI systems. We provide policy recommendations to enhance the multilingual capabilities of AI while mitigating the risks of multilingual jailbreaks. We examine how a language's level of resourcing relates to how vulnerable LLMs are to multilingual jailbreaks in that language. We do this by testing five advanced AI models across 24 official languages of the EU. Building on prior research, we propose policy actions that align with the EU legal landscape and institutional framework to address multilingual jailbreaks, while promoting linguistic inclusivity. These include mandatory assessments of multilingual capabilities and vulnerabilities, public opinion research, and state support for multilingual AI development. The measures aim to improve AI safety and functionality through EU policy initiatives, guiding the implementation of the EU AI Act and informing regulatory efforts of the European AI Office.

Natalia Loukachevitch, Ekaterina Artemova, Tatiana Batura et al.

In this paper, we present NEREL, a Russian dataset for named entity recognition and relation extraction. NEREL is significantly larger than existing Russian datasets: to date it contains 56K annotated named entities and 39K annotated relations. Its important difference from previous datasets is annotation of nested named entities, as well as relations within nested entities and at the discourse level. NEREL can facilitate development of novel models that can extract relations between nested named entities, as well as relations on both sentence and document levels. NEREL also contains the annotation of events involving named entities and their roles in the events. The NEREL collection is available via https://github.com/nerel-ds/NEREL.

Joe Booth, Vladimir Ivanov

Over the course of the last several years there was a strong interest in application of modern optimal control techniques to the field of character animation. This interest was fueled by introduction of efficient learning based algorithms for policy optimization, growth in computation power, and game engine improvements. It was shown that it is possible to generate natural looking control of a character by using two ingredients. First, the simulated agent must adhere to a motion capture dataset. And second, the character aims to track the control input from the user. The paper aims at closing the gap between the researchers and users by introducing an open source implementation of physics based character control in Unity framework that has a low entry barrier and a steep learning curve.

Aidar Valeev, Roman Garaev, Vadim Lomshakov et al.

Large Language Models demonstrate the ability to solve various programming tasks, including code generation. Typically, the performance of LLMs is measured on benchmarks with small or medium-sized context windows of thousands of lines of code. At the same time, in real-world software projects, repositories can span up to millions of LoC. This paper closes this gap by contributing to the long context code generation benchmark (YABLoCo). The benchmark featured a test set of 215 functions selected from four large repositories with thousands of functions. The dataset contained metadata of functions, contexts of the functions with different levels of dependencies, docstrings, functions bodies, and call graphs for each repository. This paper presents three key aspects of the contribution. First, the benchmark aims at function body generation in large repositories in C and C++, two languages not covered by previous benchmarks. Second, the benchmark contains large repositories from 200K to 2,000K LoC. Third, we contribute a scalable evaluation pipeline for efficient computing of the target metrics and a tool for visual analysis of generated code. Overall, these three aspects allow for evaluating code generation in large repositories in C and C++.

Vladimir Ivanov, Andrey Sadovykh, Alexandr Naumchev et al.

Requirements identification in textual documents or extraction is a tedious and error prone task that many researchers suggest automating. We manually annotated the PURE dataset and thus created a new one containing both requirements and non-requirements. Using this dataset, we fine-tuned the BERT model and compare the results with several baselines such as fastText and ELMo. In order to evaluate the model on semantically more complex documents we compare the PURE dataset results with experiments on Request For Information (RFI) documents. The RFIs often include software requirements, but in a less standardized way. The fine-tuned BERT showed promising results on PURE dataset on the binary sentence classification task. Comparing with previous and recent studies dealing with constrained inputs, our approach demonstrates high performance in terms of precision and recall metrics, while being agnostic to the unstructured textual input.

Vladimir Ivanov, Anton Krivtsov, Sergey Strelkov et al.

This study considers modern surgical navigation systems based on augmented reality technologies. Augmented reality glasses are used to construct holograms of the patient's organs from MRI and CT data, subsequently transmitted to the glasses. This, in addition to seeing the actual patient, the surgeon gains visualization inside the patient's body (bones, soft tissues, blood vessels, etc.). The solutions developed at Peter the Great St. Petersburg Polytechnic University allow reducing the invasiveness of the procedure and preserving healthy tissues. This also improves the navigation process, making it easier to estimate the location and size of the tumor to be removed. We describe the application of developed systems to different types of surgical operations (removal of a malignant brain tumor, removal of a cyst of the cervical spine). We consider the specifics of novel navigation systems designed for anesthesia, for endoscopic operations. Furthermore, we discuss the construction of novel visualization systems for ultrasound machines. Our findings indicate that the technologies proposed show potential for telemedicine.

Vitaly Ivanin, Ekaterina Artemova, Tatiana Batura et al.

We show-case an application of information extraction methods, such as named entity recognition (NER) and relation extraction (RE) to a novel corpus, consisting of documents, issued by a state agency. The main challenges of this corpus are: 1) the annotation scheme differs greatly from the one used for the general domain corpora, and 2) the documents are written in a language other than English. Unlike expectations, the state-of-the-art transformer-based models show modest performance for both tasks, either when approached sequentially, or in an end-to-end fashion. Our experiments have demonstrated that fine-tuning on a large unlabeled corpora does not automatically yield significant improvement and thus we may conclude that more sophisticated strategies of leveraging unlabelled texts are demanded. In this paper, we describe the whole developed pipeline, starting from text annotation, baseline development, and designing a shared task in hopes of improving the baseline. Eventually, we realize that the current NER and RE technologies are far from being mature and do not overcome so far challenges like ours.

Dmitry Grigorev, Vladimir Ivanov

The paper presents the solution of team "Inno" to a SEMEVAL 2020 task 11 "Detection of propaganda techniques in news articles". The goal of the second subtask is to classify textual segments that correspond to one of the 18 given propaganda techniques in news articles dataset. We tested a pure Transformer-based model with an optimized learning scheme on the ability to distinguish propaganda techniques between each other. Our model showed 0.6 and 0.58 overall F1 score on validation set and test set accordingly and non-zero F1 score on each class on both sets.

Ekaterina Artemova, Tatiana Batura, Anna Golenkovskaya et al.

In this paper we present a corpus of Russian strategic planning documents, RuREBus. This project is grounded both from language technology and e-government perspectives. Not only new language sources and tools are being developed, but also their applications to e-goverment research. We demonstrate the pipeline for creating a text corpus from scratch. First, the annotation schema is designed. Next texts are marked up using human-in-the-loop strategy, so that preliminary annotations are derived from a machine learning model and are manually corrected. The amount of annotated texts is large enough to showcase what insights can be gained from RuREBus.

Chen Wang, Chengyuan Deng, Vladimir Ivanov

Variational Autoencoders (VAEs) are powerful in data representation inference, but it cannot learn relations between features with its vanilla form and common variations. The ability to capture relations within data can provide the much needed inductive bias necessary for building more robust Machine Learning algorithms with more interpretable results. In this paper, inspired by recent advances in relational learning using Graph Neural Networks, we propose the Self-Attention Graph Variational AutoEncoder (SAG-VAE) network which can simultaneously learn feature relations and data representations in an end-to-end manner. SAG-VAE is trained by jointly inferring the posterior distribution of two types of latent variables, which denote the data representation and a shared graph structure, respectively. Furthermore, we introduce a novel self-attention graph network that improves the generative capabilities of SAG-VAE by parameterizing the generative distribution allowing SAG-VAE to generate new data via graph convolution, while still trainable via backpropagation. A learnable relational graph representation enhances SAG-VAE's robustness to perturbation and noise, while also providing deeper intuition into model performance. Experiments based on graphs show that SAG-VAE is capable of approximately retrieving edges and links between nodes based entirely on feature observations. Finally, results on image data illustrate that SAG-VAE is fairly robust against perturbations in image reconstruction and sampling.

Keith Bonawitz, Hubert Eichner, Wolfgang Grieskamp et al.

Federated Learning is a distributed machine learning approach which enables model training on a large corpus of decentralized data. We have built a scalable production system for Federated Learning in the domain of mobile devices, based on TensorFlow. In this paper, we describe the resulting high-level design, sketch some of the challenges and their solutions, and touch upon the open problems and future directions.

Joseph Alexander Brown, Vladimir Ivanov, Alan Rogers et al.

The software is often produced under significant time constraints. Our idea is to understand the effects of various software development practices on the performance of developers working in stressful environments, and identify the best operating conditions for software developed under stressful conditions collecting data through questionnaires, non-invasive software measurement tools that can collect measurable data about software engineers and the software they develop, without intervening their activities, and biophysical sensors and then try to recreated also in different processes or key development practices such conditions.

Stanislav Litvinov, Marat Mingazov, Vladislav Myachikov et al.

Software engineering education and training have obstacles caused by a lack of basic knowledge about a process of program execution. The article is devoted to the development of special tools that help to visualize the process. We analyze existing tools and propose a new approach to stack and heap visualization. The solution is able to overcome major drawbacks of existing tools and suites well for analysis of programs written in Java and C/C++.

Vasilii Artemev, Vladimir Ivanov, Manuel Mazzara et al.

Analysis of data related to software development helps to increase quality, control and predictability of software development processes and products.However, collecting such data for is a complex task. A non-invasive collection of software metrics is one of the most promising approaches to solve the task. In this paper we present an approach which consists of four parts: collect the data, store all collected data, unify the stored data and analyze the data to provide insights to the user about software product or process. We employ the approach to the development of an architecture for non-invasive software measurement system and explain its advantages and limitations.

Keith Bonawitz, Vladimir Ivanov, Ben Kreuter et al.

Secure Aggregation protocols allow a collection of mutually distrust parties, each holding a private value, to collaboratively compute the sum of those values without revealing the values themselves. We consider training a deep neural network in the Federated Learning model, using distributed stochastic gradient descent across user-held training data on mobile devices, wherein Secure Aggregation protects each user's model gradient. We design a novel, communication-efficient Secure Aggregation protocol for high-dimensional data that tolerates up to 1/3 users failing to complete the protocol. For 16-bit input values, our protocol offers 1.73x communication expansion for $2^{10}$ users and $2^{20}$-dimensional vectors, and 1.98x expansion for $2^{14}$ users and $2^{24}$ dimensional vectors.