Traian Rebedea

Aakshita Chandiramani, Aaron Blakeman, Abdullahi Olaoye et al. · amazon-science, cmu

We describe the pre-training, post-training, and quantization of Nemotron 3 Super, a 120 billion (active 12 billion) parameter hybrid Mamba-Attention Mixture-of-Experts model. Nemotron 3 Super is the first model in the Nemotron 3 family to 1) be pre-trained in NVFP4, 2) leverage LatentMoE, a new Mixture-of-Experts architecture that optimizes for both accuracy per FLOP and accuracy per parameter, and 3) include MTP layers for inference acceleration through native speculative decoding. We pre-trained Nemotron 3 Super on 25 trillion tokens followed by post-training using supervised fine tuning (SFT) and reinforcement learning (RL). The final model supports up to 1M context length and achieves comparable accuracy on common benchmarks, while also achieving up to 2.2x and 7.5x higher inference throughput compared to GPT-OSS-120B and Qwen3.5-122B, respectively. Nemotron 3 Super datasets, along with the base, post-trained, and quantized checkpoints, are open-sourced on HuggingFace.

Traian Rebedea, Razvan Dinu, Makesh Sreedhar et al.

NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems. Guardrails (or rails for short) are a specific way of controlling the output of an LLM, such as not talking about topics considered harmful, following a predefined dialogue path, using a particular language style, and more. There are several mechanisms that allow LLM providers and developers to add guardrails that are embedded into a specific model at training, e.g. using model alignment. Differently, using a runtime inspired from dialogue management, NeMo Guardrails allows developers to add programmable rails to LLM applications - these are user-defined, independent of the underlying LLM, and interpretable. Our initial results show that the proposed approach can be used with several LLM providers to develop controllable and safe LLM applications using programmable rails.

Yoshinari Fujinuma, Varun Gangal, Traian Rebedea et al.

Large language model (LLM) agents increasingly rely on reusable skills i.e. documents describing task-specific procedures. However, this introduces a new attack surface for agents to manage. We study two complementary directions for this threat. First, we evaluate guardian-based defenses: an intermediary LLM agent that acts as a mediator for skill file access (dynamic guardian) or pre-rewrites these files at build time (static guardian). Across three LLM agent families, our guardians cut attack success rate (ASR) by well over half while preserving task utility. Second, we stress test them through attack reframing using four attacks that preserve the malicious instruction but change the phrasing. For non-guardian setup, the reframing pushes the ASR up to 81.4\%, but the dynamic guardian brings it down to 18.6\%, showing that real-time mediation is a robust defense.

Mihai Masala, Marius Leordeanu, Mihai Dascalu et al.

Vision-Language Models (VLMs) largely follow the text-only LLM trajectory, excelling on English benchmarks but sharply degrading on low-resource languages, where neither large-scale image-text corpora nor culturally grounded evaluations exist. We present a systematic study of building a language-specific VLM for Romanian, covering the full pipeline from data construction to architectural choices. We translate established English VLM training and evaluation corpora into Romanian, applying machine translation to textual annotations and to in-image text, preserving visual grounding while adapting the textual content. Using this data, we train and ablate a series of VLMs to isolate the contribution of (i) vision backbones of varying scale and pretraining, (ii) language backbones from multilingual to Romanian-adapted LLMs, and (iii) OCR-style image-text data. We further curate HoraVQA, a culturally native evaluation set grounded in Romanian everyday scenes. Romanian-adapted VLMs consistently outperform their same-sized counterparts and, across all evaluated benchmarks, even surpass models from the next larger size category.

Mihai Masala, Nicolae Cudlenco, Traian Rebedea et al.

Artificial Intelligence makes great advances today and starts to bridge the gap between vision and language. However, we are still far from understanding, explaining and controlling explicitly the visual content from a linguistic perspective, because we still lack a common explainable representation between the two domains. In this work we come to address this limitation and propose the Graph of Events in Space and Time (GEST), by which we can represent, create and explain, both visual and linguistic stories. We provide a theoretical justification of our model and an experimental validation, which proves that GEST can bring a solid complementary value along powerful deep learning models. In particular, GEST can help improve at the content-level the generation of videos from text, by being easily incorporated into our novel video generation engine. Additionally, by using efficient graph matching techniques, the GEST graphs can also improve the comparisons between texts at the semantic level.

Andrei-Alexandru Preda, Dumitru-Clementin Cercel, Traian Rebedea et al.

This paper describes the solutions submitted by the UPB team to the AuTexTification shared task, featured as part of IberLEF-2023. Our team participated in the first subtask, identifying text documents produced by large language models instead of humans. The organizers provided a bilingual dataset for this subtask, comprising English and Spanish texts covering multiple domains, such as legal texts, social media posts, and how-to articles. We experimented mostly with deep learning models based on Transformers, as well as training techniques such as multi-task learning and virtual adversarial training to obtain better results. We submitted three runs, two of which consisted of ensemble models. Our best-performing model achieved macro F1-scores of 66.63% on the English dataset and 67.10% on the Spanish dataset.

Shaona Ghosh, Prasoon Varshney, Makesh Narsimhan Sreedhar et al. · amazon-science

As Large Language Models (LLMs) and generative AI become increasingly widespread, concerns about content safety have grown in parallel. Currently, there is a clear lack of high-quality, human-annotated datasets that address the full spectrum of LLM-related safety risks and are usable for commercial applications. To bridge this gap, we propose a comprehensive and adaptable taxonomy for categorizing safety risks, structured into 12 top-level hazard categories with an extension to 9 fine-grained subcategories. This taxonomy is designed to meet the diverse requirements of downstream users, offering more granular and flexible tools for managing various risk types. Using a hybrid data generation pipeline that combines human annotations with a multi-LLM "jury" system to assess the safety of responses, we obtain Aegis 2.0, a carefully curated collection of 34,248 samples of human-LLM interactions, annotated according to our proposed taxonomy. To validate its effectiveness, we demonstrate that several lightweight models, trained using parameter-efficient techniques on Aegis 2.0, achieve performance competitive with leading safety models fully fine-tuned on much larger, non-commercial datasets. In addition, we introduce a novel training blend that combines safety with topic following data.This approach enhances the adaptability of guard models, enabling them to generalize to new risk categories defined during inference. We plan to open-source Aegis 2.0 data and models to the research community to aid in the safety guardrailing of LLMs.

Rares Folea, Radu Iacob, Emil Slusanschi et al.

This paper presents a generic method for transforming the source code of various algorithms to numerical embeddings, by dynamically analysing the behaviour of computer programs against different inputs and by tailoring multiple generic complexity functions for the analysed metrics. The used algorithms embeddings are based on r-Complexity . Using the proposed code embeddings, we present an implementation of the XGBoost algorithm that achieves an average F1-score on a multi-label dataset with 11 classes, built using real-world code snippets submitted for programming competitions on the Codeforces platform.

Eduard Stefan Dinuta, Iustin Sirbu, Traian Rebedea

Safety for Large Language Models (LLMs) has been an ongoing research focus since their emergence and is even more relevant nowadays with the increasing capacity of those models. Currently, there are several guardrails in place for all public LLMs and multiple proposed datasets for training safety classifiers. However, training these safety classifiers relies on large quantities of labeled data, which can be problematic to acquire, prone to labeling errors, or often include synthetic data. To address these issues, we suggest a different approach: utilizing semi-supervised learning techniques, which leverage both labeled and unlabeled data, to improve the performance on the safety task. We analyze the improvements that these techniques can offer for both prompts given to Large Language Models and the responses to those requests. Moreover, since augmentation is the central part of semi-supervised algorithms, we demonstrate the importance of using task-specific augmentations, which significantly increase the performance when compared to general-purpose augmentation techniques.

Prasoon Varshney, Makesh Narsimhan Sreedhar, Liwei Jiang et al.

Large language models (LLMs) are typically aligned to a universal set of safety and usage principles intended for broad public acceptability. Yet, real-world applications of LLMs often take place within organizational ecosystems shaped by distinctive corporate policies, regulatory requirements, use cases, brand guidelines, and ethical commitments. This reality highlights the need for rigorous and comprehensive evaluation of LLMs with pluralistic alignment goals, an alignment paradigm that emphasizes adaptability to diverse user values and needs. In this work, we present PLURALISTIC BEHAVIOR SUITE (PBSUITE), a dynamic evaluation suite designed to systematically assess LLMs' capacity to adhere to pluralistic alignment specifications in multi-turn, interactive conversations. PBSUITE consists of (1) a diverse dataset of 300 realistic LLM behavioral policies, grounded in 30 industries; and (2) a dynamic evaluation framework for stress-testing model compliance with custom behavioral specifications under adversarial conditions. Using PBSUITE, We find that leading open- and closed-source LLMs maintain robust adherence to behavioral policies in single-turn settings (less than 4% failure rates), but their compliance weakens substantially in multi-turn adversarial interactions (up to 84% failure rates). These findings highlight that existing model alignment and safety moderation methods fall short in coherently enforcing pluralistic behavioral policies in real-world LLM interactions. Our work contributes both the dataset and analytical framework to support future research toward robust and context-aware pluralistic alignment techniques.

Vlad Negoita, Mihai Masala, Traian Rebedea

Large Language Models (LLMs) have recently exploded in popularity, often matching or outperforming human abilities on many tasks. One of the key factors in training LLMs is the availability and curation of high-quality data. Data quality is especially crucial for under-represented languages, where high-quality corpora are scarce. In this work we study the characteristics and coverage of Romanian pretraining corpora and we examine how they differ from English data. By training a lightweight multitask model on carefully LLM-annotated Romanian texts, we are able to analyze and perform multi-level filtering (e.g., educational value, topic, format) to generate high-quality pretraining datasets. Our experiments show noteworthy trends in the topics present in Romanian and English data, while also proving the effectiveness of filtering data through improved LLM pretraining performance across multiple benchmarks.

Iustin Sirbu, Robert-Adrian Popovici, Cornelia Caragea et al.

We introduce MultiMatch, a novel semi-supervised learning (SSL) algorithm combining the paradigms of co-training and consistency regularization with pseudo-labeling. At its core, MultiMatch features a pseudo-label weighting module designed for selecting and filtering pseudo-labels based on head agreement and model confidence, and weighting them according to the perceived classification difficulty. This novel module enhances and unifies three existing techniques -- heads agreement from Multihead Co-training, self-adaptive thresholds from FreeMatch, and Average Pseudo-Margins from MarginMatch -- resulting in a holistic approach that improves robustness and performance in SSL settings. Experimental results on benchmark datasets highlight the superior performance of MultiMatch, i.e., MultiMatch achieves state-of-the-art results on 8 out of 10 setups from 5 natural language processing datasets and ranks first according to the Friedman test among 21 methods. Furthermore, MultiMatch demonstrates exceptional robustness in highly imbalanced settings, outperforming the second-best approach by 3.26%, a critical advantage for real-world text classification tasks. Our code is available on GitHub.

Mihai Masala, Denis C. Ilie-Ablachim, Alexandru Dima et al.

In recent years, Large Language Models (LLMs) have achieved almost human-like performance on various tasks. While some LLMs have been trained on multilingual data, most of the training data is in English; hence, their performance in English greatly exceeds other languages. To our knowledge, we are the first to collect and translate a large collection of texts, instructions, and benchmarks and train, evaluate, and release open-source LLMs tailored for Romanian. We evaluate our methods on four different categories, including academic benchmarks, MT-Bench (manually translated), and a professionally built historical, cultural, and social benchmark adapted to Romanian. We argue for the usefulness and high performance of RoLLMs by obtaining state-of-the-art results across the board. We publicly release all resources (i.e., data, training and evaluation code, models) to support and encourage research on Romanian LLMs while concurrently creating a generalizable recipe, adequate for other low or less-resourced languages.

Ion George Dinu, Marian Cristian Mihăescu, Traian Rebedea

Architectural code smells erode software maintainability and are costly to repair manually, yet unlike localized bugs, they require cross-module reasoning about design intent that challenges both developers and automated tools. While large language model agents excel at bug fixing and code-level refactoring, their ability to repair architectural code smells remains unexplored. We present the first empirical evaluation of LLM agents on architectural code smell repair. We contribute SmellBench, a task orchestration framework that incorporates smell-type-specific optimized prompts and supports iterative multi-step execution, together with a scoring methodology that separately evaluates repair effectiveness, false positive identification, and net codebase impact. We evaluate 11 agent configurations from four model families (GPT, Claude, Gemini, Mistral) on 65 hard-severity architectural smells detected by PyExamine in the Python project scikit-learn, validated against expert judgments. Expert validation reveals that 63.1% of detected smells are false positives, while the best agent achieves a 47.7% resolution rate. Agents identify false positives with up to $κ= 0.94$ expert agreement, but repair aggressiveness and net codebase quality are inversely related: the most aggressive agent introduces 140 new smells. These findings expose a gap between current LLM capabilities in localized code transformations and the architectural understanding needed for cross-module refactoring. SmellBench provides reusable infrastructure for tracking progress on this underexplored dimension of automated software engineering. We release our code and data at https://doi.org/10.5281/zenodo.19247588.

Makesh Narsimhan Sreedhar, Traian Rebedea, Shaona Ghosh et al.

Recent advancements in instruction-tuning datasets have predominantly focused on specific tasks like mathematical or logical reasoning. There has been a notable gap in data designed for aligning language models to maintain topic relevance in conversations - a critical aspect for deploying chatbots to production. We introduce the CantTalkAboutThis dataset to help language models remain focused on the subject at hand during task-oriented interactions. It consists of synthetic dialogues on a wide range of conversation topics from different domains. These dialogues are interspersed with distractor turns that intentionally divert the chatbot from the predefined topic. Fine-tuning language models on this dataset helps make them resilient to deviating from the role assigned and improves their ability to maintain topical coherence compared to general-purpose instruction-tuned LLMs like GPT-4-turbo and Mixtral-Instruct. Additionally, preliminary observations suggest that training models on this dataset also enhance their performance on fine-grained instruction following tasks, including safety alignment.

Aishwarya Padmakumar, Leon Derczynski, Traian Rebedea et al.

Automated methods for red teaming LLMs are an important tool to identify LLM vulnerabilities that may not be covered in static benchmarks, allowing for more thorough probing. They can also adapt to each specific LLM to discover weaknesses unique to it. Most current automated red teaming methods are intended for tackling safety and content moderation. Thus, they make use of content safety models as evaluators and optimize for circumventing them, and as such, have not been tested with other adversarial intents not typically captured by these. We propose a pipeline for training a red teaming model that can generalize to arbitrary adversarial goals, including objectives it has not been directly trained on, and that does not depend on the existence of a pre-existing evaluator available at training time. We demonstrate that finetuning small models, such as Qwen3-8B, using this pipeline results in a substantial improvement in their ability to generate attacks for both in and out of domain adversarial goals.

Mihai Masala, Denis C. Ilie-Ablachim, Dragos Corlatescu et al.

In recent years, Large Language Models (LLMs) have achieved almost human-like performance on various tasks. While some LLMs have been trained on multilingual data, most of the training data is in English. Hence, their performance in English greatly exceeds their performance in other languages. This document presents our approach to training and evaluating the first foundational and chat LLM specialized for Romanian.

Iustin Sîrbu, Iulia-Renata Sîrbu, Jasmina Bogojeska et al.

Medical imaging is crucial for diagnosing, monitoring, and treating medical conditions. The medical reports of radiology images are the primary medium through which medical professionals attest their findings, but their writing is time consuming and requires specialized clinical expertise. The automated generation of radiography reports has thus the potential to improve and standardize patient care and significantly reduce clinicians workload. Through our work, we have designed and evaluated an end-to-end transformer-based method to generate accurate and factually complete radiology reports for X-ray images. Additionally, we are the first to introduce curriculum learning for end-to-end transformers in medical imaging and demonstrate its impact in obtaining improved performance. The experiments have been conducted using the MIMIC-CXR-JPG database, the largest available chest X-ray dataset. The results obtained are comparable with the current state-of-the-art on the natural language generation (NLG) metrics BLEU and ROUGE-L, while setting new state-of-the-art results on F1 examples-averaged, F1-macro and F1-micro metrics for clinical accuracy and on the METEOR metric widely used for NLG.

Mihai Masala, Traian Rebedea, Horia Velicu

In recent years,the entire field of Natural Language Processing (NLP) has enjoyed amazing novel results achieving almost human-like performance on a variety of tasks. Legal NLP domain has also been part of this process, as it has seen an impressive growth. However, general-purpose models are not readily applicable for legal domain. Due to the nature of the domain (e.g. specialized vocabulary, long documents) specific models and methods are often needed for Legal NLP. In this work we investigate both specialized and general models for predicting the final ruling of a legal case, task known as Legal Judgment Prediction (LJP). We particularly focus on methods to extend to sequence length of Transformer-based models to better understand the long documents present in legal corpora. Extensive experiments on 4 LJP datasets in Romanian, originating from 2 sources with significantly different sizes and document lengths, show that specialized models and handling long texts are critical for a good performance.

Ionel-Alexandru Hosu, Traian Rebedea, Razvan Pascanu · deepmind

One proposed mechanism to improve exploration in reinforcement learning is through the use of macro-actions. Paradoxically though, in many scenarios the naive addition of macro-actions does not lead to better exploration, but rather the opposite. It has been argued that this was caused by adding non-useful macros and multiple works have focused on mechanisms to discover effectively environment-specific useful macros. In this work, we take a slightly different perspective. We argue that the difficulty stems from the trade-offs between reducing the average number of decisions per episode versus increasing the size of the action space. Namely, one typically treats each potential macro-action as independent and atomic, hence strictly increasing the search space and making typical exploration strategies inefficient. To address this problem we propose a novel regularization term that exploits the relationship between actions and macro-actions to improve the credit assignment mechanism by reducing the effective dimension of the action space and, therefore, improving exploration. The term relies on a similarity matrix that is meta-learned jointly with learning the desired policy. We empirically validate our strategy looking at macro-actions in Atari games, and the StreetFighter II environment. Our results show significant improvements over the Rainbow-DQN baseline in all environments. Additionally, we show that the macro-action similarity is transferable to related environments. We believe this work is a small but important step towards understanding how the similarity-imposed geometry on the action space can be exploited to improve credit assignment and exploration, therefore making learning more effective.

Makesh Narsimhan Sreedhar, Traian Rebedea, Christopher Parisien

Dialogue policies play a crucial role in developing task-oriented dialogue systems, yet their development and maintenance are challenging and typically require substantial effort from experts in dialogue modeling. While in many situations, large amounts of conversational data are available for the task at hand, people lack an effective solution able to extract dialogue policies from this data. In this paper, we address this gap by first illustrating how Large Language Models (LLMs) can be instrumental in extracting dialogue policies from datasets, through the conversion of conversations into a unified intermediate representation consisting of canonical forms. We then propose a novel method for generating dialogue policies utilizing a controllable and interpretable graph-based methodology. By combining canonical forms across conversations into a flow network, we find that running graph traversal algorithms helps in extracting dialogue flows. These flows are a better representation of the underlying interactions than flows extracted by prompting LLMs. Our technique focuses on giving conversation designers greater control, offering a productivity tool to improve the process of developing dialogue policies.

Mihai Masala, Nicolae Cudlenco, Traian Rebedea et al.

One of the essential human skills is the ability to seamlessly build an inner representation of the world. By exploiting this representation, humans are capable of easily finding consensus between visual, auditory and linguistic perspectives. In this work, we set out to understand and emulate this ability through an explicit representation for both vision and language - Graphs of Events in Space and Time (GEST). GEST alows us to measure the similarity between texts and videos in a semantic and fully explainable way, through graph matching. It also allows us to generate text and videos from a common representation that provides a well understood content. In this work we show that the graph matching similarity metrics based on GEST outperform classical text generation metrics and can also boost the performance of state of art, heavily trained metrics.

Andrei-Marius Avram, Darius Catrina, Dumitru-Clementin Cercel et al.

Running large-scale pre-trained language models in computationally constrained environments remains a challenging problem yet to be addressed, while transfer learning from these models has become prevalent in Natural Language Processing tasks. Several solutions, including knowledge distillation, network quantization, or network pruning have been previously proposed; however, these approaches focus mostly on the English language, thus widening the gap when considering low-resource languages. In this work, we introduce three light and fast versions of distilled BERT models for the Romanian language: Distil-BERT-base-ro, Distil-RoBERT-base, and DistilMulti-BERT-base-ro. The first two models resulted from the individual distillation of knowledge from two base versions of Romanian BERTs available in literature, while the last one was obtained by distilling their ensemble. To our knowledge, this is the first attempt to create publicly available Romanian distilled BERT models, which were thoroughly evaluated on five tasks: part-of-speech tagging, named entity recognition, sentiment analysis, semantic textual similarity, and dialect identification. Our experimental results argue that the three distilled models offer performance comparable to their teachers, while being twice as fast on a GPU and ~35% smaller. In addition, we further test the similarity between the predictions of our students versus their teachers by measuring their label and probability loyalty, together with regression loyalty - a new metric introduced in this work.

Nicolae-Teodor Pavel, Traian Rebedea

Commit messages have an important impact in software development, especially when working in large teams. Multiple developers who have a different style of writing may often be involved in the same project. For this reason, it may be difficult to maintain a strict pattern of writing informative commit messages, with the most frequent issue being that these messages are not descriptive enough. In this paper we apply neural machine translation (NMT) techniques to convert code diffs into commit messages and we present an improved sketch-based encoder for this task. We split the approach into three parts. Firstly, we focus on finding a more suitable NMT baseline for this problem. Secondly, we show that the performance of the NMT models can be improved by training on examples containing a specific file type. Lastly, we introduce a novel sketch-based neural model inspired by recent approaches used for code generation and we show that the sketch-based encoder significantly outperforms existing state of the art solutions. The results highlight that this improvement is relevant especially for Java source code files, by examining two different datasets introduced in recent years for this task.

George-Eduard Zaharia, George-Alexandru Vlad, Dumitru-Clementin Cercel et al.

Sentiment analysis is a process widely used in opinion mining campaigns conducted today. This phenomenon presents applications in a variety of fields, especially in collecting information related to the attitude or satisfaction of users concerning a particular subject. However, the task of managing such a process becomes noticeably more difficult when it is applied in cultures that tend to combine two languages in order to express ideas and thoughts. By interleaving words from two languages, the user can express with ease, but at the cost of making the text far less intelligible for those who are not familiar with this technique, but also for standard opinion mining algorithms. In this paper, we describe the systems developed by our team for SemEval-2020 Task 9 that aims to cover two well-known code-mixed languages: Hindi-English and Spanish-English. We intend to solve this issue by introducing a solution that takes advantage of several neural network approaches, as well as pre-trained word embeddings. Our approach (multlingual BERT) achieves promising performance on the Hindi-English task, with an average F1-score of 0.6850, registered on the competition leaderboard, ranking our team 16th out of 62 participants. For the Spanish-English task, we obtained an average F1-score of 0.7064 ranking our team 17th out of 29 participants by using another multilingual Transformer-based model, XLM-RoBERTa.

Radu D. Stochiţoiu, Marian Petrica, Traian Rebedea et al.

Analysing and understanding the transmission and evolution of the COVID-19 pandemic is mandatory to be able to design the best social and medical policies, foresee their outcomes and deal with all the subsequent socio-economic effects. We address this important problem from a computational and machine learning perspective. More specifically, we want to statistically estimate all the relevant parameters for the new coronavirus COVID-19, such as the reproduction number, fatality rate or length of infectiousness period, based on Romanian patients, as well as be able to predict future outcomes. This endeavor is important, since it is well known that these factors vary across the globe, and might be dependent on many causes, including social, medical, age and genetic factors. We use a recently published improved version of SEIR, which is the classic, established model for infectious diseases. We want to infer all the parameters of the model, which govern the evolution of the pandemic in Romania, based on the only reliable, true measurement, which is the number of deaths. Once the model parameters are estimated, we are able to predict all the other relevant measures, such as the number of exposed and infectious people. To this end, we propose a self-supervised approach to train a deep convolutional network to guess the correct set of Modified-SEIR model parameters, given the observed number of daily fatalities. Then, we refine the solution with a stochastic coordinate descent approach. We compare our deep learning optimization scheme with the classic grid search approach and show great improvement in both computational time and prediction accuracy. We find an optimistic result in the case fatality rate for Romania which may be around 0.3% and we also demonstrate that our model is able to correctly predict the number of daily fatalities for up to three weeks in the future.

George-Sebastian Pîrtoacă, Traian Rebedea, Stefan Ruseti

Answering multiple-choice questions in a setting in which no supporting documents are explicitly provided continues to stand as a core problem in natural language processing. The contribution of this article is two-fold. First, it describes a method which can be used to semantically rank documents extracted from Wikipedia or similar natural language corpora. Second, we propose a model employing the semantic ranking that holds the first place in two of the most popular leaderboards for answering multiple-choice questions: ARC Easy and Challenge. To achieve this, we introduce a self-attention based neural network that latently learns to rank documents by their importance related to a given question, whilst optimizing the objective of predicting the correct answer. These documents are considered relevant contexts for the underlying question. We have published the ranked documents so that they can be used off-the-shelf to improve downstream decision models.

George-Sebastian Pirtoaca, Traian Rebedea, Stefan Ruseti

Question answering is one of the most important and difficult applications at the border of information retrieval and natural language processing, especially when we talk about complex science questions which require some form of inference to determine the correct answer. In this paper, we present a two-step method that combines information retrieval techniques optimized for question answering with deep learning models for natural language inference in order to tackle the multi-choice question answering in the science domain. For each question-answer pair, we use standard retrieval-based models to find relevant candidate contexts and decompose the main problem into two different sub-problems. First, assign correctness scores for each candidate answer based on the context using retrieval models from Lucene. Second, we use deep learning architectures to compute if a candidate answer can be inferred from some well-chosen context consisting of sentences retrieved from the knowledge base. In the end, all these solvers are combined using a simple neural network to predict the correct answer. This proposed two-step model outperforms the best retrieval-based solver by over 3% in absolute accuracy.

Florin Brad, Radu Iacob, Ionel Hosu et al.

Progress in natural language interfaces to databases (NLIDB) has been slow mainly due to linguistic issues (such as language ambiguity) and domain portability. Moreover, the lack of a large corpus to be used as a standard benchmark has made data-driven approaches difficult to develop and compare. In this paper, we revisit the problem of NLIDBs and recast it as a sequence translation problem. To this end, we introduce a large dataset extracted from the Stack Exchange Data Explorer website, which can be used for training neural natural language interfaces for databases. We also report encouraging baseline results on a smaller manually annotated test corpus, obtained using an attention-based sequence-to-sequence neural network.

Ionel-Alexandru Hosu, Traian Rebedea

This paper introduces a novel method for learning how to play the most difficult Atari 2600 games from the Arcade Learning Environment using deep reinforcement learning. The proposed method, human checkpoint replay, consists in using checkpoints sampled from human gameplay as starting points for the learning process. This is meant to compensate for the difficulties of current exploration strategies, such as epsilon-greedy, to find successful control policies in games with sparse rewards. Like other deep reinforcement learning architectures, our model uses a convolutional neural network that receives only raw pixel inputs to estimate the state value function. We tested our method on Montezuma's Revenge and Private Eye, two of the most challenging games from the Atari platform. The results we obtained show a substantial improvement compared to previous learning approaches, as well as over a random player. We also propose a method for training deep reinforcement learning agents using human gameplay experience, which we call human experience replay.