Eric Nyberg

Leonid Boytsov, Preksha Patel, Vivek Sourabh et al. · amazon-science

We carried out a reproducibility study of InPars, which is a method for unsupervised training of neural rankers (Bonifacio et al., 2022). As a by-product, we developed InPars-light, which is a simple-yet-effective modification of InPars. Unlike InPars, InPars-light uses 7x-100x smaller ranking models and only a freely available language model BLOOM, which -- as we found out -- produced more accurate rankers compared to a proprietary GPT-3 model. On all five English retrieval collections (used in the original InPars study) we obtained substantial (7%-30%) and statistically significant improvements over BM25 (in nDCG and MRR) using only a 30M parameter six-layer MiniLM-30M ranker and a single three-shot prompt. In contrast, in the InPars study only a 100x larger monoT5-3B model consistently outperformed BM25, whereas their smaller monoT5-220M model (which is still 7x larger than our MiniLM ranker) outperformed BM25 only on MS MARCO and TREC DL 2020. In the same three-shot prompting scenario, our 435M parameter DeBERTA v3 ranker was at par with the 7x larger monoT5-3B (average gain over BM25 of 1.3 vs 1.32): In fact, on three out of five datasets, DeBERTA slightly outperformed monoT5-3B. Finally, these good results were achieved by re-ranking only 100 candidate documents compared to 1000 used by Bonifacio et al. (2022). We believe that InPars-light is the first truly cost-effective prompt-based unsupervised recipe to train and deploy neural ranking models that outperform BM25. Our code and data is publicly available. https://github.com/searchivarius/inpars_light/

Kaixin Ma, Hao Cheng, Xiaodong Liu et al. · microsoft-research

We propose a novel open-domain question answering (ODQA) framework for answering single/multi-hop questions across heterogeneous knowledge sources. The key novelty of our method is the introduction of the intermediary modules into the current retriever-reader pipeline. Unlike previous methods that solely rely on the retriever for gathering all evidence in isolation, our intermediary performs a chain of reasoning over the retrieved set. Specifically, our method links the retrieved evidence with its related global context into graphs and organizes them into a candidate list of evidence chains. Built upon pretrained language models, our system achieves competitive performance on two ODQA datasets, OTT-QA and NQ, against tables and passages from Wikipedia. In particular, our model substantially outperforms the previous state-of-the-art on OTT-QA with an exact match score of 47.3 (45 % relative gain).

Jonathan Francis, Bingqing Chen, Siddha Ganju et al. · cmu, nvidia

We present the results of our autonomous racing virtual challenge, based on the newly-released Learn-to-Race (L2R) simulation framework, which seeks to encourage interdisciplinary research in autonomous driving and to help advance the state of the art on a realistic benchmark. Analogous to racing being used to test cutting-edge vehicles, we envision autonomous racing to serve as a particularly challenging proving ground for autonomous agents as: (i) they need to make sub-second, safety-critical decisions in a complex, fast-changing environment; and (ii) both perception and control must be robust to distribution shifts, novel road features, and unseen obstacles. Thus, the main goal of the challenge is to evaluate the joint safety, performance, and generalisation capabilities of reinforcement learning agents on multi-modal perception, through a two-stage process. In the first stage of the challenge, we evaluate an autonomous agent's ability to drive as fast as possible, while adhering to safety constraints. In the second stage, we additionally require the agent to adapt to an unseen racetrack through safe exploration. In this paper, we describe the new L2R Task 2.0 benchmark, with refined metrics and baseline approaches. We also provide an overview of deployment, evaluation, and rankings for the inaugural instance of the L2R Autonomous Racing Virtual Challenge (supported by Carnegie Mellon University, Arrival Ltd., AICrowd, Amazon Web Services, and Honda Research), which officially used the new L2R Task 2.0 benchmark and received over 20,100 views, 437 active participants, 46 teams, and 733 model submissions -- from 88+ unique institutions, in 58+ different countries. Finally, we release leaderboard results from the challenge and provide description of the two top-ranking approaches in cross-domain model transfer, across multiple sensor configurations and simulated races.

Kaixin Ma, Filip Ilievski, Jonathan Francis et al. · cmu

Procedural text understanding is a challenging language reasoning task that requires models to track entity states across the development of a narrative. A complete procedural understanding solution should combine three core aspects: local and global views of the inputs, and global view of outputs. Prior methods considered a subset of these aspects, resulting in either low precision or low recall. In this paper, we propose Coalescing Global and Local Information (CGLI), a new model that builds entity- and timestep-aware input representations (local input) considering the whole context (global input), and we jointly model the entity states with a structured prediction objective (global output). Thus, CGLI simultaneously optimizes for both precision and recall. We extend CGLI with additional output layers and integrate it into a story reasoning framework. Extensive experiments on a popular procedural text understanding dataset show that our model achieves state-of-the-art results; experiments on a story reasoning benchmark show the positive impact of our model on downstream reasoning.

Gyan Tatiya, Jonathan Francis, Luca Bondi et al. · cmu

Generalisation to unseen contexts remains a challenge for embodied navigation agents. In the context of semantic audio-visual navigation (SAVi) tasks, the notion of generalisation should include both generalising to unseen indoor visual scenes as well as generalising to unheard sounding objects. However, previous SAVi task definitions do not include evaluation conditions on truly novel sounding objects, resorting instead to evaluating agents on unheard sound clips of known objects; meanwhile, previous SAVi methods do not include explicit mechanisms for incorporating domain knowledge about object and region semantics. These weaknesses limit the development and assessment of models' abilities to generalise their learned experience. In this work, we introduce the use of knowledge-driven scene priors in the semantic audio-visual embodied navigation task: we combine semantic information from our novel knowledge graph that encodes object-region relations, spatial knowledge from dual Graph Encoder Networks, and background knowledge from a series of pre-training tasks -- all within a reinforcement learning framework for audio-visual navigation. We also define a new audio-visual navigation sub-task, where agents are evaluated on novel sounding objects, as opposed to unheard clips of known objects. We show improvements over strong baselines in generalisation to unseen regions and novel sounding objects, within the Habitat-Matterport3D simulation environment, under the SoundSpaces task.

Jonathan Francis, Bingqing Chen, Weiran Yao et al. · cmu

The feasibility of collecting a large amount of expert demonstrations has inspired growing research interests in learning-to-drive settings, where models learn by imitating the driving behaviour from experts. However, exclusively relying on imitation can limit agents' generalisability to novel scenarios that are outside the support of the training data. In this paper, we address this challenge by factorising the driving task, based on the intuition that modular architectures are more generalisable and more robust to changes in the environment compared to monolithic, end-to-end frameworks. Specifically, we draw inspiration from the trajectory forecasting community and reformulate the learning-to-drive task as obstacle-aware perception and grounding, distribution-aware goal prediction, and model-based planning. Firstly, we train the obstacle-aware perception module to extract salient representation of the visual context. Then, we learn a multi-modal goal distribution by performing conditional density-estimation using normalising flow. Finally, we ground candidate trajectory predictions road geometry, and plan the actions based on on vehicle dynamics. Under the CARLA simulator, we report state-of-the-art results on the CARNOVEL benchmark.

Zhiruo Wang, Zhengbao Jiang, Eric Nyberg et al.

Tables are an important form of structured data for both human and machine readers alike, providing answers to questions that cannot, or cannot easily, be found in texts. Recent work has designed special models and training paradigms for table-related tasks such as table-based question answering and table retrieval. Though effective, they add complexity in both modeling and data acquisition compared to generic text solutions and obscure which elements are truly beneficial. In this work, we focus on the task of table retrieval, and ask: "is table-specific model design necessary for table retrieval, or can a simpler text-based model be effectively used to achieve a similar result?" First, we perform an analysis on a table-based portion of the Natural Questions dataset (NQ-table), and find that structure plays a negligible role in more than 70% of the cases. Based on this, we experiment with a general Dense Passage Retriever (DPR) based on text and a specialized Dense Table Retriever (DTR) that uses table-specific model designs. We find that DPR performs well without any table-specific design and training, and even achieves superior results compared to DTR when fine-tuned on properly linearized tables. We then experiment with three modules to explicitly encode table structures, namely auxiliary row/column embeddings, hard attention masks, and soft relation-based attention biases. However, none of these yielded significant improvements, suggesting that table-specific model design may not be necessary for table retrieval.

Leonid Boytsov, David Akinpelu, Nipun Katyal et al. · amazon-science

We tested over 20 Transformer models for ranking long documents (including recent LongP models trained with FlashAttention and RankGPT models "powered" by OpenAI and Anthropic cloud APIs). We compared them with the simple FirstP baseline, which applied the same model to truncated input (up to 512 tokens). On MS MARCO, TREC DL, and Robust04 no long-document model outperformed FirstP by more than 5% (on average). We hypothesized that this lack of improvement is not due to inherent model limitations, but due to benchmark positional bias (most relevant passages tend to occur early in documents), which is known to exist in MS MARCO. To confirm this, we analyzed positional relevance distributions across four long-document corpora (with six query sets) and observed the same early-position bias. Surprisingly, we also found bias in six BEIR collections, which are typically categorized as short-document datasets. We then introduced a new diagnostic dataset, MS MARCO FarRelevant, where relevant spans were deliberately placed beyond the first 512 tokens. On this dataset, many long-context models (including RankGPT) performed at random-baseline level, suggesting overfitting to positional bias. We also experimented with debiasing training data, but with limited success. Our findings (1) highlight the need for careful benchmark design in evaluating long-context models for document ranking, (2) identify model types that are more robust to positional bias, and (3) motivate further work on approaches to debias training data. We release our code and data to support further research.

Alex Gichamba, Tewodros Kederalah Idris, Brian Ebiyau et al. · cmu

Domain-specific question answering remains challenging for language models, given the deep technical knowledge required to answer questions correctly. This difficulty is amplified for smaller language models that cannot encode as much information in their parameters as larger models. The "Specializing Large Language Models for Telecom Networks" challenge aimed to enhance the performance of two small language models, Phi-2 and Falcon-7B in telecommunication question answering. In this paper, we present our question answering systems for this challenge. Our solutions achieved leading marks of 81.9% accuracy for Phi-2 and 57.3% for Falcon-7B. We have publicly released our code and fine-tuned models.

Hugo Rodrigues, Eric Nyberg, Luisa Coheur

Question Generation (QG) is a task of Natural Language Processing (NLP) that aims at automatically generating questions from text. Many applications can benefit from automatically generated questions, but often it is necessary to curate those questions, either by selecting or editing them. This task is informative on its own, but it is typically done post-generation, and, thus, the effort is wasted. In addition, most existing systems cannot incorporate this feedback back into them easily. In this work, we present a system, GEN, that learns from such (implicit) feedback. Following a pattern-based approach, it takes as input a small set of sentence/question pairs and creates patterns which are then applied to new unseen sentences. Each generated question, after being corrected by the user, is used as a new seed in the next iteration, so more patterns are created each time. We also take advantage of the corrections made by the user to score the patterns and therefore rank the generated questions. Results show that GEN is able to improve by learning from both levels of implicit feedback when compared to the version with no learning, considering the top 5, 10, and 20 questions. Improvements go up from 10%, depending on the metric and strategy used.

James Herman, Jonathan Francis, Siddha Ganju et al.

Existing research on autonomous driving primarily focuses on urban driving, which is insufficient for characterising the complex driving behaviour underlying high-speed racing. At the same time, existing racing simulation frameworks struggle in capturing realism, with respect to visual rendering, vehicular dynamics, and task objectives, inhibiting the transfer of learning agents to real-world contexts. We introduce a new environment, where agents Learn-to-Race (L2R) in simulated competition-style racing, using multimodal information--from virtual cameras to a comprehensive array of inertial measurement sensors. Our environment, which includes a simulator and an interfacing training framework, accurately models vehicle dynamics and racing conditions. In this paper, we release the Arrival simulator for autonomous racing. Next, we propose the L2R task with challenging metrics, inspired by learning-to-drive challenges, Formula-style racing, and multimodal trajectory prediction for autonomous driving. Additionally, we provide the L2R framework suite, facilitating simulated racing on high-precision models of real-world tracks. Finally, we provide an official L2R task dataset of expert demonstrations, as well as a series of baseline experiments and reference implementations. We make all code available: https://github.com/learn-to-race/l2r.

Syeda Nahida Akter, Shrimai Prabhumoye, Matvei Novikov et al. · stanford

Large Language Models (LLMs) have shown strong reasoning capabilities, particularly when enhanced through Reinforcement Learning (RL). While prior work has successfully applied RL to mathematical reasoning -- where rules and correctness are well-defined -- generalizing these methods to broader reasoning domains remains challenging due to limited data, the lack of verifiable reward structures, and diverse task requirements. In this work, we propose NEMOTRON-CROSSTHINK, a framework that systematically incorporates multi-domain corpora, including both synthetic and real-world question-answer pairs, into RL training to improve generalization across diverse reasoning tasks. NEMOTRON-CROSSTHINK addresses key challenges by (1) incorporating data from varied sources spanning STEM, humanities, social sciences, etc.; (2) applying structured templates (e.g., multiple-choice and open-ended) to control answer-space complexity; (3) filtering for verifiable answers; and (4) optimizing data blending strategies that utilizes data from multiple sources effectively. Our approach enables scalable and verifiable reward modeling beyond mathematics and demonstrates improved accuracies on both math (MATH-500: +30.1%, AMC23:+27.5%) and non-math reasoning benchmarks (MMLU-PRO: +12.8%, GPQA-DIAMOND: +11.3%, AGIEVAL: +15.1%, SUPERGPQA: +3.8%). Moreover, NEMOTRON-CROSSTHINK exhibits significantly improved response efficiency -- using 28% fewer tokens for correct answers -- highlighting more focused and effective reasoning. Through NEMOTRON-CROSSTHINK, we demonstrate that integrating multi-domain, multi-format data in RL leads to more accurate, efficient, and generalizable LLMs.

Sahiti Yerramilli, Jayant Sravan Tamarapalli, Tanmay Girish Kulkarni et al.

Deep Learning models are incredibly data-hungry and require very large labeled datasets for supervised learning. As a consequence, these models often suffer from overfitting, limiting their ability to generalize to real-world examples. Recent advancements in diffusion models have enabled the generation of photorealistic images based on textual inputs. Leveraging the substantial datasets used to train these diffusion models, we propose a technique to utilize generated images to augment existing datasets. This paper explores various strategies for effective data augmentation to improve the out-of-domain generalization capabilities of deep learning models.

Syeda Nahida Akter, Shrimai Prabhumoye, John Kamalu et al.

The utility of synthetic data to enhance pretraining data quality and hence to improve downstream task accuracy has been widely explored in recent large language models (LLMs). Yet, these approaches fall inadequate in complex, multi-hop and mathematical reasoning tasks as the synthetic data typically fails to add complementary knowledge to the existing raw corpus. In this work, we propose a novel large-scale and diverse Math Informed syNthetic Dialogue (MIND) generation method that improves the mathematical reasoning ability of LLMs. Specifically, using MIND, we generate synthetic conversations based on OpenWebMath (OWM), resulting in a new math corpus, MIND-OWM. Our experiments with different conversational settings reveal that incorporating knowledge gaps between dialog participants is essential for generating high-quality math data. We further identify an effective way to format and integrate synthetic and raw data during pretraining to maximize the gain in mathematical reasoning, emphasizing the need to restructure raw data rather than use it as-is. Compared to pretraining just on raw data, a model pretrained on MIND-OWM shows significant boost in mathematical reasoning (GSM8K: +13.42%, MATH: +2.30%), including superior performance in specialized knowledge (MMLU: +4.55%, MMLU-STEM: +4.28%) and general purpose reasoning tasks (GENERAL REASONING: +2.51%).

Sahiti Yerramilli, Jayant Sravan Tamarapalli, Jonathan Francis et al.

Multimodal machine learning has gained significant attention in recent years due to its potential for integrating information from multiple modalities to enhance learning and decision-making processes. However, it is commonly observed that unimodal models outperform multimodal models, despite the latter having access to richer information. Additionally, the influence of a single modality often dominates the decision-making process, resulting in suboptimal performance. This research project aims to address these challenges by proposing a novel regularization term that encourages multimodal models to effectively utilize information from all modalities when making decisions. The focus of this project lies in the video-audio domain, although the proposed regularization technique holds promise for broader applications in embodied AI research, where multiple modalities are involved. By leveraging this regularization term, the proposed approach aims to mitigate the issue of unimodal dominance and improve the performance of multimodal machine learning systems. Through extensive experimentation and evaluation, the effectiveness and generalizability of the proposed technique will be assessed. The findings of this research project have the potential to significantly contribute to the advancement of multimodal machine learning and facilitate its application in various domains, including multimedia analysis, human-computer interaction, and embodied AI research.

Syeda Nahida Akter, Sangwu Lee, Yingshan Chang et al. · cmu

Verifying a question's validity before answering is crucial in real-world applications, where users may provide imperfect instructions. In this scenario, an ideal model should address the discrepancies in the query and convey them to the users rather than generating the best possible answer. Addressing this requirement, we introduce a new compositional visual question-answering dataset, VISREAS, that consists of answerable and unanswerable visual queries formulated by traversing and perturbing commonalities and differences among objects, attributes, and relations. VISREAS contains 2.07M semantically diverse queries generated automatically using Visual Genome scene graphs. The unique feature of this task, validating question answerability with respect to an image before answering, and the poor performance of state-of-the-art models inspired the design of a new modular baseline, LOGIC2VISION that reasons by producing and executing pseudocode without any external modules to generate the answer. LOGIC2VISION outperforms generative models in VISREAS (+4.82% over LLaVA-1.5; +12.23% over InstructBLIP) and achieves a significant gain in performance against the classification models.

Syeda Nahida Akter, Shrimai Prabhumoye, Eric Nyberg et al.

The prevailing paradigm for enhancing the reasoning abilities of LLMs revolves around post-training on high-quality, reasoning-intensive data. While emerging literature suggests that reasoning data is increasingly incorporated also during the mid-training stage-a practice that is relatively more proprietary and less openly characterized-the role of such data in pretraining remains unclear. In particular, due to the opaqueness of pretraining corpora in most frontier models, the effect of reasoning data introduced at different phases of pre- and/or post-training is relatively less reported in the scientific literature. This raises several important questions: Is adding reasoning data earlier during pretraining any better than introducing it during post-training? Could earlier inclusion risk overfitting and harm generalization, or instead establish durable foundations that later fine-tuning cannot recover? We conduct the first systematic study of how reasoning data-varying in scale, diversity, and quality-affects LLM performance when introduced at different stages of training. We find that front-loading reasoning data into pretraining is critical (19% avg gain), establishing foundational capabilities that cannot be fully replicated by later-stage SFT, even with more data. We uncover an asymmetric principle for optimal data allocation: pretraining benefits most from broad diversity in reasoning patterns (11% avg gain), while SFT is more sensitive to data quality (15% avg gain). We show that high-quality pretraining data has latent effects, activated only after SFT, and that naively scaling SFT data can be detrimental, washing away the benefits of early reasoning injection. Our results challenge the conventional separation of language modeling and reasoning, providing a principled guide for strategically allocating data across the entire training pipeline to build more capable models.

Syeda Nahida Akter, Aman Madaan, Sangwu Lee et al.

The potential of Vision-Language Models (VLMs) often remains underutilized in handling complex text-based problems, particularly when these problems could benefit from visual representation. Resonating with humans' ability to solve complex text-based problems by (1) creating a visual diagram from the problem and (2) deducing what steps they need to take to solve it, we propose Self-Imagine. We leverage a single Vision-Language Model (VLM) to generate a structured representation of the question using HTML, then render the HTML as an image, and finally use the same VLM to answer the question using both the question and the image. Our approach does not require any additional training data or training. We evaluate our approach on three mathematics tasks and nine general-purpose reasoning tasks using state-of-the-art (LLAVA-1.5 and GEMINI PRO) VLMs. Our approach boosts the performance of LLAVA-1.5 and GEMINI PRO on all math tasks (on average GSM8K: +3.1%; ASDIV: +3.2%; SVAMP: +6.9%) and the majority of the general-purpose reasoning tasks by 3.2% to 6.0% on average.

Alex Wilf, Syeda Nahida Akter, Leena Mathur et al.

The self-supervised objective of masking-and-predicting has led to promising performance gains on a variety of downstream tasks. However, while most approaches randomly mask tokens, there is strong intuition that deciding what to mask can substantially improve learning outcomes. We investigate this in continued pretraining setting in which pretrained models continue to pretrain on domain-specific data before performing some downstream task. We introduce Difference-Masking, a masking strategy that automatically chooses what to mask during continued pretraining by considering what makes a task domain different from the pretraining domain. Empirically, we find that Difference-Masking outperforms baselines on continued pretraining settings across four diverse language-only and multimodal video tasks.

Kaixin Ma, Hao Cheng, Yu Zhang et al.

The retrieval model is an indispensable component for real-world knowledge-intensive tasks, e.g., open-domain question answering (ODQA). As separate retrieval skills are annotated for different datasets, recent work focuses on customized methods, limiting the model transferability and scalability. In this work, we propose a modular retriever where individual modules correspond to key skills that can be reused across datasets. Our approach supports flexible skill configurations based on the target domain to boost performance. To mitigate task interference, we design a novel modularization parameterization inspired by sparse Transformer. We demonstrate that our model can benefit from self-supervised pretraining on Wikipedia and fine-tuning using multiple ODQA datasets, both in a multi-task fashion. Our approach outperforms recent self-supervised retrievers in zero-shot evaluations and achieves state-of-the-art fine-tuned retrieval performance on NQ, HotpotQA and OTT-QA.

Kaixin Ma, Hao Cheng, Xiaodong Liu et al.

The retriever-reader framework is popular for open-domain question answering (ODQA) due to its ability to use explicit knowledge. Although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text, accessing heterogeneous knowledge sources through a unified interface remains an open question. While data-to-text generation has the potential to serve as a universal interface for data and text, its feasibility for downstream tasks remains largely unknown. In this work, we bridge this gap and use the data-to-text method as a means for encoding structured knowledge for ODQA. Specifically, we propose a verbalizer-retriever-reader framework for ODQA over data and text where verbalized tables from Wikipedia and graphs from Wikidata are used as augmented knowledge sources. We show that our Unified Data and Text QA, UDT-QA, can effectively benefit from the expanded knowledge index, leading to large gains over text-only baselines. Notably, our approach sets the single-model state-of-the-art on Natural Questions. Furthermore, our analyses indicate that verbalized knowledge is preferred for answer reasoning for both adapted and hot-swap settings.

Bingqing Chen, Jonathan Francis, Jean Oh et al.

Racing demands each vehicle to drive at its physical limits, when any safety infraction could lead to catastrophic failure. In this work, we study the problem of safe reinforcement learning (RL) for autonomous racing, using the vehicle's ego-camera view and speed as input. Given the nature of the task, autonomous agents need to be able to 1) identify and avoid unsafe scenarios under the complex vehicle dynamics, and 2) make sub-second decision in a fast-changing environment. To satisfy these criteria, we propose to incorporate Hamilton-Jacobi (HJ) reachability theory, a safety verification method for general non-linear systems, into the constrained Markov decision process (CMDP) framework. HJ reachability not only provides a control-theoretic approach to learn about safety, but also enables low-latency safety verification. Though HJ reachability is traditionally not scalable to high-dimensional systems, we demonstrate that with neural approximation, the HJ safety value can be learned directly on vision context -- the highest-dimensional problem studied via the method, to-date. We evaluate our method on several benchmark tasks, including Safety Gym and Learn-to-Race (L2R), a recently-released high-fidelity autonomous racing environment. Our approach has significantly fewer constraint violations in comparison to other constrained RL baselines in Safety Gym, and achieves the new state-of-the-art results on the L2R benchmark task. We provide additional visualization of agent behavior at the following anonymized paper website: https://sites.google.com/view/safeautonomousracing/home

Kaixin Ma, Filip Ilievski, Jonathan Francis et al.

Commonsense reasoning benchmarks have been largely solved by fine-tuning language models. The downside is that fine-tuning may cause models to overfit to task-specific data and thereby forget their knowledge gained during pre-training. Recent works only propose lightweight model updates as models may already possess useful knowledge from past experience, but a challenge remains in understanding what parts and to what extent models should be refined for a given task. In this paper, we investigate what models learn from commonsense reasoning datasets. We measure the impact of three different adaptation methods on the generalization and accuracy of models. Our experiments with two models show that fine-tuning performs best, by learning both the content and the structure of the task, but suffers from overfitting and limited generalization to novel answers. We observe that alternative adaptation methods like prefix-tuning have comparable accuracy, but generalize better to unseen answers and are more robust to adversarial splits.

Yikang Li, Pulkit Goel, Varsha Kuppur Rajendra et al.

Conditional text generation has been a challenging task that is yet to see human-level performance from state-of-the-art models. In this work, we specifically focus on the Commongen benchmark, wherein the aim is to generate a plausible sentence for a given set of input concepts. Despite advances in other tasks, large pre-trained language models that are fine-tuned on this dataset often produce sentences that are syntactically correct but qualitatively deviate from a human understanding of common sense. Furthermore, generated sequences are unable to fulfill such lexical requirements as matching part-of-speech and full concept coverage. In this paper, we explore how commonsense knowledge graphs can enhance model performance, with respect to commonsense reasoning and lexically-constrained decoding. We propose strategies for enhancing the semantic correctness of the generated text, which we accomplish through: extracting commonsense relations from Conceptnet, injecting these relations into the Unified Language Model (UniLM) through attention mechanisms, and enforcing the aforementioned lexical requirements through output constraints. By performing several ablations, we find that commonsense injection enables the generation of sentences that are more aligned with human understanding, while remaining compliant with lexical requirements.

Kaixin Ma, Filip Ilievski, Jonathan Francis et al.

Recent developments in pre-trained neural language modeling have led to leaps in accuracy on commonsense question-answering benchmarks. However, there is increasing concern that models overfit to specific tasks, without learning to utilize external knowledge or perform general semantic reasoning. In contrast, zero-shot evaluations have shown promise as a more robust measure of a model's general reasoning abilities. In this paper, we propose a novel neuro-symbolic framework for zero-shot question answering across commonsense tasks. Guided by a set of hypotheses, the framework studies how to transform various pre-existing knowledge resources into a form that is most effective for pre-training models. We vary the set of language models, training regimes, knowledge sources, and data generation strategies, and measure their impact across tasks. Extending on prior work, we devise and compare four constrained distractor-sampling strategies. We provide empirical results across five commonsense question-answering tasks with data generated from five external knowledge resources. We show that, while an individual knowledge graph is better suited for specific tasks, a global knowledge graph brings consistent gains across different tasks. In addition, both preserving the structure of the task as well as generating fair and informative questions help language models learn more effectively.

Leonid Boytsov, Eric Nyberg

Our objective is to introduce to the NLP community an existing k-NN search library NMSLIB, a new retrieval toolkit FlexNeuART, as well as their integration capabilities. NMSLIB, while being one the fastest k-NN search libraries, is quite generic and supports a variety of distance/similarity functions. Because the library relies on the distance-based structure-agnostic algorithms, it can be further extended by adding new distances. FlexNeuART is a modular, extendible and flexible toolkit for candidate generation in IR and QA applications, which supports mixing of classic and neural ranking signals. FlexNeuART can efficiently retrieve mixed dense and sparse representations (with weights learned from training data), which is achieved by extending NMSLIB. In that, other retrieval systems work with purely sparse representations (e.g., Lucene), purely dense representations (e.g., FAISS and Annoy), or only perform mixing at the re-ranking stage.

Kaixin Ma, Jonathan Francis, Quanyang Lu et al.

Non-extractive commonsense QA remains a challenging AI task, as it requires systems to reason about, synthesize, and gather disparate pieces of information, in order to generate responses to queries. Recent approaches on such tasks show increased performance, only when models are either pre-trained with additional information or when domain-specific heuristics are used, without any special consideration regarding the knowledge resource type. In this paper, we perform a survey of recent commonsense QA methods and we provide a systematic analysis of popular knowledge resources and knowledge-integration methods, across benchmarks from multiple commonsense datasets. Our results and analysis show that attention-based injection seems to be a preferable choice for knowledge integration and that the degree of domain overlap, between knowledge bases and datasets, plays a crucial role in determining model success.

Leonid Boytsov, Eric Nyberg

We focus on low-dimensional non-metric search, where tree-based approaches permit efficient and accurate retrieval while having short indexing time. These methods rely on space partitioning and require a pruning rule to avoid visiting unpromising parts. We consider two known data-driven approaches to extend these rules to non-metric spaces: TriGen and a piece-wise linear approximation of the pruning rule. We propose and evaluate two adaptations of TriGen to non-symmetric similarities (TriGen does not support non-symmetric distances). We also evaluate a hybrid of TriGen and the piece-wise linear approximation pruning. We find that this hybrid approach is often more effective than either of the pruning rules. We make our software publicly available.

Leonid Boytsov, Eric Nyberg

We demonstrate that a graph-based search algorithm-relying on the construction of an approximate neighborhood graph-can directly work with challenging non-metric and/or non-symmetric distances without resorting to metric-space mapping and/or distance symmetrization, which, in turn, lead to substantial performance degradation. Although the straightforward metrization and symmetrization is usually ineffective, we find that constructing an index using a modified, e.g., symmetrized, distance can improve performance. This observation paves a way to a new line of research of designing index-specific graph-construction distance functions.

Vinayshekhar Bannihatti Kumar, Ashwin Srinivasan, Aditi Chaudhary et al.

This paper presents the submissions by Team Dr.Quad to the ACL-BioNLP 2019 shared task on Textual Inference and Question Entailment in the Medical Domain. Our system is based on the prior work Liu et al. (2019) which uses a multi-task objective function for textual entailment. In this work, we explore different strategies for generalizing state-of-the-art language understanding models to the specialized medical domain. Our results on the shared task demonstrate that incorporating domain knowledge through data augmentation is a powerful strategy for addressing challenges posed by specialized domains such as medicine.

Hemant Pugaliya, Karan Saxena, Shefali Garg et al.

Parallel deep learning architectures like fine-tuned BERT and MT-DNN, have quickly become the state of the art, bypassing previous deep and shallow learning methods by a large margin. More recently, pre-trained models from large related datasets have been able to perform well on many downstream tasks by just fine-tuning on domain-specific datasets . However, using powerful models on non-trivial tasks, such as ranking and large document classification, still remains a challenge due to input size limitations of parallel architecture and extremely small datasets (insufficient for fine-tuning). In this work, we introduce an end-to-end system, trained in a multi-task setting, to filter and re-rank answers in the medical domain. We use task-specific pre-trained models as deep feature extractors. Our model achieves the highest Spearman's Rho and Mean Reciprocal Rank of 0.338 and 0.9622 respectively, on the ACL-BioNLP workshop MediQA Question Answering shared-task.

Soumya Wadhwa, Khyathi Raghavi Chandu, Eric Nyberg

The task of Question Answering has gained prominence in the past few decades for testing the ability of machines to understand natural language. Large datasets for Machine Reading have led to the development of neural models that cater to deeper language understanding compared to information retrieval tasks. Different components in these neural architectures are intended to tackle different challenges. As a first step towards achieving generalization across multiple domains, we attempt to understand and compare the peculiarities of existing end-to-end neural models on the Stanford Question Answering Dataset (SQuAD) by performing quantitative as well as qualitative analysis of the results attained by each of them. We observed that prediction errors reflect certain model-specific biases, which we further discuss in this paper.

Soumya Wadhwa, Varsha Embar, Matthias Grabmair et al.

In this paper, we investigate the tendency of end-to-end neural Machine Reading Comprehension (MRC) models to match shallow patterns rather than perform inference-oriented reasoning on RC benchmarks. We aim to test the ability of these systems to answer questions which focus on referential inference. We propose ParallelQA, a strategy to formulate such questions using parallel passages. We also demonstrate that existing neural models fail to generalize well to this setting.

Yuan Yang, Jingcheng Yu, Ye Hu et al.

In this paper, we present LiveMedQA, a question answering system that is optimized for consumer health question. On top of the general QA system pipeline, we introduce several new features that aim to exploit domain-specific knowledge and entity structures for better performance. This includes a question type/focus analyzer based on deep text classification model, a tree-based knowledge graph for answer generation and a complementary structure-aware searcher for answer retrieval. LiveMedQA system is evaluated in the TREC 2017 LiveQA medical subtask, where it received an average score of 0.356 on a 3 point scale. Evaluation results revealed 3 substantial drawbacks in current LiveMedQA system, based on which we provide a detailed discussion and propose a few solutions that constitute the main focus of our subsequent work.

Di Wang, Nebojsa Jojic, Chris Brockett et al.

We propose simple and flexible training and decoding methods for influencing output style and topic in neural encoder-decoder based language generation. This capability is desirable in a variety of applications, including conversational systems, where successful agents need to produce language in a specific style and generate responses steered by a human puppeteer or external knowledge. We decompose the neural generation process into empirically easier sub-problems: a faithfulness model and a decoding method based on selective-sampling. We also describe training and sampling algorithms that bias the generation process with a specific language style restriction, or a topic restriction. Human evaluation results show that our proposed methods are able to restrict style and topic without degrading output quality in conversational tasks.

Varun Gangal, Harsh Jhamtani, Graham Neubig et al.

Portmanteaus are a word formation phenomenon where two words are combined to form a new word. We propose character-level neural sequence-to-sequence (S2S) methods for the task of portmanteau generation that are end-to-end-trainable, language independent, and do not explicitly use additional phonetic information. We propose a noisy-channel-style model, which allows for the incorporation of unsupervised word lists, improving performance over a standard source-to-target model. This model is made possible by an exhaustive candidate generation strategy specifically enabled by the features of the portmanteau task. Experiments find our approach superior to a state-of-the-art FST-based baseline with respect to ground truth accuracy and human evaluation.

Harsh Jhamtani, Varun Gangal, Eduard Hovy et al.

Variations in writing styles are commonly used to adapt the content to a specific context, audience, or purpose. However, applying stylistic variations is still by and large a manual process, and there have been little efforts towards automating it. In this paper we explore automated methods to transform text from modern English to Shakespearean English using an end to end trainable neural model with pointers to enable copy action. To tackle limited amount of parallel data, we pre-train embeddings of words by leveraging external dictionaries mapping Shakespearean words to modern English words as well as additional text. Our methods are able to get a BLEU score of 31+, an improvement of ~6 points above the strongest baseline. We publicly release our code to foster further research in this area.

Rui Liu, Junjie Hu, Wei Wei et al.

Deep neural networks for machine comprehension typically utilizes only word or character embeddings without explicitly taking advantage of structured linguistic information such as constituency trees and dependency trees. In this paper, we propose structural embedding of syntactic trees (SEST), an algorithm framework to utilize structured information and encode them into vector representations that can boost the performance of algorithms for the machine comprehension. We evaluate our approach using a state-of-the-art neural attention model on the SQuAD dataset. Experimental results demonstrate that our model can accurately identify the syntactic boundaries of the sentences and extract answers that are syntactically coherent over the baseline methods.

Leonid Boytsov, David Novak, Yury Malkov et al.

Retrieval pipelines commonly rely on a term-based search to obtain candidate records, which are subsequently re-ranked. Some candidates are missed by this approach, e.g., due to a vocabulary mismatch. We address this issue by replacing the term-based search with a generic k-NN retrieval algorithm, where a similarity function can take into account subtle term associations. While an exact brute-force k-NN search using this similarity function is slow, we demonstrate that an approximate algorithm can be nearly two orders of magnitude faster at the expense of only a small loss in accuracy. A retrieval pipeline using an approximate k-NN search can be more effective and efficient than the term-based pipeline. This opens up new possibilities for designing effective retrieval pipelines. Our software (including data-generating code) and derivative data based on the Stack Overflow collection is available online.

Bilegsaikhan Naidan, Leonid Boytsov, Eric Nyberg

We survey permutation-based methods for approximate k-nearest neighbor search. In these methods, every data point is represented by a ranked list of pivots sorted by the distance to this point. Such ranked lists are called permutations. The underpinning assumption is that, for both metric and non-metric spaces, the distance between permutations is a good proxy for the distance between original points. Thus, it should be possible to efficiently retrieve most true nearest neighbors by examining only a tiny subset of data points whose permutations are similar to the permutation of a query. We further test this assumption by carrying out an extensive experimental evaluation where permutation methods are pitted against state-of-the art benchmarks (the multi-probe LSH, the VP-tree, and proximity-graph based retrieval) on a variety of realistically large data set from the image and textual domain. The focus is on the high-accuracy retrieval methods for generic spaces. Additionally, we assume that both data and indices are stored in main memory. We find permutation methods to be reasonably efficient and describe a setup where these methods are most useful. To ease reproducibility, we make our software and data sets publicly available.