Dipanjan Das

Freda Shi, Mirac Suzgun, Markus Freitag et al. · deepmind

We evaluate the reasoning abilities of large language models in multilingual settings. We introduce the Multilingual Grade School Math (MGSM) benchmark, by manually translating 250 grade-school math problems from the GSM8K dataset (Cobbe et al., 2021) into ten typologically diverse languages. We find that the ability to solve MGSM problems via chain-of-thought prompting emerges with increasing model scale, and that models have strikingly strong multilingual reasoning abilities, even in underrepresented languages such as Bengali and Swahili. Finally, we show that the multilingual reasoning abilities of language models extend to other tasks such as commonsense reasoning and word-in-context semantic judgment. The MGSM benchmark is publicly available at https://github.com/google-research/url-nlp.

Bernd Bohnet, Vinh Q. Tran, Pat Verga et al. · deepmind, mit

Large language models (LLMs) have shown impressive results while requiring little or no direct supervision. Further, there is mounting evidence that LLMs may have potential in information-seeking scenarios. We believe the ability of an LLM to attribute the text that it generates is likely to be crucial in this setting. We formulate and study Attributed QA as a key first step in the development of attributed LLMs. We propose a reproducible evaluation framework for the task and benchmark a broad set of architectures. We take human annotations as a gold standard and show that a correlated automatic metric is suitable for development. Our experimental work gives concrete answers to two key questions (How to measure attribution?, and How well do current state-of-the-art methods perform on attribution?), and give some hints as to how to address a third (How to build LLMs with attribution?).

Shashi Narayan, Joshua Maynez, Reinald Kim Amplayo et al.

The ability to convey relevant and faithful information is critical for many tasks in conditional generation and yet remains elusive for neural seq-to-seq models whose outputs often reveal hallucinations and fail to correctly cover important details. In this work, we advocate planning as a useful intermediate representation for rendering conditional generation less opaque and more grounded. Our work proposes a new conceptualization of text plans as a sequence of question-answer (QA) pairs. We enhance existing datasets (e.g., for summarization) with a QA blueprint operating as a proxy for both content selection (i.e.,~what to say) and planning (i.e.,~in what order). We obtain blueprints automatically by exploiting state-of-the-art question generation technology and convert input-output pairs into input-blueprint-output tuples. We develop Transformer-based models, each varying in how they incorporate the blueprint in the generated output (e.g., as a global plan or iteratively). Evaluation across metrics and datasets demonstrates that blueprint models are more factual than alternatives which do not resort to planning and allow tighter control of the generation output.

Shashi Narayan, Gonçalo Simões, Yao Zhao et al.

We propose Composition Sampling, a simple but effective method to generate diverse outputs for conditional generation of higher quality compared to previous stochastic decoding strategies. It builds on recently proposed plan-based neural generation models (Narayan et al, 2021) that are trained to first create a composition of the output and then generate by conditioning on it and the input. Our approach avoids text degeneration by first sampling a composition in the form of an entity chain and then using beam search to generate the best possible text grounded to this entity chain. Experiments on summarization (CNN/DailyMail and XSum) and question generation (SQuAD), using existing and newly proposed automatic metrics together with human-based evaluation, demonstrate that Composition Sampling is currently the best available decoding strategy for generating diverse meaningful outputs.

Priyanka Agrawal, Chris Alberti, Fantine Huot et al.

The availability of large, high-quality datasets has been one of the main drivers of recent progress in question answering (QA). Such annotated datasets however are difficult and costly to collect, and rarely exist in languages other than English, rendering QA technology inaccessible to underrepresented languages. An alternative to building large monolingual training datasets is to leverage pre-trained language models (PLMs) under a few-shot learning setting. Our approach, QAmeleon, uses a PLM to automatically generate multilingual data upon which QA models are trained, thus avoiding costly annotation. Prompt tuning the PLM for data synthesis with only five examples per language delivers accuracy superior to translation-based baselines, bridges nearly 60% of the gap between an English-only baseline and a fully supervised upper bound trained on almost 50,000 hand labeled examples, and always leads to substantial improvements compared to fine-tuning a QA model directly on labeled examples in low resource settings. Experiments on the TyDiQA-GoldP and MLQA benchmarks show that few-shot prompt tuning for data synthesis scales across languages and is a viable alternative to large-scale annotation.

Fantine Huot, Joshua Maynez, Shashi Narayan et al.

While conditional generation models can now generate natural language well enough to create fluent text, it is still difficult to control the generation process, leading to irrelevant, repetitive, and hallucinated content. Recent work shows that planning can be a useful intermediate step to render conditional generation less opaque and more grounded. We present a web browser-based demonstration for query-focused summarization that uses a sequence of question-answer pairs, as a blueprint plan for guiding text generation (i.e., what to say and in what order). We illustrate how users may interact with the generated text and associated plan visualizations, e.g., by editing and modifying the blueprint in order to improve or control the generated output. A short video demonstrating our system is available at https://goo.gle/text-blueprint-demo.

Reinald Kim Amplayo, Kellie Webster, Michael Collins et al.

Large language models (LLMs) have been shown to perform well in answering questions and in producing long-form texts, both in few-shot closed-book settings. While the former can be validated using well-known evaluation metrics, the latter is difficult to evaluate. We resolve the difficulties to evaluate long-form output by doing both tasks at once -- to do question answering that requires long-form answers. Such questions tend to be multifaceted, i.e., they may have ambiguities and/or require information from multiple sources. To this end, we define query refinement prompts that encourage LLMs to explicitly express the multifacetedness in questions and generate long-form answers covering multiple facets of the question. Our experiments on two long-form question answering datasets, ASQA and AQuAMuSe, show that using our prompts allows us to outperform fully finetuned models in the closed book setting, as well as achieve results comparable to retrieve-then-generate open-book models.

Aileen Cheng, Alon Jacovi, Amir Globerson et al.

We introduce The FACTS Leaderboard, an online leaderboard suite and associated set of benchmarks that comprehensively evaluates the ability of language models to generate factually accurate text across diverse scenarios. The suite provides a holistic measure of factuality by aggregating the performance of models on four distinct sub-leaderboards: (1) FACTS Multimodal, which measures the factuality of responses to image-based questions; (2) FACTS Parametric, which assesses models' world knowledge by answering closed-book factoid questions from internal parameters; (3) FACTS Search, which evaluates factuality in information-seeking scenarios, where the model must use a search API; and (4) FACTS Grounding (v2), which evaluates whether long-form responses are grounded in provided documents, featuring significantly improved judge models. Each sub-leaderboard employs automated judge models to score model responses, and the final suite score is an average of the four components, designed to provide a robust and balanced assessment of a model's overall factuality. The FACTS Leaderboard Suite will be actively maintained, containing both public and private splits to allow for external participation while guarding its integrity. It can be found at https://www.kaggle.com/benchmarks/google/facts .

Gemini Team, Petko Georgiev, Ving Ian Lei et al. · deepmind, mila

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Alon Jacovi, Andrew Wang, Chris Alberti et al. · deepmind

We introduce FACTS Grounding, an online leaderboard and associated benchmark that evaluates language models' ability to generate text that is factually accurate with respect to given context in the user prompt. In our benchmark, each prompt includes a user request and a full document, with a maximum length of 32k tokens, requiring long-form responses. The long-form responses are required to be fully grounded in the provided context document while fulfilling the user request. Models are evaluated using automated judge models in two phases: (1) responses are disqualified if they do not fulfill the user request; (2) they are judged as accurate if the response is fully grounded in the provided document. The automated judge models were comprehensively evaluated against a held-out test-set to pick the best prompt template, and the final factuality score is an aggregate of multiple judge models to mitigate evaluation bias. The FACTS Grounding leaderboard will be actively maintained over time, and contains both public and private splits to allow for external participation while guarding the integrity of the leaderboard. It can be found at https://www.kaggle.com/facts-leaderboard.

Nikita Gupta, Riju Chatterjee, Lukas Haas et al.

We introduce DeepSearchQA, a 900-prompt benchmark for evaluating agents on difficult multi-step information-seeking tasks across 17 different fields. Unlike traditional benchmarks that target single answer retrieval or broad-spectrum factuality, DeepSearchQA features a dataset of challenging, handcrafted tasks designed to evaluate an agent's ability to execute complex search plans to generate exhaustive answer lists. This shift in design explicitly tests three critical, yet under-evaluated capabilities: 1) systematic collation of fragmented information from disparate sources, 2) de-duplication and entity resolution to ensure precision, and 3) the ability to reason about stopping criteria within an open-ended search space. Each task is structured as a causal chain, where discovering information for one step is dependent on the successful completion of the previous one, stressing long-horizon planning and context retention. All tasks are grounded in the open web with objectively verifiable answer sets. Our comprehensive evaluation of state-of-the-art agent architectures reveals significant performance limitations: even the most advanced models struggle to balance high recall with precision. We observe distinct failure modes ranging from premature stopping (under-retrieval) to hedging behaviors, where agents cast an overly wide net of low-confidence answers to artificially boost recall. These findings highlight critical headroom in current agent designs and position DeepSearchQA as an essential diagnostic tool for driving future research toward more robust, deep-research capabilities.

Lukas Haas, Gal Yona, Giovanni D'Antonio et al.

We introduce SimpleQA Verified, a 1,000-prompt benchmark for evaluating Large Language Model (LLM) short-form factuality based on OpenAI's SimpleQA. It addresses critical limitations in OpenAI's benchmark, including noisy and incorrect labels, topical biases, and question redundancy. SimpleQA Verified was created through a rigorous multi-stage filtering process involving de-duplication, topic balancing, and source reconciliation to produce a more reliable and challenging evaluation set, alongside improvements in the autorater prompt. On this new benchmark, Gemini 2.5 Pro achieves a state-of-the-art F1-score of 55.6, outperforming other frontier models, including GPT-5. This work provides the research community with a higher-fidelity tool to track genuine progress in parametric model factuality and to mitigate hallucinations. The benchmark dataset, evaluation code, and leaderboard are available at: https://www.kaggle.com/benchmarks/deepmind/simpleqa-verified.

Chaitanya Malaviya, Priyanka Agrawal, Kuzman Ganchev et al.

Experts in various fields routinely perform methodical writing tasks to plan, organize, and report their work. From a clinician writing a differential diagnosis for a patient, to a teacher writing a lesson plan for students, these tasks are pervasive, requiring to methodically generate structured long-form output for a given input. We develop a typology of methodical tasks structured in the form of a task objective, procedure, input, and output, and introduce DoLoMiTes, a novel benchmark with specifications for 519 such tasks elicited from hundreds of experts from across 25 fields. Our benchmark further contains specific instantiations of methodical tasks with concrete input and output examples (1,857 in total) which we obtain by collecting expert revisions of up to 10 model-generated examples of each task. We use these examples to evaluate contemporary language models highlighting that automating methodical tasks is a challenging long-form generation problem, as it requires performing complex inferences, while drawing upon the given context as well as domain knowledge.

Shivam Singh, Karthik Swaminathan, Raghav Arora et al.

An agent assisting humans in daily living activities can collaborate more effectively by anticipating upcoming tasks. Data-driven methods represent the state of the art in task anticipation, planning, and related problems, but these methods are resource-hungry and opaque. Our prior work introduced a proof of concept framework that used an LLM to anticipate 3 high-level tasks that served as goals for a classical planning system that computed a sequence of low-level actions for the agent to achieve these goals. This paper describes DaTAPlan, our framework that significantly extends our prior work toward human-robot collaboration. Specifically, DaTAPlan planner computes actions for an agent and a human to collaboratively and jointly achieve the tasks anticipated by the LLM, and the agent automatically adapts to unexpected changes in human action outcomes and preferences. We evaluate DaTAPlan capabilities in a realistic simulation environment, demonstrating accurate task anticipation, effective human-robot collaboration, and the ability to adapt to unexpected changes. Project website: https://dataplan-hrc.github.io

Gemini Team, Rohan Anil, Sebastian Borgeaud et al.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

Elizabeth Clark, Shruti Rijhwani, Sebastian Gehrmann et al.

Reliable automatic evaluation of summarization systems is challenging due to the multifaceted and subjective nature of the task. This is especially the case for languages other than English, where human evaluations are scarce. In this work, we introduce SEAHORSE, a dataset for multilingual, multifaceted summarization evaluation. SEAHORSE consists of 96K summaries with human ratings along 6 dimensions of text quality: comprehensibility, repetition, grammar, attribution, main ideas, and conciseness, covering 6 languages, 9 systems and 4 datasets. As a result of its size and scope, SEAHORSE can serve both as a benchmark to evaluate learnt metrics, as well as a large-scale resource for training such metrics. We show that metrics trained with SEAHORSE achieve strong performance on the out-of-domain meta-evaluation benchmarks TRUE (Honovich et al., 2022) and mFACE (Aharoni et al., 2022). We make the SEAHORSE dataset and metrics publicly available for future research on multilingual and multifaceted summarization evaluation.

Hannah Rashkin, Vitaly Nikolaev, Matthew Lamm et al.

With recent improvements in natural language generation (NLG) models for various applications, it has become imperative to have the means to identify and evaluate whether NLG output is only sharing verifiable information about the external world. In this work, we present a new evaluation framework entitled Attributable to Identified Sources (AIS) for assessing the output of natural language generation models, when such output pertains to the external world. We first define AIS and introduce a two-stage annotation pipeline for allowing annotators to appropriately evaluate model output according to AIS guidelines. We empirically validate this approach on generation datasets spanning three tasks (two conversational QA datasets, a summarization dataset, and a table-to-text dataset) via human evaluation studies that suggest that AIS could serve as a common framework for measuring whether model-generated statements are supported by underlying sources. We release guidelines for the human evaluation studies.

Dipanjan Das, Priyanka Bose, Nicola Ruaro et al.

Non-Fungible Tokens (NFTs) have emerged as a way to collect digital art as well as an investment vehicle. Despite having been popularized only recently, NFT markets have witnessed several high-profile (and high-value) asset sales and a tremendous growth in trading volumes over the last year. Unfortunately, these marketplaces have not yet received much security scrutiny. Instead, most academic research has focused on attacks against decentralized finance (DeFi) protocols and automated techniques to detect smart contract vulnerabilities. To the best of our knowledge, we are the first to study the market dynamics and security issues of the multi-billion dollar NFT ecosystem. In this paper, we first present a systematic overview of how the NFT ecosystem works, and we identify three major actors: marketplaces, external entities, and users. We perform an in-depth analysis of the top 8 marketplaces (ranked by transaction volume) to discover potential issues associated with such marketplaces. Many of these issues can lead to substantial financial losses. We also collected a large amount of asset and event data pertaining to the NFTs being traded in the examined marketplaces. We automatically analyze this data to understand how the entities external to the blockchain are able to interfere with NFT markets, leading to serious consequences, and quantify the malicious trading behaviors carried out by users under the cloak of anonymity.

Hannah Rashkin, David Reitter, Gaurav Singh Tomar et al.

Knowledge-grounded dialogue systems are intended to convey information that is based on evidence provided in a given source text. We discuss the challenges of training a generative neural dialogue model for such systems that is controlled to stay faithful to the evidence. Existing datasets contain a mix of conversational responses that are faithful to selected evidence as well as more subjective or chit-chat style responses. We propose different evaluation measures to disentangle these different styles of responses by quantifying the informativeness and objectivity. At training time, additional inputs based on these evaluation measures are given to the dialogue model. At generation time, these additional inputs act as stylistic controls that encourage the model to generate responses that are faithful to the provided evidence. We also investigate the usage of additional controls at decoding time using resampling techniques. In addition to automatic metrics, we perform a human evaluation study where raters judge the output of these controlled generation models to be generally more objective and faithful to the evidence compared to baseline dialogue systems.

Thibault Sellam, Steve Yadlowsky, Jason Wei et al.

Experiments with pre-trained models such as BERT are often based on a single checkpoint. While the conclusions drawn apply to the artifact tested in the experiment (i.e., the particular instance of the model), it is not always clear whether they hold for the more general procedure which includes the architecture, training data, initialization scheme, and loss function. Recent work has shown that repeating the pre-training process can lead to substantially different performance, suggesting that an alternate strategy is needed to make principled statements about procedures. To enable researchers to draw more robust conclusions, we introduce the MultiBERTs, a set of 25 BERT-Base checkpoints, trained with similar hyper-parameters as the original BERT model but differing in random weight initialization and shuffling of training data. We also define the Multi-Bootstrap, a non-parametric bootstrap method for statistical inference designed for settings where there are multiple pre-trained models and limited test data. To illustrate our approach, we present a case study of gender bias in coreference resolution, in which the Multi-Bootstrap lets us measure effects that may not be detected with a single checkpoint. We release our models and statistical library along with an additional set of 140 intermediate checkpoints captured during pre-training to facilitate research on learning dynamics.

Priyanka Bose, Dipanjan Das, Yanju Chen et al.

This paper presents SAILFISH, a scalable system for automatically finding state-inconsistency bugs in smart contracts. To make the analysis tractable, we introduce a hybrid approach that includes (i) a light-weight exploration phase that dramatically reduces the number of instructions to analyze, and (ii) a precise refinement phase based on symbolic evaluation guided by our novel value-summary analysis, which generates extra constraints to over-approximate the side effects of whole-program execution, thereby ensuring the precision of the symbolic evaluation. We developed a prototype of SAILFISH and evaluated its ability to detect two state-inconsistency flaws, viz., reentrancy and transaction order dependence (TOD) in Ethereum smart contracts. Further, we present detection rules for other kinds of smart contract flaws that SAILFISH can be extended to detect. Our experiments demonstrate the efficiency of our hybrid approach as well as the benefit of the value summary analysis. In particular, we show that S SAILFISH outperforms five state-of-the-art smart contract analyzers (SECURITY, MYTHRIL, OYENTE, SEREUM and VANDAL ) in terms of performance, and precision. In total, SAILFISH discovered 47 previously unknown vulnerable smart contracts out of 89,853 smart contracts from ETHERSCAN .

Eunsol Choi, Jennimaria Palomaki, Matthew Lamm et al.

Models for question answering, dialogue agents, and summarization often interpret the meaning of a sentence in a rich context and use that meaning in a new context. Taking excerpts of text can be problematic, as key pieces may not be explicit in a local window. We isolate and define the problem of sentence decontextualization: taking a sentence together with its context and rewriting it to be interpretable out of context, while preserving its meaning. We describe an annotation procedure, collect data on the Wikipedia corpus, and use the data to train models to automatically decontextualize sentences. We present preliminary studies that show the value of sentence decontextualization in a user facing task, and as preprocessing for systems that perform document understanding. We argue that decontextualization is an important subtask in many downstream applications, and that the definitions and resources provided can benefit tasks that operate on sentences that occur in a richer context.

Sebastian Gehrmann, Tosin Adewumi, Karmanya Aggarwal et al.

We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for which we are organizing a shared task at our ACL 2021 Workshop and to which we invite the entire NLG community to participate.

Thibault Sellam, Amy Pu, Hyung Won Chung et al.

The quality of machine translation systems has dramatically improved over the last decade, and as a result, evaluation has become an increasingly challenging problem. This paper describes our contribution to the WMT 2020 Metrics Shared Task, the main benchmark for automatic evaluation of translation. We make several submissions based on BLEURT, a previously published metric based on transfer learning. We extend the metric beyond English and evaluate it on 14 language pairs for which fine-tuning data is available, as well as 4 "zero-shot" language pairs, for which we have no labelled examples. Additionally, we focus on English to German and demonstrate how to combine BLEURT's predictions with those of YiSi and use alternative reference translations to enhance the performance. Empirical results show that the models achieve competitive results on the WMT Metrics 2019 Shared Task, indicating their promise for the 2020 edition.

Vignesh Prasad, Dipanjan Das, Brojeshwar Bhowmick

Recent advances in deep learning have shown their ability to learn strong feature representations for images. The task of image clustering naturally requires good feature representations to capture the distribution of the data and subsequently differentiate data points from one another. Often these two aspects are dealt with independently and thus traditional feature learning alone does not suffice in partitioning the data meaningfully. Variational Autoencoders (VAEs) naturally lend themselves to learning data distributions in a latent space. Since we wish to efficiently discriminate between different clusters in the data, we propose a method based on VAEs where we use a Gaussian Mixture prior to help cluster the images accurately. We jointly learn the parameters of both the prior and the posterior distributions. Our method represents a true Gaussian Mixture VAE. This way, our method simultaneously learns a prior that captures the latent distribution of the images and a posterior to help discriminate well between data points. We also propose a novel reparametrization of the latent space consisting of a mixture of discrete and continuous variables. One key takeaway is that our method generalizes better across different datasets without using any pre-training or learnt models, unlike existing methods, allowing it to be trained from scratch in an end-to-end manner. We verify our efficacy and generalizability experimentally by achieving state-of-the-art results among unsupervised methods on a variety of datasets. To the best of our knowledge, we are the first to pursue image clustering using VAEs in a purely unsupervised manner on real image datasets.

Ankur P. Parikh, Xuezhi Wang, Sebastian Gehrmann et al.

We present ToTTo, an open-domain English table-to-text dataset with over 120,000 training examples that proposes a controlled generation task: given a Wikipedia table and a set of highlighted table cells, produce a one-sentence description. To obtain generated targets that are natural but also faithful to the source table, we introduce a dataset construction process where annotators directly revise existing candidate sentences from Wikipedia. We present systematic analyses of our dataset and annotation process as well as results achieved by several state-of-the-art baselines. While usually fluent, existing methods often hallucinate phrases that are not supported by the table, suggesting that this dataset can serve as a useful research benchmark for high-precision conditional text generation.

Junghyun Min, R. Thomas McCoy, Dipanjan Das et al.

Pretrained neural models such as BERT, when fine-tuned to perform natural language inference (NLI), often show high accuracy on standard datasets, but display a surprising lack of sensitivity to word order on controlled challenge sets. We hypothesize that this issue is not primarily caused by the pretrained model's limitations, but rather by the paucity of crowdsourced NLI examples that might convey the importance of syntactic structure at the fine-tuning stage. We explore several methods to augment standard training sets with syntactically informative examples, generated by applying syntactic transformations to sentences from the MNLI corpus. The best-performing augmentation method, subject/object inversion, improved BERT's accuracy on controlled examples that diagnose sensitivity to word order from 0.28 to 0.73, without affecting performance on the MNLI test set. This improvement generalized beyond the particular construction used for data augmentation, suggesting that augmentation causes BERT to recruit abstract syntactic representations.

Thibault Sellam, Dipanjan Das, Ankur P. Parikh

Text generation has made significant advances in the last few years. Yet, evaluation metrics have lagged behind, as the most popular choices (e.g., BLEU and ROUGE) may correlate poorly with human judgments. We propose BLEURT, a learned evaluation metric based on BERT that can model human judgments with a few thousand possibly biased training examples. A key aspect of our approach is a novel pre-training scheme that uses millions of synthetic examples to help the model generalize. BLEURT provides state-of-the-art results on the last three years of the WMT Metrics shared task and the WebNLG Competition dataset. In contrast to a vanilla BERT-based approach, it yields superior results even when the training data is scarce and out-of-distribution.

Bhuwan Dhingra, Manaal Faruqui, Ankur Parikh et al.

Automatically constructed datasets for generating text from semi-structured data (tables), such as WikiBio, often contain reference texts that diverge from the information in the corresponding semi-structured data. We show that metrics which rely solely on the reference texts, such as BLEU and ROUGE, show poor correlation with human judgments when those references diverge. We propose a new metric, PARENT, which aligns n-grams from the reference and generated texts to the semi-structured data before computing their precision and recall. Through a large scale human evaluation study of table-to-text models for WikiBio, we show that PARENT correlates with human judgments better than existing text generation metrics. We also adapt and evaluate the information extraction based evaluation proposed by Wiseman et al (2017), and show that PARENT has comparable correlation to it, while being easier to use. We show that PARENT is also applicable when the reference texts are elicited from humans using the data from the WebNLG challenge.

Ian Tenney, Patrick Xia, Berlin Chen et al.

Contextualized representation models such as ELMo (Peters et al., 2018a) and BERT (Devlin et al., 2018) have recently achieved state-of-the-art results on a diverse array of downstream NLP tasks. Building on recent token-level probing work, we introduce a novel edge probing task design and construct a broad suite of sub-sentence tasks derived from the traditional structured NLP pipeline. We probe word-level contextual representations from four recent models and investigate how they encode sentence structure across a range of syntactic, semantic, local, and long-range phenomena. We find that existing models trained on language modeling and translation produce strong representations for syntactic phenomena, but only offer comparably small improvements on semantic tasks over a non-contextual baseline.

Ian Tenney, Dipanjan Das, Ellie Pavlick

Pre-trained text encoders have rapidly advanced the state of the art on many NLP tasks. We focus on one such model, BERT, and aim to quantify where linguistic information is captured within the network. We find that the model represents the steps of the traditional NLP pipeline in an interpretable and localizable way, and that the regions responsible for each step appear in the expected sequence: POS tagging, parsing, NER, semantic roles, then coreference. Qualitative analysis reveals that the model can and often does adjust this pipeline dynamically, revising lower-level decisions on the basis of disambiguating information from higher-level representations.

Hao Peng, Ankur P. Parikh, Manaal Faruqui et al.

We propose a novel conditioned text generation model. It draws inspiration from traditional template-based text generation techniques, where the source provides the content (i.e., what to say), and the template influences how to say it. Building on the successful encoder-decoder paradigm, it first encodes the content representation from the given input text; to produce the output, it retrieves exemplar text from the training data as "soft templates," which are then used to construct an exemplar-specific decoder. We evaluate the proposed model on abstractive text summarization and data-to-text generation. Empirical results show that this model achieves strong performance and outperforms comparable baselines.

Dipanjan Das, Ratul Ghosh, Brojeshwar Bhowmick

Despite significant advances in clustering methods in recent years, the outcome of clustering of a natural image dataset is still unsatisfactory due to two important drawbacks. Firstly, clustering of images needs a good feature representation of an image and secondly, we need a robust method which can discriminate these features for making them belonging to different clusters such that intra-class variance is less and inter-class variance is high. Often these two aspects are dealt with independently and thus the features are not sufficient enough to partition the data meaningfully. In this paper, we propose a method where we discover these features required for the separation of the images using deep autoencoder. Our method learns the image representation features automatically for the purpose of clustering and also select a coherent image and an incoherent image simultaneously for a given image so that the feature representation learning can learn better discriminative features for grouping the similar images in a cluster and at the same time separating the dissimilar images across clusters. Experiment results show that our method produces significantly better result than the state-of-the-art methods and we also show that our method is more generalized across different dataset without using any pre-trained model like other existing methods.

Vignesh Prasad, Dipanjan Das, Brojeshwar Bhowmick

Deep approaches to predict monocular depth and ego-motion have grown in recent years due to their ability to produce dense depth from monocular images. The main idea behind them is to optimize the photometric consistency over image sequences by warping one view into another, similar to direct visual odometry methods. One major drawback is that these methods infer depth from a single view, which might not effectively capture the relation between pixels. Moreover, simply minimizing the photometric loss does not ensure proper pixel correspondences, which is a key factor for accurate depth and pose estimations. In contrast, we propose a 2-view depth network to infer the scene depth from consecutive frames, thereby learning inter-pixel relationships. To ensure better correspondences, thereby better geometric understanding, we propose incorporating epipolar constraints to make the learning more geometrically sound. We use the Essential matrix obtained using Nist'er's Five Point Algorithm, to enforce meaningful geometric constraints, rather than using it as training labels. This allows us to use lesser no. of trainable parameters compared to state-of-the-art methods. The proposed method results in better depth images and pose estimates, which capture the scene structure and motion in a better way. Such a geometrically constrained learning performs successfully even in cases where simply minimizing the photometric error would fail.

Jan A. Botha, Manaal Faruqui, John Alex et al.

Split and rephrase is the task of breaking down a sentence into shorter ones that together convey the same meaning. We extract a rich new dataset for this task by mining Wikipedia's edit history: WikiSplit contains one million naturally occurring sentence rewrites, providing sixty times more distinct split examples and a ninety times larger vocabulary than the WebSplit corpus introduced by Narayan et al. (2017) as a benchmark for this task. Incorporating WikiSplit as training data produces a model with qualitatively better predictions that score 32 BLEU points above the prior best result on the WebSplit benchmark.

Manaal Faruqui, Ellie Pavlick, Ian Tenney et al.

We release a corpus of 43 million atomic edits across 8 languages. These edits are mined from Wikipedia edit history and consist of instances in which a human editor has inserted a single contiguous phrase into, or deleted a single contiguous phrase from, an existing sentence. We use the collected data to show that the language generated during editing differs from the language that we observe in standard corpora, and that models trained on edits encode different aspects of semantics and discourse than models trained on raw, unstructured text. We release the full corpus as a resource to aid ongoing research in semantics, discourse, and representation learning.

Manaal Faruqui, Dipanjan Das

Understanding search queries is a hard problem as it involves dealing with "word salad" text ubiquitously issued by users. However, if a query resembles a well-formed question, a natural language processing pipeline is able to perform more accurate interpretation, thus reducing downstream compounding errors. Hence, identifying whether or not a query is well formed can enhance query understanding. Here, we introduce a new task of identifying a well-formed natural language question. We construct and release a dataset of 25,100 publicly available questions classified into well-formed and non-wellformed categories and report an accuracy of 70.7% on the test set. We also show that our classifier can be used to improve the performance of neural sequence-to-sequence models for generating questions for reading comprehension.

Gaurav Singh Tomar, Thyago Duque, Oscar Täckström et al.

We present a solution to the problem of paraphrase identification of questions. We focus on a recent dataset of question pairs annotated with binary paraphrase labels and show that a variant of the decomposable attention model (Parikh et al., 2016) results in accurate performance on this task, while being far simpler than many competing neural architectures. Furthermore, when the model is pretrained on a noisy dataset of automatically collected question paraphrases, it obtains the best reported performance on the dataset.

Kenton Lee, Shimi Salant, Tom Kwiatkowski et al.

The reading comprehension task, that asks questions about a given evidence document, is a central problem in natural language understanding. Recent formulations of this task have typically focused on answer selection from a set of candidates pre-defined manually or through the use of an external NLP pipeline. However, Rajpurkar et al. (2016) recently released the SQuAD dataset in which the answers can be arbitrary strings from the supplied text. In this paper, we focus on this answer extraction task, presenting a novel model architecture that efficiently builds fixed length representations of all spans in the evidence document with a recurrent network. We show that scoring explicit span representations significantly improves performance over other approaches that factor the prediction into separate predictions about words or start and end markers. Our approach improves upon the best published results of Wang & Jiang (2016) by 5% and decreases the error of Rajpurkar et al.'s baseline by > 50%.

Ankur P. Parikh, Oscar Täckström, Dipanjan Das et al.

We propose a simple neural architecture for natural language inference. Our approach uses attention to decompose the problem into subproblems that can be solved separately, thus making it trivially parallelizable. On the Stanford Natural Language Inference (SNLI) dataset, we obtain state-of-the-art results with almost an order of magnitude fewer parameters than previous work and without relying on any word-order information. Adding intra-sentence attention that takes a minimum amount of order into account yields further improvements.