Phil Blunsom

Aishwarya Agrawal, Ivana Kajić, Emanuele Bugliarello et al. · deepmind

Vision-and-language (V&L) models pretrained on large-scale multimodal data have demonstrated strong performance on various tasks such as image captioning and visual question answering (VQA). The quality of such models is commonly assessed by measuring their performance on unseen data that typically comes from the same distribution as the training data. However, when evaluated under out-of-distribution (out-of-dataset) settings for VQA, we observe that these models exhibit poor generalization. We comprehensively evaluate two pretrained V&L models under different settings (i.e. classification and open-ended text generation) by conducting cross-dataset evaluations. We find that these models tend to learn to solve the benchmark, rather than learning the high-level skills required by the VQA task. We also find that in most cases generative models are less susceptible to shifts in data distribution compared to discriminative ones, and that multimodal pretraining is generally helpful for OOD generalization. Finally, we revisit assumptions underlying the use of automatic VQA evaluation metrics, and empirically show that their stringent nature repeatedly penalizes models for correct responses.

Satwik Bhattamishra, Arkil Patel, Varun Kanade et al. · mila

Despite the widespread success of Transformers on NLP tasks, recent works have found that they struggle to model several formal languages when compared to recurrent models. This raises the question of why Transformers perform well in practice and whether they have any properties that enable them to generalize better than recurrent models. In this work, we conduct an extensive empirical study on Boolean functions to demonstrate the following: (i) Random Transformers are relatively more biased towards functions of low sensitivity. (ii) When trained on Boolean functions, both Transformers and LSTMs prioritize learning functions of low sensitivity, with Transformers ultimately converging to functions of lower sensitivity. (iii) On sparse Boolean functions which have low sensitivity, we find that Transformers generalize near perfectly even in the presence of noisy labels whereas LSTMs overfit and achieve poor generalization accuracy. Overall, our results provide strong quantifiable evidence that suggests differences in the inductive biases of Transformers and recurrent models which may help explain Transformer's effective generalization performance despite relatively limited expressiveness.

Arkil Patel, Satwik Bhattamishra, Phil Blunsom et al. · mila

Compositional generalization is a fundamental trait in humans, allowing us to effortlessly combine known phrases to form novel sentences. Recent works have claimed that standard seq-to-seq models severely lack the ability to compositionally generalize. In this paper, we focus on one-shot primitive generalization as introduced by the popular SCAN benchmark. We demonstrate that modifying the training distribution in simple and intuitive ways enables standard seq-to-seq models to achieve near-perfect generalization performance, thereby showing that their compositional generalization abilities were previously underestimated. We perform detailed empirical analysis of this phenomenon. Our results indicate that the generalization performance of models is highly sensitive to the characteristics of the training data which should be carefully considered while designing such benchmarks in future.

Adam Liška, Tomáš Kočiský, Elena Gribovskaya et al.

Knowledge and language understanding of models evaluated through question answering (QA) has been usually studied on static snapshots of knowledge, like Wikipedia. However, our world is dynamic, evolves over time, and our models' knowledge becomes outdated. To study how semi-parametric QA models and their underlying parametric language models (LMs) adapt to evolving knowledge, we construct a new large-scale dataset, StreamingQA, with human written and generated questions asked on a given date, to be answered from 14 years of time-stamped news articles. We evaluate our models quarterly as they read new articles not seen in pre-training. We show that parametric models can be updated without full retraining, while avoiding catastrophic forgetting. For semi-parametric models, adding new articles into the search space allows for rapid adaptation, however, models with an outdated underlying LM under-perform those with a retrained LM. For questions about higher-frequency named entities, parametric updates are particularly beneficial. In our dynamic world, the StreamingQA dataset enables a more realistic evaluation of QA models, and our experiments highlight several promising directions for future research.

Laurent Sartran, Samuel Barrett, Adhiguna Kuncoro et al.

We introduce Transformer Grammars (TGs), a novel class of Transformer language models that combine (i) the expressive power, scalability, and strong performance of Transformers and (ii) recursive syntactic compositions, which here are implemented through a special attention mask and deterministic transformation of the linearized tree. We find that TGs outperform various strong baselines on sentence-level language modeling perplexity, as well as on multiple syntax-sensitive language modeling evaluation metrics. Additionally, we find that the recursive syntactic composition bottleneck which represents each sentence as a single vector harms perplexity on document-level language modeling, providing evidence that a different kind of memory mechanism -- one that is independent of composed syntactic representations -- plays an important role in current successful models of long text.

Satwik Bhattamishra, Arkil Patel, Phil Blunsom et al. · mila

In order to understand the in-context learning phenomenon, recent works have adopted a stylized experimental framework and demonstrated that Transformers can learn gradient-based learning algorithms for various classes of real-valued functions. However, the limitations of Transformers in implementing learning algorithms, and their ability to learn other forms of algorithms are not well understood. Additionally, the degree to which these capabilities are confined to attention-based models is unclear. Furthermore, it remains to be seen whether the insights derived from these stylized settings can be extrapolated to pretrained Large Language Models (LLMs). In this work, we take a step towards answering these questions by demonstrating the following: (a) On a test-bed with a variety of Boolean function classes, we find that Transformers can nearly match the optimal learning algorithm for 'simpler' tasks, while their performance deteriorates on more 'complex' tasks. Additionally, we find that certain attention-free models perform (almost) identically to Transformers on a range of tasks. (b) When provided a teaching sequence, i.e. a set of examples that uniquely identifies a function in a class, we show that Transformers learn more sample-efficiently. Interestingly, our results show that Transformers can learn to implement two distinct algorithms to solve a single task, and can adaptively select the more sample-efficient algorithm depending on the sequence of in-context examples. (c) Lastly, we show that extant LLMs, e.g. LLaMA-2, GPT-4, can compete with nearest-neighbor baselines on prediction tasks that are guaranteed to not be in their training set.

Qi Liu, Zihuiwen Ye, Tao Yu et al.

The task of context-dependent text-to-SQL aims to convert multi-turn user utterances to formal SQL queries. This is a challenging task due to both the scarcity of training data from which to learn complex contextual dependencies and to generalize to unseen databases. In this paper we explore augmenting the training datasets using self-play, which leverages contextual information to synthesize new interactions to adapt the model to new databases. We first design a SQL-to-text model conditioned on a sampled goal query, which represents a user's intent, that then converses with a text-to-SQL semantic parser to generate new interactions. We then filter the synthesized interactions and retrain the models with the augmented data. We find that self-play improves the accuracy of a strong baseline on SParC and CoSQL, two widely used cross-domain text-to-SQL datasets. Our analysis shows that self-play simulates various conversational thematic relations, enhances cross-domain generalization and improves beam-search.

Ahmet Üstün, Viraat Aryabumi, Zheng-Xin Yong et al.

Recent breakthroughs in large language models (LLMs) have centered around a handful of data-rich languages. What does it take to broaden access to breakthroughs beyond first-class citizen languages? Our work introduces Aya, a massively multilingual generative language model that follows instructions in 101 languages of which over 50% are considered as lower-resourced. Aya outperforms mT0 and BLOOMZ on the majority of tasks while covering double the number of languages. We introduce extensive new evaluation suites that broaden the state-of-art for multilingual eval across 99 languages -- including discriminative and generative tasks, human evaluation, and simulated win rates that cover both held-out tasks and in-distribution performance. Furthermore, we conduct detailed investigations on the optimal finetuning mixture composition, data pruning, as well as the toxicity, bias, and safety of our models. We open-source our instruction datasets and our model at https://hf.co/CohereForAI/aya-101

Kyle Duffy, Satwik Bhattamishra, Phil Blunsom

Large-scale pre-training has made progress in many fields of natural language processing, though little is understood about the design of pre-training datasets. We propose a methodology for obtaining a quantitative understanding of structural overlap between machine translation tasks. We apply our methodology to the natural language to Bash semantic parsing task (NLBash) and show that it is largely reducible to lexical alignment. We also find that there is strong structural overlap between NLBash and natural language to SQL. Additionally, we perform a study varying compute expended during pre-training on the English to German machine translation task and find that more compute expended during pre-training does not always correspond semantic representations with stronger transfer to NLBash.

Tom Hosking, Phil Blunsom, Max Bartolo

Human feedback has become the de facto standard for evaluating the performance of Large Language Models, and is increasingly being used as a training objective. However, it is not clear which properties of a generated output this single `preference' score captures. We hypothesise that preference scores are subjective and open to undesirable biases. We critically analyse the use of human feedback for both training and evaluation, to verify whether it fully captures a range of crucial error criteria. We find that while preference scores have fairly good coverage, they under-represent important aspects like factuality. We further hypothesise that both preference scores and error annotation may be affected by confounders, and leverage instruction-tuned models to generate outputs that vary along two possible confounding dimensions: assertiveness and complexity. We find that the assertiveness of an output skews the perceived rate of factuality errors, indicating that human annotations are not a fully reliable evaluation metric or training objective. Finally, we offer preliminary evidence that using human feedback as a training objective disproportionately increases the assertiveness of model outputs. We encourage future work to carefully consider whether preference scores are well aligned with the desired objective.

John Dang, Shivalika Singh, Daniel D'souza et al.

We introduce the Aya Expanse model family, a new generation of 8B and 32B parameter multilingual language models, aiming to address the critical challenge of developing highly performant multilingual models that match or surpass the capabilities of monolingual models. By leveraging several years of research at Cohere For AI and Cohere, including advancements in data arbitrage, multilingual preference training, and model merging, Aya Expanse sets a new state-of-the-art in multilingual performance. Our evaluations on the Arena-Hard-Auto dataset, translated into 23 languages, demonstrate that Aya Expanse 8B and 32B outperform leading open-weight models in their respective parameter classes, including Gemma 2, Qwen 2.5, and Llama 3.1, achieving up to a 76.6% win-rate. Notably, Aya Expanse 32B outperforms Llama 3.1 70B, a model with twice as many parameters, achieving a 54.0% win-rate. In this short technical report, we present extended evaluation results for the Aya Expanse model family and release their open-weights, together with a new multilingual evaluation dataset m-ArenaHard.

Saurabh Dash, Yiyang Nan, John Dang et al.

Building multimodal language models is fundamentally challenging: it requires aligning vision and language modalities, curating high-quality instruction data, and avoiding the degradation of existing text-only capabilities once vision is introduced. These difficulties are further magnified in the multilingual setting, where the need for multimodal data in different languages exacerbates existing data scarcity, machine translation often distorts meaning, and catastrophic forgetting is more pronounced. To address the aforementioned challenges, we introduce novel techniques spanning both data and modeling. First, we develop a synthetic annotation framework that curates high-quality, diverse multilingual multimodal instruction data, enabling Aya Vision models to produce natural, human-preferred responses to multimodal inputs across many languages. Complementing this, we propose a cross-modal model merging technique that mitigates catastrophic forgetting, effectively preserving text-only capabilities while simultaneously enhancing multimodal generative performance. Aya-Vision-8B achieves best-in-class performance compared to strong multimodal models such as Qwen-2.5-VL-7B, Pixtral-12B, and even much larger Llama-3.2-90B-Vision. We further scale this approach with Aya-Vision-32B, which outperforms models more than twice its size, such as Molmo-72B and LLaMA-3.2-90B-Vision. Our work advances multilingual progress on the multi-modal frontier, and provides insights into techniques that effectively bend the need for compute while delivering extremely high performance.

Zihuiwen Ye, Luckeciano Carvalho Melo, Younesse Kaddar et al.

Complex multi-step reasoning tasks, such as solving mathematical problems, remain challenging for large language models (LLMs). While outcome supervision is commonly used, process supervision via process reward models (PRMs) provides intermediate rewards to verify step-wise correctness in solution traces. However, as proxies for human judgement, PRMs suffer from reliability issues, including susceptibility to reward hacking. In this work, we propose leveraging uncertainty quantification (UQ) to enhance the reliability of step-wise verification with generative reward models for mathematical reasoning tasks. We introduce CoT Entropy, a novel UQ method that outperforms existing approaches in quantifying a PRM's uncertainty in step-wise verification. Our results demonstrate that incorporating uncertainty estimates improves the robustness of judge-LM PRMs, leading to more reliable verification.

Satwik Bhattamishra, Phil Blunsom, Varun Kanade

We study the learnability of languages in the Next Symbol Prediction (NSP) setting, where a learner receives only positive examples from a language together with, for every prefix, (i) whether the prefix itself is in the language and (ii) which next symbols can lead to an accepting string. This setting has been used in prior works to empirically analyze neural sequence models, and additionally, we observe that efficient algorithms for the NSP setting can be used to learn the (truncated) support of language models. We formalize the setting so as to make it amenable to PAC-learning analysis. While the setting provides a much richer set of labels than the conventional classification setting, we show that learning concept classes such as DFAs and Boolean formulas remains computationally hard. The proof is via a construction that makes almost all additional labels uninformative, yielding a reduction from the conventional learning problem to learning with NSP labels. Under cryptographic assumptions, the reduction implies that the problem of learning DFAs is computationally hard in the NSP setting.

Satwik Bhattamishra, Michael Hahn, Phil Blunsom et al.

Transformer architectures have been widely adopted in foundation models. Due to their high inference costs, there is renewed interest in exploring the potential of efficient recurrent architectures (RNNs). In this paper, we analyze the differences in the representational capabilities of Transformers and RNNs across several tasks of practical relevance, including index lookup, nearest neighbor, recognizing bounded Dyck languages, and string equality. For the tasks considered, our results show separations based on the size of the model required for different architectures. For example, we show that a one-layer Transformer of logarithmic width can perform index lookup, whereas an RNN requires a hidden state of linear size. Conversely, while constant-size RNNs can recognize bounded Dyck languages, we show that one-layer Transformers require a linear size for this task. Furthermore, we show that two-layer Transformers of logarithmic size can perform decision tasks such as string equality or disjointness, whereas both one-layer Transformers and recurrent models require linear size for these tasks. We also show that a log-size two-layer Transformer can implement the nearest neighbor algorithm in its forward pass; on the other hand recurrent models require linear size. Our constructions are based on the existence of $N$ nearly orthogonal vectors in $O(\log N)$ dimensional space and our lower bounds are based on reductions from communication complexity problems. We supplement our theoretical results with experiments that highlight the differences in the performance of these architectures on practical-size sequences.

Arash Ahmadian, Saurabh Dash, Hongyu Chen et al.

Emergent properties have been widely adopted as a term to describe behavior not present in smaller models but observed in larger models. Recent work suggests that the trade-off incurred by quantization is also an emergent property, with sharp drops in performance in models over 6B parameters. In this work, we ask "are quantization cliffs in performance solely a factor of scale?" Against a backdrop of increased research focus on why certain emergent properties surface at scale, this work provides a useful counter-example. We posit that it is possible to optimize for a quantization friendly training recipe that suppresses large activation magnitude outliers. Here, we find that outlier dimensions are not an inherent product of scale, but rather sensitive to the optimization conditions present during pre-training. This both opens up directions for more efficient quantization, and poses the question of whether other emergent properties are inherent or can be altered and conditioned by optimization and architecture design choices. We successfully quantize models ranging in size from 410M to 52B with minimal degradation in performance.

Qi Liu, Dani Yogatama, Phil Blunsom

We present a memory-augmented approach to condition an autoregressive language model on a knowledge graph. We represent the graph as a collection of relation triples and retrieve relevant relations for a given context to improve text generation. Experiments on WikiText-103, WMT19, and enwik8 English datasets demonstrate that our approach produces a better language model in terms of perplexity and bits per character. We also show that relational memory improves coherence, is complementary to token-based memory, and enables causal interventions. Our model provides a simple yet effective way to combine an autoregressive language model with a knowledge graph for a more coherent and logical generation.

Xiang Lorraine Li, Adhiguna Kuncoro, Jordan Hoffmann et al.

Language models (LMs) trained on large amounts of data have shown impressive performance on many NLP tasks under the zero-shot and few-shot setup. Here we aim to better understand the extent to which such models learn commonsense knowledge -- a critical component of many NLP applications. We conduct a systematic and rigorous zero-shot and few-shot commonsense evaluation of large pre-trained LMs, where we: (i) carefully control for the LMs' ability to exploit potential surface cues and annotation artefacts, and (ii) account for variations in performance that arise from factors that are not related to commonsense knowledge. Our findings highlight the limitations of pre-trained LMs in acquiring commonsense knowledge without task-specific supervision; furthermore, using larger models or few-shot evaluation are insufficient to achieve human-level commonsense performance.

Qi Liu, Lei Yu, Laura Rimell et al.

Direct decoding for task-oriented dialogue is known to suffer from the explaining-away effect, manifested in models that prefer short and generic responses. Here we argue for the use of Bayes' theorem to factorize the dialogue task into two models, the distribution of the context given the response, and the prior for the response itself. This approach, an instantiation of the noisy channel model, both mitigates the explaining-away effect and allows the principled incorporation of large pretrained models for the response prior. We present extensive experiments showing that a noisy channel model decodes better responses compared to direct decoding and that a two stage pretraining strategy, employing both open-domain and task-oriented dialogue data, improves over randomly initialized models.

Angeliki Lazaridou, Adhiguna Kuncoro, Elena Gribovskaya et al.

Our world is open-ended, non-stationary, and constantly evolving; thus what we talk about and how we talk about it change over time. This inherent dynamic nature of language contrasts with the current static language modelling paradigm, which trains and evaluates models on utterances from overlapping time periods. Despite impressive recent progress, we demonstrate that Transformer-XL language models perform worse in the realistic setup of predicting future utterances from beyond their training period, and that model performance becomes increasingly worse with time. We find that, while increasing model size alone -- a key driver behind recent progress -- does not solve this problem, having models that continually update their knowledge with new information can indeed mitigate this performance degradation over time. Hence, given the compilation of ever-larger language modelling datasets, combined with the growing list of language-model-based NLP applications that require up-to-date factual knowledge about the world, we argue that now is the right time to rethink the static way in which we currently train and evaluate our language models, and develop adaptive language models that can remain up-to-date with respect to our ever-changing and non-stationary world. We publicly release our dynamic, streaming language modelling benchmarks for WMT and arXiv to facilitate language model evaluation that takes temporal dynamics into account.

Gábor Melis, András György, Phil Blunsom

A common failure mode of density models trained as variational autoencoders is to model the data without relying on their latent variables, rendering these variables useless. Two contributing factors, the underspecification of the model and the looseness of the variational lower bound, have been studied separately in the literature. We weave these two strands of research together, specifically the tighter bounds of Monte-Carlo objectives and constraints on the mutual information between the observable and the latent variables. Estimating the mutual information as the average Kullback-Leibler divergence between the easily available variational posterior $q(z|x)$ and the prior does not work with Monte-Carlo objectives because $q(z|x)$ is no longer a direct approximation to the model's true posterior $p(z|x)$. Hence, we construct estimators of the Kullback-Leibler divergence of the true posterior from the prior by recycling samples used in the objective, with which we train models of continuous and discrete latents at much improved rate-distortion and no posterior collapse. While alleviated, the tradeoff between modelling the data and using the latents still remains, and we urge for evaluating inference methods across a range of mutual information values.

Oana-Maria Camburu, Eleonora Giunchiglia, Jakob Foerster et al.

For neural models to garner widespread public trust and ensure fairness, we must have human-intelligible explanations for their predictions. Recently, an increasing number of works focus on explaining the predictions of neural models in terms of the relevance of the input features. In this work, we show that feature-based explanations pose problems even for explaining trivial models. We show that, in certain cases, there exist at least two ground-truth feature-based explanations, and that, sometimes, neither of them is enough to provide a complete view of the decision-making process of the model. Moreover, we show that two popular classes of explainers, Shapley explainers and minimal sufficient subsets explainers, target fundamentally different types of ground-truth explanations, despite the apparently implicit assumption that explainers should look for one specific feature-based explanation. These findings bring an additional dimension to consider in both developing and choosing explainers.

Adhiguna Kuncoro, Lingpeng Kong, Daniel Fried et al.

Textual representation learners trained on large amounts of data have achieved notable success on downstream tasks; intriguingly, they have also performed well on challenging tests of syntactic competence. Given this success, it remains an open question whether scalable learners like BERT can become fully proficient in the syntax of natural language by virtue of data scale alone, or whether they still benefit from more explicit syntactic biases. To answer this question, we introduce a knowledge distillation strategy for injecting syntactic biases into BERT pretraining, by distilling the syntactically informative predictions of a hierarchical---albeit harder to scale---syntactic language model. Since BERT models masked words in bidirectional context, we propose to distill the approximate marginal distribution over words in context from the syntactic LM. Our approach reduces relative error by 2-21% on a diverse set of structured prediction tasks, although we obtain mixed results on the GLUE benchmark. Our findings demonstrate the benefits of syntactic biases, even in representation learners that exploit large amounts of data, and contribute to a better understanding of where syntactic biases are most helpful in benchmarks of natural language understanding.

Daniel Fried, Jean-Baptiste Alayrac, Phil Blunsom et al.

We apply a generative segmental model of task structure, guided by narration, to action segmentation in video. We focus on unsupervised and weakly-supervised settings where no action labels are known during training. Despite its simplicity, our model performs competitively with previous work on a dataset of naturalistic instructional videos. Our model allows us to vary the sources of supervision used in training, and we find that both task structure and narrative language provide large benefits in segmentation quality.

Gunnar A. Sigurdsson, Jean-Baptiste Alayrac, Aida Nematzadeh et al.

There are thousands of actively spoken languages on Earth, but a single visual world. Grounding in this visual world has the potential to bridge the gap between all these languages. Our goal is to use visual grounding to improve unsupervised word mapping between languages. The key idea is to establish a common visual representation between two languages by learning embeddings from unpaired instructional videos narrated in the native language. Given this shared embedding we demonstrate that (i) we can map words between the languages, particularly the 'visual' words; (ii) that the shared embedding provides a good initialization for existing unsupervised text-based word translation techniques, forming the basis for our proposed hybrid visual-text mapping algorithm, MUVE; and (iii) our approach achieves superior performance by addressing the shortcomings of text-based methods -- it is more robust, handles datasets with less commonality, and is applicable to low-resource languages. We apply these methods to translate words from English to French, Korean, and Japanese -- all without any parallel corpora and simply by watching many videos of people speaking while doing things.

Kazuya Kawakami, Luyu Wang, Chris Dyer et al.

Unsupervised speech representation learning has shown remarkable success at finding representations that correlate with phonetic structures and improve downstream speech recognition performance. However, most research has been focused on evaluating the representations in terms of their ability to improve the performance of speech recognition systems on read English (e.g. Wall Street Journal and LibriSpeech). This evaluation methodology overlooks two important desiderata that speech representations should have: robustness to domain shifts and transferability to other languages. In this paper we learn representations from up to 8000 hours of diverse and noisy speech data and evaluate the representations by looking at their robustness to domain shifts and their ability to improve recognition performance in many languages. We find that our representations confer significant robustness advantages to the resulting recognition systems: we see significant improvements in out-of-domain transfer relative to baseline feature sets and the features likewise provide improvements in 25 phonetically diverse languages including tonal languages and low-resource languages.

Oana-Maria Camburu, Brendan Shillingford, Pasquale Minervini et al.

To increase trust in artificial intelligence systems, a promising research direction consists of designing neural models capable of generating natural language explanations for their predictions. In this work, we show that such models are nonetheless prone to generating mutually inconsistent explanations, such as "Because there is a dog in the image" and "Because there is no dog in the [same] image", exposing flaws in either the decision-making process of the model or in the generation of the explanations. We introduce a simple yet effective adversarial framework for sanity checking models against the generation of inconsistent natural language explanations. Moreover, as part of the framework, we address the problem of adversarial attacks with full target sequences, a scenario that was not previously addressed in sequence-to-sequence attacks. Finally, we apply our framework on a state-of-the-art neural natural language inference model that provides natural language explanations for its predictions. Our framework shows that this model is capable of generating a significant number of inconsistent explanations.

Lei Yu, Laurent Sartran, Wojciech Stokowiec et al.

We show that Bayes' rule provides an effective mechanism for creating document translation models that can be learned from only parallel sentences and monolingual documents---a compelling benefit as parallel documents are not always available. In our formulation, the posterior probability of a candidate translation is the product of the unconditional (prior) probability of the candidate output document and the "reverse translation probability" of translating the candidate output back into the source language. Our proposed model uses a powerful autoregressive language model as the prior on target language documents, but it assumes that each sentence is translated independently from the target to the source language. Crucially, at test time, when a source document is observed, the document language model prior induces dependencies between the translations of the source sentences in the posterior. The model's independence assumption not only enables efficient use of available data, but it additionally admits a practical left-to-right beam-search algorithm for carrying out inference. Experiments show that our model benefits from using cross-sentence context in the language model, and it outperforms existing document translation approaches.

Chris Dyer, Gábor Melis, Phil Blunsom

A series of recent papers has used a parsing algorithm due to Shen et al. (2018) to recover phrase-structure trees based on proxies for "syntactic depth." These proxy depths are obtained from the representations learned by recurrent language models augmented with mechanisms that encourage the (unsupervised) discovery of hierarchical structure latent in natural language sentences. Using the same parser, we show that proxies derived from a conventional LSTM language model produce trees comparably well to the specialized architectures used in previous work. However, we also provide a detailed analysis of the parsing algorithm, showing (1) that it is incomplete---that is, it can recover only a fraction of possible trees---and (2) that it has a marked bias for right-branching structures which results in inflated performance in right-branching languages like English. Our analysis shows that evaluating with biased parsing algorithms can inflate the apparent structural competence of language models.

Gábor Melis, Tomáš Kočiský, Phil Blunsom

Many advances in Natural Language Processing have been based upon more expressive models for how inputs interact with the context in which they occur. Recurrent networks, which have enjoyed a modicum of success, still lack the generalization and systematicity ultimately required for modelling language. In this work, we propose an extension to the venerable Long Short-Term Memory in the form of mutual gating of the current input and the previous output. This mechanism affords the modelling of a richer space of interactions between inputs and their context. Equivalently, our model can be viewed as making the transition function given by the LSTM context-dependent. Experiments demonstrate markedly improved generalization on language modelling in the range of 3-4 perplexity points on Penn Treebank and Wikitext-2, and 0.01-0.05 bpc on four character-based datasets. We establish a new state of the art on all datasets with the exception of Enwik8, where we close a large gap between the LSTM and Transformer models.

Vid Kocijan, Oana-Maria Camburu, Ana-Maria Cretu et al.

Pronoun resolution is a major area of natural language understanding. However, large-scale training sets are still scarce, since manually labelling data is costly. In this work, we introduce WikiCREM (Wikipedia CoREferences Masked) a large-scale, yet accurate dataset of pronoun disambiguation instances. We use a language-model-based approach for pronoun resolution in combination with our WikiCREM dataset. We compare a series of models on a collection of diverse and challenging coreference resolution problems, where we match or outperform previous state-of-the-art approaches on 6 out of 7 datasets, such as GAP, DPR, WNLI, PDP, WinoBias, and WinoGender. We release our model to be used off-the-shelf for solving pronoun disambiguation.

Adhiguna Kuncoro, Chris Dyer, Laura Rimell et al.

Prior work has shown that, on small amounts of training data, syntactic neural language models learn structurally sensitive generalisations more successfully than sequential language models. However, their computational complexity renders scaling difficult, and it remains an open question whether structural biases are still necessary when sequential models have access to ever larger amounts of training data. To answer this question, we introduce an efficient knowledge distillation (KD) technique that transfers knowledge from a syntactic language model trained on a small corpus to an LSTM language model, hence enabling the LSTM to develop a more structurally sensitive representation of the larger training data it learns from. On targeted syntactic evaluations, we find that, while sequential LSTMs perform much better than previously reported, our proposed technique substantially improves on this baseline, yielding a new state of the art. Our findings and analysis affirm the importance of structural biases, even in models that learn from large amounts of data.

Dani Yogatama, Cyprien de Masson d'Autume, Jerome Connor et al.

We define general linguistic intelligence as the ability to reuse previously acquired knowledge about a language's lexicon, syntax, semantics, and pragmatic conventions to adapt to new tasks quickly. Using this definition, we analyze state-of-the-art natural language understanding models and conduct an extensive empirical investigation to evaluate them against these criteria through a series of experiments that assess the task-independence of the knowledge being acquired by the learning process. In addition to task performance, we propose a new evaluation metric based on an online encoding of the test data that quantifies how quickly an existing agent (model) learns a new task. Our results show that while the field has made impressive progress in terms of model architectures that generalize to many tasks, these models still require a lot of in-domain training examples (e.g., for fine tuning, training task-specific modules), and are prone to catastrophic forgetting. Moreover, we find that far from solving general tasks (e.g., document question answering), our models are overfitting to the quirks of particular datasets (e.g., SQuAD). We discuss missing components and conjecture on how to make progress toward general linguistic intelligence.

Oana-Maria Camburu, Tim Rocktäschel, Thomas Lukasiewicz et al.

In order for machine learning to garner widespread public adoption, models must be able to provide interpretable and robust explanations for their decisions, as well as learn from human-provided explanations at train time. In this work, we extend the Stanford Natural Language Inference dataset with an additional layer of human-annotated natural language explanations of the entailment relations. We further implement models that incorporate these explanations into their training process and output them at test time. We show how our corpus of explanations, which we call e-SNLI, can be used for various goals, such as obtaining full sentence justifications of a model's decisions, improving universal sentence representations and transferring to out-of-domain NLI datasets. Our dataset thus opens up a range of research directions for using natural language explanations, both for improving models and for asserting their trust.

Kazuya Kawakami, Chris Dyer, Phil Blunsom

We propose a segmental neural language model that combines the generalization power of neural networks with the ability to discover word-like units that are latent in unsegmented character sequences. In contrast to previous segmentation models that treat word segmentation as an isolated task, our model unifies word discovery, learning how words fit together to form sentences, and, by conditioning the model on visual context, how words' meanings ground in representations of non-linguistic modalities. Experiments show that the unconditional model learns predictive distributions better than character LSTM models, discovers words competitively with nonparametric Bayesian word segmentation models, and that modeling language conditional on visual context improves performance on both.

Changhao Chen, Yishu Miao, Chris Xiaoxuan Lu et al.

Inertial information processing plays a pivotal role in ego-motion awareness for mobile agents, as inertial measurements are entirely egocentric and not environment dependent. However, they are affected greatly by changes in sensor placement/orientation or motion dynamics, and it is infeasible to collect labelled data from every domain. To overcome the challenges of domain adaptation on long sensory sequences, we propose a novel framework that extracts domain-invariant features of raw sequences from arbitrary domains, and transforms to new domains without any paired data. Through the experiments, we demonstrate that it is able to efficiently and effectively convert the raw sequence from a new unlabelled target domain into an accurate inertial trajectory, benefiting from the physical motion knowledge transferred from the labelled source domain. We also conduct real-world experiments to show our framework can reconstruct physically meaningful trajectories from raw IMU measurements obtained with a standard mobile phone in various attachments.

Andrew Trask, Felix Hill, Scott Reed et al.

Neural networks can learn to represent and manipulate numerical information, but they seldom generalize well outside of the range of numerical values encountered during training. To encourage more systematic numerical extrapolation, we propose an architecture that represents numerical quantities as linear activations which are manipulated using primitive arithmetic operators, controlled by learned gates. We call this module a neural arithmetic logic unit (NALU), by analogy to the arithmetic logic unit in traditional processors. Experiments show that NALU-enhanced neural networks can learn to track time, perform arithmetic over images of numbers, translate numerical language into real-valued scalars, execute computer code, and count objects in images. In contrast to conventional architectures, we obtain substantially better generalization both inside and outside of the range of numerical values encountered during training, often extrapolating orders of magnitude beyond trained numerical ranges.

Tomáš Kočiský, Jonathan Schwarz, Phil Blunsom et al.

Reading comprehension (RC)---in contrast to information retrieval---requires integrating information and reasoning about events, entities, and their relations across a full document. Question answering is conventionally used to assess RC ability, in both artificial agents and children learning to read. However, existing RC datasets and tasks are dominated by questions that can be solved by selecting answers using superficial information (e.g., local context similarity or global term frequency); they thus fail to test for the essential integrative aspect of RC. To encourage progress on deeper comprehension of language, we present a new dataset and set of tasks in which the reader must answer questions about stories by reading entire books or movie scripts. These tasks are designed so that successfully answering their questions requires understanding the underlying narrative rather than relying on shallow pattern matching or salience. We show that although humans solve the tasks easily, standard RC models struggle on the tasks presented here. We provide an analysis of the dataset and the challenges it presents.

Felix Hill, Stephen Clark, Karl Moritz Hermann et al.

Neural network-based systems can now learn to locate the referents of words and phrases in images, answer questions about visual scenes, and execute symbolic instructions as first-person actors in partially-observable worlds. To achieve this so-called grounded language learning, models must overcome challenges that infants face when learning their first words. While it is notable that models with no meaningful prior knowledge overcome these obstacles, researchers currently lack a clear understanding of how they do so, a problem that we attempt to address in this paper. For maximum control and generality, we focus on a simple neural network-based language learning agent, trained via policy-gradient methods, which can interpret single-word instructions in a simulated 3D world. Whilst the goal is not to explicitly model infant word learning, we take inspiration from experimental paradigms in developmental psychology and apply some of these to the artificial agent, exploring the conditions under which established human biases and learning effects emerge. We further propose a novel method for visualising semantic representations in the agent.

Gábor Melis, Chris Dyer, Phil Blunsom

Ongoing innovations in recurrent neural network architectures have provided a steady influx of apparently state-of-the-art results on language modelling benchmarks. However, these have been evaluated using differing code bases and limited computational resources, which represent uncontrolled sources of experimental variation. We reevaluate several popular architectures and regularisation methods with large-scale automatic black-box hyperparameter tuning and arrive at the somewhat surprising conclusion that standard LSTM architectures, when properly regularised, outperform more recent models. We establish a new state of the art on the Penn Treebank and Wikitext-2 corpora, as well as strong baselines on the Hutter Prize dataset.

Yishu Miao, Edward Grefenstette, Phil Blunsom

Topic models have been widely explored as probabilistic generative models of documents. Traditional inference methods have sought closed-form derivations for updating the models, however as the expressiveness of these models grows, so does the difficulty of performing fast and accurate inference over their parameters. This paper presents alternative neural approaches to topic modelling by providing parameterisable distributions over topics which permit training by backpropagation in the framework of neural variational inference. In addition, with the help of a stick-breaking construction, we propose a recurrent network that is able to discover a notionally unbounded number of topics, analogous to Bayesian non-parametric topic models. Experimental results on the MXM Song Lyrics, 20NewsGroups and Reuters News datasets demonstrate the effectiveness and efficiency of these neural topic models.

Tsung-Hsien Wen, Yishu Miao, Phil Blunsom et al.

Developing a dialogue agent that is capable of making autonomous decisions and communicating by natural language is one of the long-term goals of machine learning research. Traditional approaches either rely on hand-crafting a small state-action set for applying reinforcement learning that is not scalable or constructing deterministic models for learning dialogue sentences that fail to capture natural conversational variability. In this paper, we propose a Latent Intention Dialogue Model (LIDM) that employs a discrete latent variable to learn underlying dialogue intentions in the framework of neural variational inference. In a goal-oriented dialogue scenario, these latent intentions can be interpreted as actions guiding the generation of machine responses, which can be further refined autonomously by reinforcement learning. The experimental evaluation of LIDM shows that the model out-performs published benchmarks for both corpus-based and human evaluation, demonstrating the effectiveness of discrete latent variable models for learning goal-oriented dialogues.

Wang Ling, Dani Yogatama, Chris Dyer et al.

Solving algebraic word problems requires executing a series of arithmetic operations---a program---to obtain a final answer. However, since programs can be arbitrarily complicated, inducing them directly from question-answer pairs is a formidable challenge. To make this task more feasible, we solve these problems by generating answer rationales, sequences of natural language and human-readable mathematical expressions that derive the final answer through a series of small steps. Although rationales do not explicitly specify programs, they provide a scaffolding for their structure via intermediate milestones. To evaluate our approach, we have created a new 100,000-sample dataset of questions, answers and rationales. Experimental results show that indirect supervision of program learning via answer rationales is a promising strategy for inducing arithmetic programs.

Jan Buys, Phil Blunsom

Parsing sentences to linguistically-expressive semantic representations is a key goal of Natural Language Processing. Yet statistical parsing has focused almost exclusively on bilexical dependencies or domain-specific logical forms. We propose a neural encoder-decoder transition-based parser which is the first full-coverage semantic graph parser for Minimal Recursion Semantics (MRS). The model architecture uses stack-based embedding features, predicting graphs jointly with unlexicalized predicates and their token alignments. Our parser is more accurate than attention-based baselines on MRS, and on an additional Abstract Meaning Representation (AMR) benchmark, and GPU batch processing makes it an order of magnitude faster than a high-precision grammar-based parser. Further, the 86.69% Smatch score of our MRS parser is higher than the upper-bound on AMR parsing, making MRS an attractive choice as a semantic representation.

Kazuya Kawakami, Chris Dyer, Phil Blunsom

Fixed-vocabulary language models fail to account for one of the most characteristic statistical facts of natural language: the frequent creation and reuse of new word types. Although character-level language models offer a partial solution in that they can create word types not attested in the training corpus, they do not capture the "bursty" distribution of such words. In this paper, we augment a hierarchical LSTM language model that generates sequences of word tokens character by character with a caching mechanism that learns to reuse previously generated words. To validate our model we construct a new open-vocabulary language modeling corpus (the Multilingual Wikipedia Corpus, MWC) from comparable Wikipedia articles in 7 typologically diverse languages and demonstrate the effectiveness of our model across this range of languages.

Dani Yogatama, Chris Dyer, Wang Ling et al.

We empirically characterize the performance of discriminative and generative LSTM models for text classification. We find that although RNN-based generative models are more powerful than their bag-of-words ancestors (e.g., they account for conditional dependencies across words in a document), they have higher asymptotic error rates than discriminatively trained RNN models. However we also find that generative models approach their asymptotic error rate more rapidly than their discriminative counterparts---the same pattern that Ng & Jordan (2001) proved holds for linear classification models that make more naive conditional independence assumptions. Building on this finding, we hypothesize that RNN-based generative classification models will be more robust to shifts in the data distribution. This hypothesis is confirmed in a series of experiments in zero-shot and continual learning settings that show that generative models substantially outperform discriminative models.

Dani Yogatama, Phil Blunsom, Chris Dyer et al.

We use reinforcement learning to learn tree-structured neural networks for computing representations of natural language sentences. In contrast with prior work on tree-structured models in which the trees are either provided as input or predicted using supervision from explicit treebank annotations, the tree structures in this work are optimized to improve performance on a downstream task. Experiments demonstrate the benefit of learning task-specific composition orders, outperforming both sequential encoders and recursive encoders based on treebank annotations. We analyze the induced trees and show that while they discover some linguistically intuitive structures (e.g., noun phrases, simple verb phrases), they are different than conventional English syntactic structures.

Zichao Yang, Phil Blunsom, Chris Dyer et al.

We propose a general class of language models that treat reference as an explicit stochastic latent variable. This architecture allows models to create mentions of entities and their attributes by accessing external databases (required by, e.g., dialogue generation and recipe generation) and internal state (required by, e.g. language models which are aware of coreference). This facilitates the incorporation of information that can be accessed in predictable locations in databases or discourse context, even when the targets of the reference may be rare words. Experiments on three tasks shows our model variants based on deterministic attention.

Tomáš Kočiský, Gábor Melis, Edward Grefenstette et al.

We present a novel semi-supervised approach for sequence transduction and apply it to semantic parsing. The unsupervised component is based on a generative model in which latent sentences generate the unpaired logical forms. We apply this method to a number of semantic parsing tasks focusing on domains with limited access to labelled training data and extend those datasets with synthetically generated logical forms.

Lei Yu, Jan Buys, Phil Blunsom

We introduce an online neural sequence to sequence model that learns to alternate between encoding and decoding segments of the input as it is read. By independently tracking the encoding and decoding representations our algorithm permits exact polynomial marginalization of the latent segmentation during training, and during decoding beam search is employed to find the best alignment path together with the predicted output sequence. Our model tackles the bottleneck of vanilla encoder-decoders that have to read and memorize the entire input sequence in their fixed-length hidden states before producing any output. It is different from previous attentive models in that, instead of treating the attention weights as output of a deterministic function, our model assigns attention weights to a sequential latent variable which can be marginalized out and permits online generation. Experiments on abstractive sentence summarization and morphological inflection show significant performance gains over the baseline encoder-decoders.