Search

Human readers can accurately count how many letters are in a word (e.g., 7 in ``buffalo''), remove a letter from a given position (e.g., ``bufflo'') or add a new one. The human brain of readers must have therefore learned to disentangle information related to the position of a letter and its identity. Such disentanglement is necessary for the compositional, unbounded, ability of humans to create and parse new strings, with any combination of letters appearing in any positions. Do modern deep neural models also possess this crucial compositional ability? Here, we tested whether neural models that achieve state-of-the-art on disentanglement of features in visual input can also disentangle letter position and letter identity when trained on images of written words. Specifically, we trained beta variational autoencoder ($\beta$-VAE) to reconstruct images of letter strings and evaluated their disentanglement performance using CompOrth - a new benchmark that we created for studying compositional learning and zero-shot generalization in visual models for orthography. The benchmark suggests a set of tests, of increasing complexity, to evaluate the degree of disentanglement between orthographic features of written words in deep neural models. Using CompOrth, we conducted a set of experiments to analyze the generalization ability of these models, in particular, to unseen word length and to unseen combinations of letter identities and letter positions. We found that while models effectively disentangle surface features, such as horizontal and vertical `retinal' locations of words within an image, they dramatically fail to disentangle letter position and letter identity and lack any notion of word length. Together, this study demonstrates the shortcomings of state-of-the-art $\beta$-VAE models compared to humans and proposes a new challenge and a corresponding benchmark to evaluate neural models.

Originally formalized with symbolic representations, syntactic trees may also be effectively represented in the activations of large language models (LLMs). Indeed, a 'Structural Probe' can find a subspace of neural activations, where syntactically related words are relatively close to one-another. However, this syntactic code remains incomplete: the distance between the Structural Probe word embeddings can represent the existence but not the type and direction of syntactic relations. Here, we hypothesize that syntactic relations are, in fact, coded by the relative direction between nearby embeddings. To test this hypothesis, we introduce a 'Polar Probe' trained to read syntactic relations from both the distance and the direction between word embeddings. Our approach reveals three main findings. First, our Polar Probe successfully recovers the type and direction of syntactic relations, and substantially outperforms the Structural Probe by nearly two folds. Second, we confirm that this polar coordinate system exists in a low-dimensional subspace of the intermediate layers of many LLMs and becomes increasingly precise in the latest frontier models. Third, we demonstrate with a new benchmark that similar syntactic relations are coded similarly across the nested levels of syntactic trees. Overall, this work shows that LLMs spontaneously learn a geometry of neural activations that explicitly represents the main symbolic structures of linguistic theory.

Do architectural differences significantly affect the way models represent and process language? We propose a new approach, based on metric-learning encoding models (MLEMs), as a first step to answer this question. The approach provides a feature-based comparison of how any two layers of any two models represent linguistic information. We apply the method to BERT, GPT-2 and Mamba. Unlike previous methods, MLEMs offer a transparent comparison, by identifying the specific linguistic features responsible for similarities and differences. More generally, the method uses formal, symbolic descriptions of a domain, and use these to compare neural representations. As such, the approach can straightforwardly be extended to other domains, such as speech and vision, and to other neural systems, including human brains., Comment: 7 pages, 6 figures

We introduce Metric-Learning Encoding Models (MLEMs) as a new approach to understand how neural systems represent the theoretical features of the objects they process. As a proof-of-concept, we apply MLEMs to neural representations extracted from BERT, and track a wide variety of linguistic features (e.g., tense, subject person, clause type, clause embedding). We find that: (1) linguistic features are ordered: they separate representations of sentences to different degrees in different layers; (2) neural representations are organized hierarchically: in some layers, we find clusters of representations nested within larger clusters, following successively important linguistic features; (3) linguistic features are disentangled in middle layers: distinct, selective units are activated by distinct linguistic features. Methodologically, MLEMs are superior (4) to multivariate decoding methods, being more robust to type-I errors, and (5) to univariate encoding methods, in being able to predict both local and distributed representations. Together, this demonstrates the utility of Metric-Learning Encoding Methods for studying how linguistic features are neurally encoded in language models and the advantage of MLEMs over traditional methods. MLEMs can be extended to other domains (e.g. vision) and to other neural systems, such as the human brain., Comment: 17 pages, 13 figures

During language acquisition, children follow a typical sequence of learning stages, whereby they first learn to categorize phonemes before they develop their lexicon and eventually master increasingly complex syntactic structures. However, the computational principles that lead to this learning trajectory remain largely unknown. To investigate this, we here compare the learning trajectories of deep language models to those of children. Specifically, we test whether, during its training, GPT-2 exhibits stages of language acquisition comparable to those observed in children aged between 18 months and 6 years. For this, we train 48 GPT-2 models from scratch and evaluate their syntactic and semantic abilities at each training step, using 96 probes curated from the BLiMP, Zorro and BIG-Bench benchmarks. We then compare these evaluations with the behavior of 54 children during language production. Our analyses reveal three main findings. First, similarly to children, the language models tend to learn linguistic skills in a systematic order. Second, this learning scheme is parallel: the language tasks that are learned last improve from the very first training steps. Third, some - but not all - learning stages are shared between children and these language models. Overall, these results shed new light on the principles of language acquisition, and highlight important divergences in how humans and modern algorithms learn to process natural language., Comment: Accepted to ACL 2023. *Equal Contribution

Two fundamental questions in neurolinguistics concerns the brain regions that integrate information beyond the lexical level, and the size of their window of integration. To address these questions we introduce a new approach named masked-attention generation. It uses GPT-2 transformers to generate word embeddings that capture a fixed amount of contextual information. We then tested whether these embeddings could predict fMRI brain activity in humans listening to naturalistic text. The results showed that most of the cortex within the language network is sensitive to contextual information, and that the right hemisphere is more sensitive to longer contexts than the left. Masked-attention generation supports previous analyses of context-sensitivity in the brain, and complements them by quantifying the window size of context integration per voxel., Comment: 2 pages, 2 figures, CCN 2023

A fundamental question in neurolinguistics concerns the brain regions involved in syntactic and semantic processing during speech comprehension, both at the lexical (word processing) and supra-lexical levels (sentence and discourse processing). To what extent are these regions separated or intertwined? To address this question, we trained a lexical language model, Glove, and a supra-lexical language model, GPT-2, on a text corpus from which we selectively removed either syntactic or semantic information. We then assessed to what extent these information-restricted models were able to predict the time-courses of fMRI signal of humans listening to naturalistic text. We also manipulated the size of contextual information provided to GPT-2 in order to determine the windows of integration of brain regions involved in supra-lexical processing. Our analyses show that, while most brain regions involved in language are sensitive to both syntactic and semantic variables, the relative magnitudes of these effects vary a lot across these regions. Furthermore, we found an asymmetry between the left and right hemispheres, with semantic and syntactic processing being more dissociated in the left hemisphere than in the right, and the left and right hemispheres showing respectively greater sensitivity to short and long contexts. The use of information-restricted NLP models thus shed new light on the spatial organization of syntactic processing, semantic processing and compositionality., Comment: 19 pages, 8 figures, 10 pages of Appendix, 5 appendix figures

Neural Language Models (NLMs) have made tremendous advances during the last years, achieving impressive performance on various linguistic tasks. Capitalizing on this, studies in neuroscience have started to use NLMs to study neural activity in the human brain during language processing. However, many questions remain unanswered regarding which factors determine the ability of a neural language model to capture brain activity (aka its 'brain score'). Here, we make first steps in this direction and examine the impact of test loss, training corpus and model architecture (comparing GloVe, LSTM, GPT-2 and BERT), on the prediction of functional Magnetic Resonance Imaging timecourses of participants listening to an audiobook. We find that (1) untrained versions of each model already explain significant amount of signal in the brain by capturing similarity in brain responses across identical words, with the untrained LSTM outperforming the transformerbased models, being less impacted by the effect of context; (2) that training NLP models improves brain scores in the same brain regions irrespective of the model's architecture; (3) that Perplexity (test loss) is not a good predictor of brain score; (4) that training data have a strong influence on the outcome and, notably, that off-the-shelf models may lack statistical power to detect brain activations. Overall, we outline the impact of modeltraining choices, and suggest good practices for future studies aiming at explaining the human language system using neural language models.

Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit "breakthrough" behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting., Comment: 27 pages, 17 figures + references and appendices, repo: https://github.com/google/BIG-bench

Recursive processing is considered a hallmark of human linguistic abilities. A recent study evaluated recursive processing in recurrent neural language models (RNN-LMs) and showed that such models perform below chance level on embedded dependencies within nested constructions -- a prototypical example of recursion in natural language. Here, we study if state-of-the-art Transformer LMs do any better. We test four different Transformer LMs on two different types of nested constructions, which differ in whether the embedded (inner) dependency is short or long range. We find that Transformers achieve near-perfect performance on short-range embedded dependencies, significantly better than previous results reported for RNN-LMs and humans. However, on long-range embedded dependencies, Transformers' performance sharply drops below chance level. Remarkably, the addition of only three words to the embedded dependency caused Transformers to fall from near-perfect to below-chance performance. Taken together, our results reveal Transformers' shortcoming when it comes to recursive, structure-based, processing., Comment: None

One of the fundamental principles of contemporary linguistics states that language processing requires the ability to extract recursively nested tree structures. However, it remains unclear whether and how this code could be implemented in neural circuits. Recent advances in Recurrent Neural Networks (RNNs), which achieve near-human performance in some language tasks, provide a compelling model to address such questions. Here, we present a new framework to study recursive processing in RNNs, using subject-verb agreement as a probe into the representations of the neural network. We trained six distinct types of RNNs on a simplified probabilistic context-free grammar designed to independently manipulate the length of a sentence and the depth of its syntactic tree. All RNNs generalized to subject-verb dependencies longer than those seen during training. However, none systematically generalized to deeper tree structures, even those with a structural bias towards learning nested tree (i.e., stack-RNNs). In addition, our analyses revealed primacy and recency effects in the generalization patterns of LSTM-based models, showing that these models tend to perform well on the outer- and innermost parts of a center-embedded tree structure, but poorly on its middle levels. Finally, probing the internal states of the model during the processing of sentences with nested tree structures, we found a complex encoding of grammatical agreement information (e.g. grammatical number), in which all the information for multiple words nouns was carried by a single unit. Taken together, these results indicate how neural networks may extract bounded nested tree structures, without learning a systematic recursive rule.

Recursive processing in sentence comprehension is considered a hallmark of human linguistic abilities. However, its underlying neural mechanisms remain largely unknown. We studied whether a modern artificial neural network trained with "deep learning" methods mimics a central aspect of human sentence processing, namely the storing of grammatical number and gender information in working memory and its use in long-distance agreement (e.g., capturing the correct number agreement between subject and verb when they are separated by other phrases). Although the network, a recurrent architecture with Long Short-Term Memory units, was solely trained to predict the next word in a large corpus, analysis showed the emergence of a very sparse set of specialized units that successfully handled local and long-distance syntactic agreement for grammatical number. However, the simulations also showed that this mechanism does not support full recursion and fails with some long-range embedded dependencies. We tested the model's predictions in a behavioral experiment where humans detected violations in number agreement in sentences with systematic variations in the singular/plural status of multiple nouns, with or without embedding. Human and model error patterns were remarkably similar, showing that the model echoes various effects observed in human data. However, a key difference was that, with embedded long-range dependencies, humans remained above chance level, while the model's systematic errors brought it below chance. Overall, our study shows that exploring the ways in which modern artificial neural networks process sentences leads to precise and testable hypotheses about human linguistic performance.

Recent work has shown that LSTMs trained on a generic language modeling objective capture syntax-sensitive generalizations such as long-distance number agreement. We have however no mechanistic understanding of how they accomplish this remarkable feat. Some have conjectured it depends on heuristics that do not truly take hierarchical structure into account. We present here a detailed study of the inner mechanics of number tracking in LSTMs at the single neuron level. We discover that long-distance number information is largely managed by two `number units'. Importantly, the behaviour of these units is partially controlled by other units independently shown to track syntactic structure. We conclude that LSTMs are, to some extent, implementing genuinely syntactic processing mechanisms, paving the way to a more general understanding of grammatical encoding in LSTMs., Comment: To appear in Proceedings of NAACL, Minneapolis, MN, 2019

Metric functions for phoneme perception capture the similarity structure among phonemes in a given language and therefore play a central role in phonology and psycho-linguistics. Various phenomena depend on phoneme similarity, such as spoken word recognition or serial recall from verbal working memory. This study presents a new framework for learning a metric function for perceptual distances among pairs of phonemes. Previous studies have proposed various metric functions, from simple measures counting the number of phonetic dimensions that two phonemes share (place-, manner-of-articulation and voicing), to more sophisticated ones such as deriving perceptual distances based on the number of natural classes that both phonemes belong to. However, previous studies have manually constructed the metric function, which may lead to unsatisfactory account of the empirical data. This study presents a framework to derive the metric function from behavioral data on phoneme perception using learning algorithms. We first show that this approach outperforms previous metrics suggested in the literature in predicting perceptual distances among phoneme pairs. We then study several metric functions derived by the learning algorithms and show how perceptual saliencies of phonological features can be derived from them. For English, we show that the derived perceptual saliencies are in accordance with a previously described order among phonological features and show how the framework extends the results to more features. Finally, we explore how the metric function and perceptual saliencies of phonological features may vary across languages. To this end, we compare results based on two English datasets and a new dataset that we have collected for Hebrew.