Transformer architectures have achieved great success in solving natural language tasks, which learn strong language representations from large-scale unlabeled texts. In this paper, we seek to go further beyond and explore a new logical inductive bias for better language representation learning. Logic reasoning is known as a formal methodology to reach answers from given knowledge and facts. Inspired by such a view, we develop a novel neural architecture named FOLNet (First-Order Logic Network), to encode this new inductive bias. We construct a set of neural logic operators as learnable Horn clauses, which are further forward-chained into a fully differentiable neural architecture (FOLNet). Interestingly, we find that the self-attention module in transformers can be composed by two of our neural logic operators, which probably explains their strong reasoning performance. Our proposed FOLNet has the same input and output interfaces as other pretrained models and thus could be pretrained/finetuned by using similar losses. It also allows FOLNet to be used in a plug-and-play manner when replacing other pretrained models. With our logical inductive bias, the same set of ``logic deduction skills'' learned through pretraining are expected to be equally capable of solving diverse downstream tasks. For this reason, FOLNet learns language representations that have much stronger transfer capabilities. Experimental results on several language understanding tasks show that our pretrained FOLNet model outperforms the existing strong transformer-based approaches.
There have been a lot of interest in the scaling properties of Transformer models. However, not much has been done on the front of investigating the effect of scaling properties of different inductive biases and model architectures. Do model architectures scale differently? If so, how does inductive bias affect scaling behaviour? How does this influence upstream (pretraining) and downstream (transfer)? This paper conducts a systematic study of scaling behaviour of ten diverse model architectures such as Transformers, Switch Transformers, Universal Transformers, Dynamic convolutions, Performers…
Vector Institute（向量研究所）；University of Toronto（多伦多大学）；Google Research（谷歌研究院）；University of Cambridge（剑桥大学）
While designing inductive bias in neural architectures has been widely studied, we hypothesize that transformer networks are flexible enough to learn inductive bias from suitable generic tasks. Here, we replace architecture engineering by encoding inductive bias in the form of datasets. Inspired by Peirce's view that deduction, induction, and abduction are the primitives of reasoning, we design three synthetic tasks that are intended to require the model to have these three abilities. We specifically design these tasks to be synthetic and devoid of mathematical knowledge to ensure that only th…
Aligning Inductive Bias for Data-Efficient Generalization in State Space Models
School of Computer Science, National Key Laboratory for Multimedia Information Processing, Peking University, China（计算机学院、多媒体信息处理国家重点实验室、北京大学）；School of Physics, Peking University, China（物理学院、北京大学）
Harvard University（哈佛大学）；Microsoft Research NYC（微软研究院纽约）
Self-attention, an architectural motif designed to model long-range interactions in sequential data, has driven numerous recent breakthroughs in natural language processing and beyond. This work provides a theoretical analysis of the inductive biases of self-attention modules. Our focus is to rigorously establish which functions and long-range dependencies self-attention blocks prefer to represent. Our main result shows that bounded-norm Transformer networks "create sparse variables": a single self-attention head can represent a sparse function of the input sequence, with sample complexity sca…
Learned optimizers -- neural networks that are trained to act as optimizers -- have the potential to dramatically accelerate training of machine learning models. However, even when meta-trained across thousands of tasks at huge computational expense, blackbox learned optimizers often struggle with stability and generalization when applied to tasks unlike those in their meta-training set. In this paper, we use tools from dynamical systems to investigate the inductive biases and stability properties of optimization algorithms, and apply the resulting insights to designing inductive biases for bl…
Structural Disentanglement in Bilinear MLPs via Architectural Inductive Bias
Department of Electrical Engineering, Indian Institute of Technology Roorkee (IIT Roorkee), India；Department of Material Science \& Engineering, IIT Roorkee, India；Mehta Family School of Data Science
Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance
Qwen Large Model Application Team, Alibaba（阿里巴巴大模型应用团队）；The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；Shenzhen Research Institute of Big Data（深圳大数据研究院）
From CNNs to attention mechanisms, encoding inductive biases into neural networks has been a fruitful source of improvement in machine learning. Adding auxiliary losses to the main objective function is a general way of encoding biases that can help networks learn better representations. However, since auxiliary losses are minimized only on training data, they suffer from the same generalization gap as regular task losses. Moreover, by adding a term to the loss function, the model optimizes a different objective than the one we care about. In this work we address both problems: first, we take …
Graph Inductive Biases in Transformers without Message Passing
McGill University（麦吉尔大学）；University of Oxford（牛津大学）；Massachusetts Institute of Technology（麻省理工学院）；MetaAI（Meta人工智能实验室）；Mila - Quebec AI Institute（米拉-魁北克人工智能研究所）；Canada CIFAR AI Chair（加拿大CIFAR人工智能讲席）；FiveAI（FiveAI公司）；International Laboratory on Learning Systems（国际学习系统实验室）
本文提出 GRIT，通过相对位置编码、节点对注意力和度信息注入，在无消息传递的情况下赋予 Graph Transformer 图归纳偏置，并取得理论表达力与 SOTA 性能。
Characterising the Inductive Biases of Neural Networks on Boolean Data
Department of Theoretical Physics, University of Oxford, UK（理论物理系，牛津大学）；Department of Theoretical Chemistry, University of Oxford, UK（理论化学系，牛津大学）；Green Templeton College, University of Oxford, UK（格林·泰尔顿学院，牛津大学）；Department of Computer Science, University of Oxford, UK（计算机科学系，牛津大学）
Current deep learning approaches have shown good in-distribution generalization performance, but struggle with out-of-distribution generalization. This is especially true in the case of tasks involving abstract relations like recognizing rules in sequences, as we find in many intelligence tests. Recent work has explored how forcing relational representations to remain distinct from sensory representations, as it seems to be the case in the brain, can help artificial systems. Building on this work, we further explore and formalize the advantages afforded by 'partitioned' representations of rela…
Attention is a general reasoning mechanism than can flexibly deal with image information, but its memory requirements had made it so far impractical for high resolution image generation. We present Grid Partitioned Attention (GPA), a new approximate attention algorithm that leverages a sparse inductive bias for higher computational and memory efficiency in image domains: queries attend only to few keys, spatially close queries attend to close keys due to correlations. Our paper introduces the new attention layer, analyzes its complexity and how the trade-off between memory usage and model powe…
Johns Hopkins University（约翰斯·霍普金斯大学）；University of California, Berkeley（加州大学伯克利分校）；Microsoft Research AI（微软人工智能研究院）；Princeton University（普林斯顿大学）
How do learners acquire languages from the limited data available to them? This process must involve some inductive biases - factors that affect how a learner generalizes - but it is unclear which inductive biases can explain observed patterns in language acquisition. To facilitate computational modeling aimed at addressing this question, we introduce a framework for giving particular linguistic inductive biases to a neural network model; such a model can then be used to empirically explore the effects of those inductive biases. This framework disentangles universal inductive biases, which are…
Learners that are exposed to the same training data might generalize differently due to differing inductive biases. In neural network models, inductive biases could in theory arise from any aspect of the model architecture. We investigate which architectural factors affect the generalization behavior of neural sequence-to-sequence models trained on two syntactic tasks, English question formation and English tense reinflection. For both tasks, the training set is consistent with a generalization based on hierarchical structure and a generalization based on linear order. All architectural factor…
University of Cambridge（剑桥大学）；ETH Zürich（苏黎世联邦理工学院）
Since language models are used to model a wide variety of languages, it is natural to ask whether the neural architectures used for the task have inductive biases towards modeling particular types of languages. Investigation of these biases has proved complicated due to the many variables that appear in the experimental setup. Languages vary in many typological dimensions, and it is difficult to single out one or two to investigate without the others acting as confounders. We propose a novel method for investigating the inductive biases of language models using artificial languages. These lang…
People use rich prior knowledge about the world in order to efficiently learn new concepts. These priors - also known as "inductive biases" - pertain to the space of internal models considered by a learner, and they help the learner make inferences that go beyond the observed data. A recent study found that deep neural networks optimized for object recognition develop the shape bias (Ritter et al., 2017), an inductive bias possessed by children that plays an important role in early word learning. However, these networks use unrealistically large quantities of training data, and the conditions …
Acoustics-to-word models are end-to-end speech recognizers that use words as targets without relying on pronunciation dictionaries or graphemes. These models are notoriously difficult to train due to the lack of linguistic knowledge. It is also unclear how the amount of training data impacts the optimization and generalization of such models. In this work, we study the optimization and generalization of acoustics-to-word models under different amounts of training data. In addition, we study three types of inductive bias, leveraging a pronunciation dictionary, word boundary annotations, and con…
We study how masking and predicting tokens in an unsupervised fashion can give rise to linguistic structures and downstream performance gains. Recent theories have suggested that pretrained language models acquire useful inductive biases through masks that implicitly act as cloze reductions for downstream tasks. While appealing, we show that the success of the random masking strategy used in practice cannot be explained by such cloze-like masks alone. We construct cloze-like masks using task-specific lexicons for three different classification datasets and show that the majority of pretrained …
Pre-trained contextual representations have led to dramatic performance improvements on a range of downstream tasks. Such performance improvements have motivated researchers to quantify and understand the linguistic information encoded in these representations. In general, researchers quantify the amount of linguistic information through probing, an endeavor which consists of training a supervised model to predict a linguistic property directly from the contextual representations. Unfortunately, this definition of probing has been subject to extensive criticism in the literature, and has been …
Tsinghua University（清华大学）；Mila（米拉研究所）；Université de Montréal（蒙特利尔大学）；CIFAR AI Chair（CIFAR人工智能讲席）
Various forms of regularization in learning tasks strive for different notions of simplicity. This paper presents a spectral regularization technique, which attaches a unique inductive bias to sequence modeling based on an intuitive concept of simplicity defined in the Chomsky hierarchy. From fundamental connections between Hankel matrices and regular grammars, we propose to use the trace norm of the Hankel matrix, the tightest convex relaxation of its rank, as the spectral regularizer. To cope with the fact that the Hankel matrix is bi-infinite, we propose an unbiased stochastic estimator for…
End-to-end automatic speech recognition (ASR) commonly transcribes audio signals into sequences of characters while its performance is evaluated by measuring the word-error rate (WER). This suggests that predicting sequences of words directly may be helpful instead. However, training with word-level supervision can be more difficult due to the sparsity of examples per label class. In this paper we analyze an end-to-end ASR model that combines a word-and-character representation in a multi-task learning (MTL) framework. We show that it improves on the WER and study how the word-level model can …
University of Edinburgh（爱丁堡大学）；University of British Columbia（不列颠哥伦比亚大学）；University of Cambridge（剑桥大学）；DeepMind（深度思维）
Referential games and reconstruction games are the most common game types for studying emergent languages. We investigate how the type of the language game affects the emergent language in terms of: i) language compositionality and ii) transfer of an emergent language to a task different from its origin, which we refer to as language expressivity. With empirical experiments on a handcrafted symbolic dataset, we show that languages emerged from different games have different compositionality and further different expressivity.
University of Cambridge（剑桥大学）；MILA（蒙特利尔学习算法研究所）；McGill University（麦吉尔大学）；Polytechnique Montréal（蒙特利尔综合理工学院）；ServiceNow（ServiceNow公司）
In order to communicate, humans flatten a complex representation of ideas and their attributes into a single word or a sentence. We investigate the impact of representation learning in artificial agents by developing graph referential games. We empirically show that agents parametrized by graph neural networks develop a more compositional language compared to bag-of-words and sequence models, which allows them to systematically generalize to new combinations of familiar features.
Shaping Shared Languages: Human and Large Language Models' Inductive Biases in Emergent Communication
Leiden Institute of Advanced Computer Science, Leiden University, Netherlands（莱顿高级计算机科学研究所，莱顿大学，荷兰）；School of Computing, University of Kent, United Kingdom（计算学院，肯特大学，英国）；Cognitive Psychology Unit, Institute of Psychology, Leiden, the Netherlands（认知心理学单元，心理学研究所，莱顿，荷兰）
Graph Alignment Topology as an Inductive Bias for Grounding Detection
Department of Pediatrics, University of Illinois College of Medicine Peoria（伊利诺伊大学皮奥里亚医学院儿科部）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Siebel School of Computing and Data Science, University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校计算机与数据科学学院）；Carle Illinois College of Medicine, University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校卡莱医学院）
Large natural language models (such as GPT-3 or T5) demonstrate impressive abilities across a range of general NLP tasks. Here, we show that the knowledge embedded in such models provides a useful inductive bias, not just on traditional NLP tasks, but also in the nontraditional task of training a symbolic reasoning engine. We observe that these engines learn quickly and generalize in a natural way that reflects human intuition. For example, training such a system to model block-stacking might naturally generalize to stacking other types of objects because of structure in the real world that ha…
