Recent studies have shown that as Transformer-based language models become larger and are trained on very large amounts of data, the fit of their surprisal estimates to naturalistic human reading times degrades. The current work presents a series of analyses showing that word frequency is a key explanatory factor underlying these two trends. First, residual errors from four language model families on four corpora show that the inverse correlation between model size and fit to reading times is the strongest on the subset of least frequent words, which is driven by excessively accurate predictions of larger model variants. Additionally, training dynamics reveal that during later training steps, all model variants learn to predict rare words and that larger model variants do so more accurately, which explains the detrimental effect of both training data amount and model size on fit to reading times. Finally, a feature attribution analysis demonstrates that larger model variants are able to accurately predict rare words based on both an effectively longer context window size as well as stronger local associations compared to smaller model variants. Taken together, these results indicate that Transformer-based language models' surprisal estimates diverge from human-like expectations due to the superhumanly complex associations they learn for predicting rare words.
This work presents a detailed linguistic analysis into why larger Transformer-based pre-trained language models with more parameters and lower perplexity nonetheless yield surprisal estimates that are less predictive of human reading times. First, regression analyses show a strictly monotonic, positive log-linear relationship between perplexity and fit to reading times for the more recently released five GPT-Neo variants and eight OPT variants on two separate datasets, replicating earlier results limited to just GPT-2 (Oh et al., 2022). Subsequently, analysis of residual errors reveals a syste…
Transformer-based large language models are trained to make predictions about the next word by aggregating representations of previous tokens through their self-attention mechanism. In the field of cognitive modeling, such attention patterns have recently been interpreted as embodying the process of cue-based retrieval, in which attention over multiple targets is taken to generate interference and latency during retrieval. Under this framework, this work first defines an entropy-based predictor that quantifies the diffuseness of self-attention, as well as distance-based predictors that capture…
Surprisal theory has provided a unifying framework for understanding many phenomena in sentence processing (Hale, 2001; Levy, 2008a), positing that a word's conditional probability given all prior context fully determines processing difficulty. Problematically for this claim, one local statistic, word frequency, has also been shown to affect processing, even when conditional probability given context is held constant. Here, we ask whether other local statistics have a role in processing, or whether word frequency is a special case. We present the first clear evidence that more complex local st…
Human reading behavior is sensitive to surprisal: more predictable words tend to be read faster. Unexpectedly, this applies not only to the surprisal of the word that is currently being read, but also to the surprisal of upcoming (successor) words that have not been fixated yet. This finding has been interpreted as evidence that readers can extract lexical information parafoveally. Calling this interpretation into question, Angele et al. (2015) showed that successor effects appear even in contexts in which those successor words are not yet visible. They hypothesized that successor surprisal pr…
Stanford University（斯坦福大学）；University of Edinburgh（爱丁堡大学）；University of Washington（华盛顿大学）；Harvard University（哈佛大学）
Intuitively, human readers cope easily with errors in text; typos, misspelling, word substitutions, etc. do not unduly disrupt natural reading. Previous work indicates that letter transpositions result in increased reading times, but it is unclear if this effect generalizes to more natural errors. In this paper, we report an eye-tracking study that compares two error types (letter transpositions and naturally occurring misspelling) and two error rates (10% or 50% of all words contain errors). We find that human readers show unimpaired comprehension in spite of these errors, but error words cau…
When humans read text, they fixate some words and skip others. However, there have been few attempts to explain skipping behavior with computational models, as most existing work has focused on predicting reading times (e.g.,~using surprisal). In this paper, we propose a novel approach that models both skipping and reading, using an unsupervised architecture that combines a neural attention with autoencoding, trained on raw text using reinforcement learning. Our model explains human reading behavior as a tradeoff between precision of language understanding (encoding the input accurately) and e…
Multiple-choice Machine Reading Comprehension (MRC) is an important and challenging Natural Language Understanding (NLU) task, in which a machine must choose the answer to a question from a set of choices, with the question placed in context of text passages or dialog. In the last a couple of years the NLU field has been revolutionized with the advent of models based on the Transformer architecture, which are pretrained on massive amounts of unsupervised data and then fine-tuned for various supervised learning NLU tasks. Transformer models have come to dominate a wide variety of leader-boards …
Why are language models less surprised than humans? Testing the Parse Multiplicity Mismatch Hypothesis
Department of Linguistics, New York University（纽约大学语言学系）；Center for Data Science, New York University（纽约大学数据科学中心）；Department of Linguistics, University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校语言学系）
We advance a novel explanation of similarity-based interference effects in subject-verb and reflexive pronoun agreement processing, grounded in surprisal values computed from a pretrained large-scale Transformer model, GPT-2. Specifically, we show that surprisal of the verb or reflexive pronoun predicts facilitatory interference effects in ungrammatical sentences, where a distractor noun that matches in number with the verb or pronoun leads to faster reading times, despite the distractor not participating in the agreement relation. We review the human empirical evidence for such effects, inclu…
Can LLMs Augment Low-Resource Reading Comprehension Datasets? Opportunities and Challenges
University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Carnegie Mellon University（卡内基梅隆大学）；Georgia Institute of Technology（佐治亚理工学院）；Stanford University（斯坦福大学）；Amazon AI（亚马逊人工智能）
To Know or Not To Know? Analyzing Self-Consistency of Large Language Models under Ambiguity
LMU Munich（慕尼黑大学）；Faculty of Philological and Cultural Studies, University of Vienna（维也纳大学语言与文化学院）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；UniVie Doctoral School Computer Science（维也纳大学计算机博士学院）
By positing a relationship between naturalistic reading times and information-theoretic surprisal, surprisal theory (Hale, 2001; Levy, 2008) provides a natural interface between language models and psycholinguistic models. This paper re-evaluates a claim due to Goodkind and Bicknell (2018) that a language model's ability to model reading times is a linear function of its perplexity. By extending Goodkind and Bicknell's analysis to modern neural architectures, we show that the proposed relation does not always hold for Long Short-Term Memory networks, Transformers, and pre-trained models. We in…
SiLLM: Large Language Models for Simultaneous Machine Translation
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；University of Chinese Academy of Sciences（中国科学院大学）；Soochow University（苏州大学）
Human-like fleeting memory improves language learning but impairs reading time prediction in transformer language models
University of Amsterdam, Amsterdam Brain and Cognition（阿姆斯特丹大学，阿姆斯特丹脑与认知中心）；Vrije Universiteit Amsterdam, Department of Informatics（阿姆斯特丹自由大学，信息学院）；Max Planck Institute for Psycholinguistics（马克斯·普朗克心理学语言学研究所）
Structure leads and dominates comprehension in naturalistic reading
Department of Brain and Cognitive Sciences, University of Rochester（罗切斯特大学脑科学与认知科学系）；Department of Linguistics and Modern Languages, the Chinese University of Hong Kong（香港中文大学语言学与现代语言系）；Department of Language Science, University of California Irvine（加州大学 Irvine 分校语言科学系）
