Psychological research consistently finds that human ratings of words across diverse semantic scales can be reduced to a low-dimensional form with relatively little information loss. We find that the semantic associations encoded in the embedding matrices of large language models (LLMs) exhibit a similar structure. We show that the projections of words on semantic directions defined by antonym pairs (e.g. kind - cruel) correlate highly with human ratings, and further find that these projections effectively reduce to a 3-dimensional subspace within LLM embeddings, closely resembling the patterns derived from human survey responses. Moreover, we find that shifting tokens along one semantic direction causes off-target effects on geometrically aligned features proportional to their cosine similarity. These findings suggest that semantic features are entangled within LLMs similarly to how they are interconnected in human language, and a great deal of semantic information, despite its apparent complexity, is surprisingly low-dimensional. Furthermore, accounting for this semantic structure may prove essential for avoiding unintended consequences when steering features.

Semantic dimensions of sound have been playing a central role in understanding the nature of auditory sensory experience as well as the broader relation between perception, language, and meaning. Accordingly, and given the recent proliferation of large language models (LLMs), here we asked whether such models exhibit an organisation of perceptual semantics similar to those observed in humans. Specifically, we prompted ChatGPT, a chatbot based on a state-of-the-art LLM, to rate musical instrument sounds on a set of 20 semantic scales. We elicited multiple responses in separate chats, analogous to having multiple human raters. ChatGPT generated semantic profiles that only partially correlated with human ratings, yet showed robust agreement along well-known psychophysical dimensions of musical sounds such as brightness (bright-dark) and pitch height (deep-high). Exploratory factor analysis suggested the same dimensionality but different spatial configuration of a latent factor space between the chatbot and human ratings. Unexpectedly, the chatbot showed degrees of internal variability that were comparable in magnitude to that of human ratings. Our work highlights the potential of LLMs to capture salient dimensions of human sensory experience.

Model bias triggered by long-tailed data has been widely studied. However, measure based on the number of samples cannot explicate three phenomena simultaneously: (1) Given enough data, the classification performance gain is marginal with additional samples. (2) Classification performance decays precipitously as the number of training samples decreases when there is insufficient data. (3) Model trained on sample-balanced datasets still has different biases for different classes. In this work, we define and quantify the semantic scale of classes, which is used to measure the feature diversity of classes. It is exciting to find experimentally that there is a marginal effect of semantic scale, which perfectly describes the first two phenomena. Further, the quantitative measurement of semantic scale imbalance is proposed, which can accurately reflect model bias on multiple datasets, even on sample-balanced data, revealing a novel perspective for the study of class imbalance. Due to the prevalence of semantic scale imbalance, we propose semantic-scale-balanced learning, including a general loss improvement scheme and a dynamic re-weighting training framework that overcomes the challenge of calculating semantic scales in real-time during iterations. Comprehensive experiments show that dynamic semantic-scale-balanced learning consistently enables the model to perform superiorly on large-scale long-tailed and non-long-tailed natural and medical datasets, which is a good starting point for mitigating the prevalent but unnoticed model bias.