Do Large Language Models Mirror Cognitive Language Processing?

2024-02-28

arXiv.org

AI Summary

Understanding the neural alignment between large language models (LLMs) and human language processing remains a fundamental challenge in cognitive neuroscience and AI. Method: We systematically investigated representational alignment between 23 state-of-the-art LLMs and fMRI-measured neural responses in human language cortex using representational similarity analysis (RSA). Contribution/Results: We provide the first causal characterization of how pretraining data scale, model scaling, alignment-specific fine-tuning, and prompt design affect LLM–brain correspondence. Pretraining scale and supervised alignment training significantly enhance alignment; explicit semantic prompts improve cross-subject consistency, whereas noisy prompts degrade it; and benchmark scores (e.g., MMLU) strongly correlate with neural alignment (r > 0.8). These findings establish neural alignment as an objective, quantifiable metric for assessing the cognitive interpretability of LLMs, offering a principled evaluation framework for brain-inspired AI development.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Large Language Models (LLMs) have demonstrated remarkable abilities in text comprehension and logical reasoning, indicating that the text representations learned by LLMs can facilitate their language processing capabilities. In neuroscience, brain cognitive processing signals are typically utilized to study human language processing. Therefore, it is natural to ask how well the text embeddings from LLMs align with the brain cognitive processing signals, and how training strategies affect the LLM-brain alignment? In this paper, we employ Representational Similarity Analysis (RSA) to measure the alignment between 23 mainstream LLMs and fMRI signals of the brain to evaluate how effectively LLMs simulate cognitive language processing. We empirically investigate the impact of various factors (e.g., pre-training data size, model scaling, alignment training, and prompts) on such LLM-brain alignment. Experimental results indicate that pre-training data size and model scaling are positively correlated with LLM-brain similarity, and alignment training can significantly improve LLM-brain similarity. Explicit prompts contribute to the consistency of LLMs with brain cognitive language processing, while nonsensical noisy prompts may attenuate such alignment. Additionally, the performance of a wide range of LLM evaluations (e.g., MMLU, Chatbot Arena) is highly correlated with the LLM-brain similarity.

Problem

• Large Language Models
• Neural Similarity
• Training Methods

Innovation

• Representational Similarity Analysis (RSA)
• Large Language Models (LLMs) vs. Human Brain
• Training Parameters Impact