SCOP: Evaluating the Comprehension Process of Large Language Models from a Cognitive View

2025-06-05

AI Summary

Existing LLM understanding evaluations predominantly focus on output correctness while neglecting alignment between model reasoning processes and human expert cognition, undermining assessment reliability. Method: We propose SCOP, the first cognitive-process-oriented evaluation framework for language understanding, grounded in cognitive psychology. SCOP defines five core understanding competencies, constructs a rigorous, multi-dimensional bias-aware test suite, and integrates human annotation with controllable generation techniques. Contribution/Results: Through fine-grained behavioral analysis and attribution-based diagnosis, we conduct the first systematic empirical study across major open- and closed-source LLMs. Our findings reveal a critical unreliability phenomenon—“correct answers with flawed reasoning”—where models frequently outperform humans locally yet fail to achieve expert-level process consistency. SCOP establishes both a theoretical foundation and an empirical benchmark for understanding-driven model training and evaluation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Despite the great potential of large language models(LLMs) in machine comprehension, it is still disturbing to fully count on them in real-world scenarios. This is probably because there is no rational explanation for whether the comprehension process of LLMs is aligned with that of experts. In this paper, we propose SCOP to carefully examine how LLMs perform during the comprehension process from a cognitive view. Specifically, it is equipped with a systematical definition of five requisite skills during the comprehension process, a strict framework to construct testing data for these skills, and a detailed analysis of advanced open-sourced and closed-sourced LLMs using the testing data. With SCOP, we find that it is still challenging for LLMs to perform an expert-level comprehension process. Even so, we notice that LLMs share some similarities with experts, e.g., performing better at comprehending local information than global information. Further analysis reveals that LLMs can be somewhat unreliable -- they might reach correct answers through flawed comprehension processes. Based on SCOP, we suggest that one direction for improving LLMs is to focus more on the comprehension process, ensuring all comprehension skills are thoroughly developed during training.

Problem

• Evaluating LLM comprehension alignment with expert cognitive processes
• Assessing LLM performance on five requisite comprehension skills
• Identifying unreliable LLM reasoning despite correct answers

Innovation

• Systematic definition of five comprehension skills
• Strict framework for testing data construction
• Detailed analysis of open and closed-source LLMs