Can LLMs Really Understand Item Difficulty Levels? Implications for Automated Item Generation Using LLMs
Summary
This paper investigates whether LLMs can accurately predict item difficulty levels in large-scale reading and writing tests, finding that GPT-4.1 achieves moderate accuracy but is outperformed by ConvBERT, and that LLMs tend to underestimate difficulty for hard items.
View Cached Full Text
Cached at: 08/03/26, 07:31 AM
# Can LLMs Really Understand Item Difficulty Levels? Implications for Automated Item Generation Using LLMs Source: [https://arxiv.org/abs/2607.28634](https://arxiv.org/abs/2607.28634) [View PDF](https://arxiv.org/pdf/2607.28634) > Abstract:The estimation of item difficulty plays a key role in both formative assessment and large\-scale high\-stakes summative assessments\. This study explores how large language models \(LLMs\) perform in predicting item difficulty levels using items from a large\-scale Reading and Writing test\. The study investigated various prompting strategies and parameter settings across multiple LLMs\. LLM performance was compared with encoder\-only language models and feature\-based supervised machine learning models\. Zero\-shot GPT\-4\.1 with a temperature of 0 yielded the highest item difficulty level prediction accuracy, with a quadratic weighted kappa \(QWK\) of 0\.578\. However, LLMs' prediction accuracy was lower than that of ConvBERT \(QWK = 0\.625\), which outperformed the best feature\-based supervised machine learning model\. Further analysis showed that all LLMs struggled to label hard items; in particular, the current advanced GPT\-5\.4 tended to underestimate item difficulty levels\. Dimension reduction of embeddings showed that item embeddings from different difficulty levels were mixed together, indicating that semantic information from items alone is likely insufficient for item difficulty level prediction\. The findings suggest that if LLMs cannot understand item difficulty levels as evidenced by empirical data and tend to treat most items as easy when their own capabilities increase, caution should be exercised when using LLMs to generate items with targeted difficulty levels\. ## Submission history From: Xinyi Wang \[[view email](https://arxiv.org/show-email/0174468d/2607.28634)\] **\[v1\]**Sun, 17 May 2026 10:17:02 UTC \(867 KB\)
Similar Articles
LLMs Struggle to Measure What Distinguishes Students of Different Proficiency Levels: A Study of Item Discrimination in Reading Comprehension Assessment
This paper evaluates 42 large language models on their ability to measure item discrimination in reading comprehension assessments, finding weak alignment with human-calibrated measures and highlighting it as an open challenge for psychometric evaluation.
The Easy Trap: Why LLMs Underestimate Misconception-Driven Difficulty
This paper investigates how LLM-generated difficulty ratings for math items align with actual student performance, finding that LLMs systematically underestimate difficulty for items driven by learner misconceptions, a phenomenon termed the 'Easy Trap'.
Distinguishing Artificial from Authentic: Evaluating LLMs for Detecting LLM-Generated Content
This paper investigates how well LLMs can detect their own generated content in educational contexts, finding that detection accuracy varies by task type and is unreliable for short-answer questions.
Cognitive Episodes in LLM Reasoning Traces Enable Interpretable Human Item Difficulty Prediction
Introduces Epi2Diff, a framework that maps LLM reasoning traces into cognitive episodes to predict human item difficulty, outperforming baselines and providing interpretable process evidence.
Can LLMs Judge Better Than They Generate? Evaluating Task Asymmetry, Mechanistic Interpretability and Transferability for In-Context QA
This paper tests the assumption that LLMs judge better than they generate in in-context QA, finding generation accuracy exceeds self-evaluation on most benchmarks, with evaluation attending less to context. The findings challenge core assumptions in self-evaluation pipelines.