Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Prompt Transfer for Dual-Aspect Cross Domain Cognitive Diagnosis

Fei Liu, Yizhong Zhang, Shuochen Liu, Shengwei Ji, Kui Yu, Le Wu

TL;DR

This paper tackles dual-aspect cross-domain cognitive diagnosis (CDCD) by proposing PromptCD, a soft-prompt transfer framework that enables robust knowledge transfer across student- and exercise-aspect CDCD. It introduces two instantiations, PromptCD-S and PromptCD-E, and a two-stage training regime (pre-training on source domains, fine-tuning on few-shot target data) to achieve scenario-agnostic adaptation. Extensive experiments on real-world datasets show significant improvements over traditional cognitive diagnosis models and existing CDCD methods across AUC, RMSE, and other metrics, with the Ours+ variant offering the strongest performance. The approach also includes prompt-to-representation mapping, visualization analyses, and a demonstration of personalized recommendations, highlighting its practical value for guiding learning and exercise selection.

Abstract

Cognitive Diagnosis (CD) aims to evaluate students' cognitive states based on their interaction data, enabling downstream applications such as exercise recommendation and personalized learning guidance. However, existing methods often struggle with accuracy drops in cross-domain cognitive diagnosis (CDCD), a practical yet challenging task. While some efforts have explored exercise-aspect CDCD, such as crosssubject scenarios, they fail to address the broader dual-aspect nature of CDCD, encompassing both student- and exerciseaspect variations. This diversity creates significant challenges in developing a scenario-agnostic framework. To address these gaps, we propose PromptCD, a simple yet effective framework that leverages soft prompt transfer for cognitive diagnosis. PromptCD is designed to adapt seamlessly across diverse CDCD scenarios, introducing PromptCD-S for student-aspect CDCD and PromptCD-E for exercise-aspect CDCD. Extensive experiments on real-world datasets demonstrate the robustness and effectiveness of PromptCD, consistently achieving superior performance across various CDCD scenarios. Our work offers a unified and generalizable approach to CDCD, advancing both theoretical and practical understanding in this critical domain. The implementation of our framework is publicly available at https://github.com/Publisher-PromptCD/PromptCD.

Prompt Transfer for Dual-Aspect Cross Domain Cognitive Diagnosis

TL;DR

This paper tackles dual-aspect cross-domain cognitive diagnosis (CDCD) by proposing PromptCD, a soft-prompt transfer framework that enables robust knowledge transfer across student- and exercise-aspect CDCD. It introduces two instantiations, PromptCD-S and PromptCD-E, and a two-stage training regime (pre-training on source domains, fine-tuning on few-shot target data) to achieve scenario-agnostic adaptation. Extensive experiments on real-world datasets show significant improvements over traditional cognitive diagnosis models and existing CDCD methods across AUC, RMSE, and other metrics, with the Ours+ variant offering the strongest performance. The approach also includes prompt-to-representation mapping, visualization analyses, and a demonstration of personalized recommendations, highlighting its practical value for guiding learning and exercise selection.

Abstract

Cognitive Diagnosis (CD) aims to evaluate students' cognitive states based on their interaction data, enabling downstream applications such as exercise recommendation and personalized learning guidance. However, existing methods often struggle with accuracy drops in cross-domain cognitive diagnosis (CDCD), a practical yet challenging task. While some efforts have explored exercise-aspect CDCD, such as crosssubject scenarios, they fail to address the broader dual-aspect nature of CDCD, encompassing both student- and exerciseaspect variations. This diversity creates significant challenges in developing a scenario-agnostic framework. To address these gaps, we propose PromptCD, a simple yet effective framework that leverages soft prompt transfer for cognitive diagnosis. PromptCD is designed to adapt seamlessly across diverse CDCD scenarios, introducing PromptCD-S for student-aspect CDCD and PromptCD-E for exercise-aspect CDCD. Extensive experiments on real-world datasets demonstrate the robustness and effectiveness of PromptCD, consistently achieving superior performance across various CDCD scenarios. Our work offers a unified and generalizable approach to CDCD, advancing both theoretical and practical understanding in this critical domain. The implementation of our framework is publicly available at https://github.com/Publisher-PromptCD/PromptCD.

Paper Structure

This paper contains 25 sections, 7 equations, 8 figures, 7 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Cognitive Diagnosis
  4. Cross-Domain Cognitive Diagnosis (CDCD)
  5. Proposed Framework
  6. Overall Architecture
  7. Source Prompt Enhancement
  8. Source-To-Target Transfer
  9. Proposed Instantiations
  10. Student-Aspect CDCD: PromptCD-S
  11. Exercise-Aspect CDCD: PromptCD-E
  12. Experiments
  13. Experimental Settings
  14. Datasets
  15. Baselines
  16. ...and 10 more sections

Figures (8)

  • Figure 1: Performance comparison of MIRT in in-domain scenarios A and B versus cross-domain scenarios C and D.
  • Figure 2: Traditional cognitive diagnosis versus cross-domain cognitive diagnosis.
  • Figure 3: Illustration of overlapping and non-overlapping sets in CDCD.
  • Figure 4: Overall architecture of the proposed PromptCD framework, including the pre-training and fine-tuning stages.
  • Figure 5: The average score distribution across different schools and subjects in the SLP dataset
  • ...and 3 more figures

Theorems & Definitions (2)

  • Definition 1: Overlapping Set
  • Definition 2: Non-Overlapping Set