Zero-1-to-3: Domain-level Zero-shot Cognitive Diagnosis via One Batch of Early-bird Students towards Three Diagnostic Objectives

This repository contains the implementation for the paper titled Zero-1-to-3: Domain-level Zero-shot Cognitive Diagnosis via One Batch of Early-bird Students towards Three Diagnostic Objectives, published at AAAI 2024 [Paper].

Authors: Weibo Gao, Qi Liu, Hao Wang, et al.

Email: weibogao@mail.ustc.edu.cn

Environment Settings

We use PyTorch as the backend.

• Torch version: '1.7.1'

Datasets

[1] Dbe-kt22: A knowledge tracing dataset based on online student evaluation. arXiv'22
[2] XES3G5M: a knowledge tracing benchmark dataset with auxiliary information. NeurIPS'23

Running

1. Select a model for running, e.g., Zero-NCDM

    CD Zero-NCDM
   

2. Pre-training and testing the model, in multiple source domains

   python train.py
   

3. Fine-tuning the model using early-bird students' logs in the target domain:

   python fine_tune.py
   

4. Generating the simulated logs

   python generate_coll_data.py
   

5. Fine-tuning using the simulated data

   python fine_tune_step_2.py
   

Related Works

• RCD: Relation Map Driven Cognitive Diagnosis for Intelligent Education Systems (SIGIR'2021) [Paper] [Code] [Presentation Video]

• Leveraging Transferable Knowledge Concept Graph Embedding for Cold-Start Cognitive Diagnosis (SIGIR'2023) [Paper] [Code] [Presentation Video]

• FedJudge: Federated Legal Large Language Model [Paper] [Code]

Last Update Date: March 14, 2024