---
dataset_info:
  config_name: embeddings
  features:
  - name: response_index
    dtype: int64
  - name: embedding
    struct:
    - name: text
      sequence: float64
  splits:
  - name: train
    num_bytes: 100791911220
    num_examples: 3074799
  download_size: 23826024498
  dataset_size: 100791911220
configs:
- config_name: embeddings
  data_files:
  - split: train
    path: embeddings/train-*
license: cc-by-4.0
language:
- en
- vi
tags:
- coding
- computer-science
pretty_name: codeinsight
size_categories:
- 10K<n<100K
---

The data is collected from students in the Department of Computer Science at the VNU-HCM University of Technology (Vietnam) in the 2023 and 2024 academic years. We collect data in two courses, Data Structure and Algorithm (DSA, Fall semester) and Programming Fundamental (PF, Spring semester). Most students are in Programming Fundamentals in their first year and Data Structure in their second year. DSA has PF as a prerequisite, and PF has Introduction to Computing as a prerequisite. The dataset comprises interaction data from 781 students participating in a six-week programming course, divided into three main groups. The largest group, ``L,'' includes 572 students from 15 classes. These are regular students with typical academic backgrounds. The second group, ``CC,'' comprises 184 students from three classes who entered the program with lower entrance scores and paid higher tuition fees. Finally, the ``DT'' group consists of 25 students who are either retaking the course due to prior failure or to improve their previous scores. Throughout the course, students engage in weekly programming assignments that involve a variety of topics, starting with object-oriented programming (OOP), recursion, array lists, and singly linked lists in the first week. The subsequent weeks introduce more advanced data structures and algorithms, including doubly linked lists, stacks, queues, sorting algorithms, binary trees, AVL trees, and search algorithms, culminating in topics like hash functions and graph theory by the final week. Each week begins with an exam covering the previous week’s material. The dataset captures student actions while working on programming exercises in an interactive coding environment. The actions logged include starting an attempt, prechecking, saving, checking, and finishing attempts. Prechecks allow students to test their code against a public set of test cases, providing feedback on whether their code compiles and passes basic tests. Students see the input, output, and feedback on their code's performance. In contrast, submissions are evaluated against a private set of test cases that assess the robustness of their solutions, with more challenging test cases and potential penalties for multiple submissions. Here, students only see a numerical evaluation score.