# QuesNet: A Unified Representation for Heterogeneous Test Questions

arXiv: https://arxiv.org/abs/1905.10949 · HTML: https://mlsanity.com/paper/1905.10949/

## Key facts

- Cited by 18 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 27 May 2019.

## Built on by (most cited first, 18 of 18)

- [Knowledge Tracing: A Survey](https://mlsanity.com/paper/2201.06953/) (2022)
- [A Survey of Knowledge Tracing: Models, Variants, and Applications](https://mlsanity.com/paper/2105.15106/) (2021)
- [Difficulty-Focused Contrastive Learning for Knowledge Tracing with a Large Language Model-Based Difficulty Prediction](https://mlsanity.com/paper/2312.11890/) (2023)
- [No Task Left Behind: Multi-Task Learning of Knowledge Tracing and Option Tracing for Better Student Assessment](https://mlsanity.com/paper/2204.14006/) (2022)
- [Automate Knowledge Concept Tagging on Math Questions with LLMs](https://mlsanity.com/paper/2403.17281/) (2024)
- [Survey of Computerized Adaptive Testing: A Machine Learning Perspective](https://mlsanity.com/paper/2404.00712/) (2024)
- [Towards a Holistic Understanding of Mathematical Questions with Contrastive Pre-training](https://mlsanity.com/paper/2301.07558/) (2023)
- [Knowledge Tagging with Large Language Model based Multi-Agent System](https://mlsanity.com/paper/2409.08406/) (2024)
- [KARL: Knowledge-Aware Retrieval and Representations aid Retention and Learning in Students](https://mlsanity.com/paper/2402.12291/) (2024)
- [Knowledge Transfer by Discriminative Pre-training for Academic Performance Prediction](https://mlsanity.com/paper/2107.04009/) (2021)
- [An Empirical Study of Finding Similar Exercises](https://mlsanity.com/paper/2111.08322/) (2021)
- [Lost in Translation: Reimagining the Machine Learning Life Cycle in Education](https://mlsanity.com/paper/2209.03929/) (2022)
- [EduNLP: Towards a Unified and Modularized Library for Educational Resources](https://mlsanity.com/paper/2406.01276/) (2024)
- [Neural-Symbolic Knowledge Tracing: Injecting Educational Knowledge into Deep Learning for Responsible Learner Modelling](https://mlsanity.com/paper/2604.08263/) (2026)
- [APGKT: Exploiting Associative Path on Skills Graph for Knowledge Tracing](https://mlsanity.com/paper/2210.08971/) (2022)
- [TQ-Net: Mixed Contrastive Representation Learning For Heterogeneous Test Questions](https://mlsanity.com/paper/2303.08039/) (2023)
- [Finding Similar Exercises in Retrieval Manner](https://mlsanity.com/paper/2303.11163/) (2023)
- [MBP-KT: Learning Global Collaborative Information from Meta-Behavioral Pattern for Enhanced Knowledge Tracing](https://mlsanity.com/paper/2605.08697/) (2026)

## Abstract (arXiv)

Understanding learning materials (e.g. test questions) is a crucial issue in online learning systems, which can promote many applications in education domain. Unfortunately, many supervised approaches suffer from the problem of scarce human labeled data, whereas abundant unlabeled resources are highly underutilized. To alleviate this problem, an effective solution is to use pre-trained representations for question understanding. However, existing pre-training methods in NLP area are infeasible to learn test question representations due to several domain-specific characteristics in education. First, questions usually comprise of heterogeneous data including content text, images and side information. Second, there exists both basic linguistic information as well as domain logic and knowledge. To this end, in this paper, we propose a novel pre-training method, namely QuesNet, for comprehensively learning question representations. Specifically, we first design a unified framework to aggregate question information with its heterogeneous inputs into a comprehensive vector. Then we propose a two-level hierarchical pre-training algorithm to learn better understanding of test questions in an unsupervised way. Here, a novel holed language model objective is developed to extract low-level linguistic features, and a domain-oriented objective is proposed to learn high-level logic and knowledge. Moreover, we show that QuesNet has good capability of being fine-tuned in many question-based tasks. We conduct extensive experiments on large-scale real-world question data, where the experimental results clearly demonstrate the effectiveness of QuesNet for question understanding as well as its superior applicability.