UIST 2026 (Accepted)
MAIC-UI: Making Interactive Courseware with Generative UI
About this paper
A system for generating and incrementally editing interactive courseware from textbooks, slides, and PDFs.
THU-MAIC · AI for education
Publications
Research on educational agents, language models, and learning environments.
Browse selected work below, or visit Google Scholar ↗ for the full list. Preprints are labelled explicitly.
2026
Journal of Computer Science and Technology (JCST), 41(1): 394–414
From MOOC to MAIC: Reimagine Online Teaching and Learning Through LLM-Driven Agents
About this paper
A multi-agent approach to interactive online teaching and learning, extending the MOOC model with LLM-driven educational agents.
ACL 2026
SimPBL: A Multi-Agent Framework for Project-Based Learning
ACL 2026
From Knowing to Teaching: Scaffolding Pedagogical Decisions for LLM Agent
ACL 2026
Beyond Self-Report: Bridging the Intention-Behavior Gap in Critical Thinking Assessment via Interpretable Multi-Agent System
The Web Conference (WWW) 2026
Personalized Learning Path Planning through Goal-Driven Learner State Modeling
AIED 2026
Decoding Student Dialogue: A Multi-Dimensional Comparison and Bias Analysis of Large Language Models as Annotation Tools
AI Open, 7: 142–151
Which type of students can LLMs act? Investigating authentic simulation with graph-based Human–AI collaborative system
Computers & Education, 239: 105472
Mapping student-AI interaction dynamics in multi-agent learning environments: Supporting personalized learning and reducing performance gaps
About this paper
An analysis of student–AI interaction patterns in a multi-agent course and their relationships with learning gains and prior knowledge.
Journal of Research on Technology in Education
Unpacking interaction profiles and strategies in human-AI collaborative problem solving: a cognitive distribution and regulation perspective
About this paper
A study of human–AI collaborative problem solving through the lenses of distributed cognition and regulation of learning.
arXiv preprint, 2026
CogEvol: Towards Efficient and Reliable Learning Environment Generation
About this paper
Models for generating complete slides and interactive learning pages in a single pass.
2025
NAACL 2025
Simulating Classroom Education with LLM-Empowered Agents
Corresponding author
KDD 2025
Awaking the Slides: A Tuning-free and Knowledge-regulated AI Tutoring System via Language Model Coordination
Corresponding author
About this paper
Slide2Lecture turns existing lecture slides into structured teaching actions and interactive tutoring sessions.
CIKM 2025 Demo
EduCraft: A System for Generating Pedagogical Lecture Scripts from Long-Context Multimodal Presentations
Co-first author
About this paper
A system that converts long, multimodal presentations into pedagogically structured lecture scripts.
CIKM 2025 Demo
VocQuiz: Vocabulary Question Generation for English Language Education
Corresponding author
Frontiers of Digital Education, 2(4): 34
Explainable Few-Shot Knowledge Tracing
Co-first author
About this paper
An LLM-based knowledge-tracing framework for interpreting student mastery from a small number of learning records.
ACM Multimedia 2025
LongWriter-V: Enabling Ultra-Long and High-Fidelity Generation in Vision-Language Models
Corresponding author
About this paper
Training data and iterative preference optimization for long, faithful outputs from vision-language models.
KDD 2025
Knowledge-to-Jailbreak: Investigating Knowledge-driven Jailbreaking Attacks for Large Language Models
Corresponding author
KDD 2025
SoAy: A Solution-based LLM API-using Methodology for Academic Information Seeking
ACL 2025
CoT-based Synthesizer: Enhancing LLM Performance through Answer Synthesis
ACL 2025
Dynamic Scaling of Unit Tests for Code Reward Modeling
Findings of ACL 2025
TableLLM: Enabling Tabular Data Manipulation by LLMs in Real Office Usage Scenarios
学位与研究生教育, 2025年第9期
基于生成式人工智能新特征的研究生教育融合创新探析
arXiv preprint, 2025
Addressing Situated Teaching Needs: A Multi-Agent Framework for Automated Slide Adaptation
arXiv preprint, 2025
Handling Students Dropouts in an LLM-driven Interactive Online Course Using Language Models
arXiv preprint, 2025
Learning in Context: Personalizing Educational Content with Large Language Models to Enhance Student Learning
ICLS 2025
AI as Learning Partners: Students' Interactions and Perceptions in a Simulated Classroom with Multiple LLM-Powered Agents
About this paper
This study explores how students interact with and perceive multiple AI agents in a simulated classroom environment. Findings reveal that students actively engage in both cognitive and regulatory activities, holding generally positive perceptions of AI agents' cognitive support and personalization capabilities. However, limitations in emotional resonance were identified, highlighting the potential of multi-agent systems to enrich educational experiences while underscoring the need for improved emotional engagement design.
2024
EMNLP 2024 Industry
CharacterGLM: Customizing Social Characters with Large Language Models
LREC-COLING 2024
Evaluating Generative Language Models in Information Extraction as Subjective Question Correction
Expert Systems with Applications, 239: 122321
Exploring sequence-to-sequence taxonomy expansion via language model probing
KDD 2024
OAG-Bench: A Human-Curated Benchmark for Academic Graph Mining
NeurIPS 2024 Datasets and Benchmarks
SpreadsheetBench: Towards Challenging Real World Spreadsheet Manipulation
LREC-COLING 2024
Untangle the KNOT: Interweaving Conflicting Knowledge and Reasoning Skills in Large Language Models
ACL 2024
WaterBench: Towards Holistic Evaluation of Watermarks for Large Language Models
现代教育技术, 2024年第12期
多智能体协同交互的高临场感在线学习环境构建
EMNLP 2024 Demo
LM-Interview: An Easy-to-use Smart Interviewer System via Knowledge-guided Language Model Exploitation
About this paper
We present LM-Interview, a knowledge-guided language model system that automates the full pipeline of semi-structured interviews, including interview guide construction, intelligent dialogue execution, and multimodal data analysis. By adopting a state-action-reward paradigm for fine-grained control of interview flow, the system supports both qualitative and quantitative analysis dimensions. Experiments in real-world scenarios demonstrate that LM-Interview achieves performance comparable to experienced human interviewers, offering an efficient and scalable solution for qualitative research data collection.
ICLR 2024
KoLA: Carefully Benchmarking World Knowledge of Large Language Models
About this paper
We construct a Knowledge-oriented LLM Assessment benchmark (KoLA), in which we carefully design three crucial factors: (1) Ability Modeling (2) Evolving Data, (3) Standardized Evaluation. We evaluate 21 open-source and commercial LLMs and obtain some intriguing findings.
COLING 2024
A Cause-Effect Look at Alleviating Hallucination of Knowledge-grounded Dialogue Generation
About this paper
In this paper, we analyze the causal story behind this problem with counterfactual reasoning methods. Based on the causal effect analysis, we propose a possible solution for alleviating the hallucination in KGD by exploiting the dialogue-knowledge interaction.
ACL 2024
Transferable and Efficient Non-Factual Content Detection via Probe Training with Offline Consistency Checking
About this paper
This paper proposes PINOSE, which trains a probing model on offline self-consistency checking results, thereby circumventing the need for human-annotated data and achieving transferability across diverse data distributions.
2023
SIGIR 2023
MoocRadar: A Fine-grained and Multi-aspect Knowledge Repository for Improving Cognitive Student Modeling in MOOCs
About this paper
In this paper, we present MoocRadar, a fine-grained, multi-aspect knowledge repository consisting of 2,513 exercise questions, 5,600 knowledge concepts, and over 12 million behavioral records. Specifically, we propose a framework to guarantee a high-quality and comprehensive annotation of fine-grained concepts and cognitive labels.
ACL 2023
Distantly Supervised Course Concept Extraction in MOOCs with Academic Discipline
About this paper
We present a novel three-stage framework DS-MOCE, which leverages the power of pre-trained language models explicitly and implicitly and employs discipline-embedding models with a self-train strategy based on label generation refinement across different domains.
ACL 2023 Demo
VisKoP: Visual Knowledge oriented Programming for Interactive Knowledge Base Question Answering (Best Demo Paper Award)
About this paper
We present Visual Knowledge oriented Programming platform (VisKoP), a knowledge base question answering (KBQA) system that integrates human into the loop to edit and debug the knowledge base (KB) queries.
EMNLP 2023
Exploring the Cognitive Knowledge Structure of Large Language Models: An Educational Diagnostic Assessment Approach
About this paper
In this paper, based on educational diagnostic assessment method, we conduct an evaluation using MoocRadar, a meticulously annotated human test dataset based on Bloom Taxonomy.
EMNLP 2023
Mastering the Task of Open Information Extraction with Large Language Models and Consistent Reasoning Environment (Outstanding Paper)
About this paper
As the large language models (LLMs) have exhibited remarkable in-context learning capabilities, a question arises as to whether the task of OIE can be effectively tackled with this paradigm? In this paper, we explore solving the OIE problem by constructing an appropriate reasoning environment for LLMs.
KDD 2023
GLM-Dialog: Noise-tolerant Pre-training for Knowledge-grounded Dialogue Generation
About this paper
We present GLM-Dialog, a large-scale language model (LLM) with 10B parameters capable of knowledge-grounded conversation in Chinese using a search engine to access the Internet knowledge. GLM-Dialog offers a series of applicable techniques for exploiting various external knowledge including both helpful and noisy knowledge, enabling the creation of robust knowledge-grounded dialogue LLMs with limited proper datasets.
CIKM 2023
LittleMu: Deploying an Online Virtual Teaching Assistant via Heterogeneous Sources Integration and Chain of Teach Prompts
About this paper
In this paper, we present a virtual MOOC teaching assistant, LittleMu with minimum labeled training data, to provide question answering and chit-chat services.
CIKM 2023
GOAL: A Challenging Knowledge-grounded Video Captioning Benchmark for Real-time Soccer Commentary Generation
About this paper
In this paper, we present GOAL, a benchmark of over 8.9k soccer video clips, 22k sentences, and 42k knowledge triples for proposing a challenging new task setting as Knowledge-grounded Video Captioning (KGVC).
NeuraIPS 2023 Benchmarking Track
Benchmarking Foundation Models with Language-Model-as-an-Examiner
About this paper
(1) We instruct the LM examiner to generate questions across a multitude of domains to probe for a broad acquisition, and raise follow-up questions to engage in a more in-depth assessment. (2) Upon evaluation, the examiner combines both scoring and ranking measurements, providing a reliable result as it aligns closely with human annotations. (3) We additionally propose a decentralized Peer-examination method to address the biases in a single examiner.
Findings of ACL 2023
Learn to Not Link: Exploring NIL Prediction in Entity Linking
About this paper
We propose an entity linking dataset NEL focuses on the NIL prediction problem. NEL takes entities that share an alias with other entities as seeds, collects relevant mention context in the Wikipedia corpus, and ensures the presence of mentions linking to NIL by human annotation and entity masking.
2022
KDD 2022
XDAI: A Tuning‑free Framework for Exploiting the Pre‑trained Language Models in Knowledge Grounded Dialogue Generation
About this paper
We propose XDAI, a knowledge-grounded dialogue system that is equipped with the prompt-aware tuning-free PLM exploitation and supported by the ready-to-use open-domain external knowledge resources plus the easy-to-change domain-specific mechanism.
ACL 2022
Program Transfer for Answering Complex Questions over Knowledge Bases
About this paper
In this paper, we propose the approach of program transfer, which aims to leverage the valuable program annotations on the rich-resourced KBs as external supervision signals to aid program induction for the low-resourced KBs that lack program annotations.
ACL 2022
Subgraph Retrieval Enhanced Model for Multi-hop Knowledge Base Question Answering
About this paper
This paper proposes a trainable subgraph retriever (SR) decoupled from the subsequent reasoning process, which enables a plug-and-play framework to enhance any subgraph-oriented KBQA model.
ACL 2022
HOSMEL: A Hot-Swappable Modularized Entity Linking Toolkit for Chinese
About this paper
We investigate the usage of entity linking (EL) in downstream tasks and present the first modularized EL toolkit for easy task adaptation. Different from the existing EL methods that deal with all the features simultaneously, we modularize the whole model into separate parts with each feature.
COLING 2022
UPER: Boosting Multi-Document Summarization with an Unsupervised Prompt-based Extractor
About this paper
To extract documents effectively, we construct prompting templates that invoke the underlying knowledge in Pre-trained Language Model (PLM) to calculate the document and keyword’s perplexity, which can assess the document’s semantic salience. Our unsupervised approach can be applied as a plug-in to boost other metrics for evaluating a document’s salience, thus improving the subsequent abstract generation.
CIKM 2022
CStory: A Chinese Large-scale News Storyline Dataset
About this paper
In this paper, we construct CStory, a large-scale Chinese news storyline dataset, which contains 11, 978 news articles, 112, 549 manually labeled storyline relation pairs, and 49, 832 evidence sentences for annotation judgment. We conduct extensive experiments on CStory using various algorithms and find that constructing news storylines is challenging even for pre-trained language models.
CIKM 2022
ICLEA: Interactive Contrastive Learning for Self-supervised Entity Alignment
About this paper
In this paper, we propose an interactive contrastive learning model for self-supervised EA. The model encodes not only structures and semantics of entities (including entity name, entity description, and entity neighborhood), but also conducts cross-KG contrastive learning by building pseudo-aligned entity pairs.
Arxiv
Towards a General Pre-training Framework for Adaptive Learning in MOOCs
About this paper
To realize the idea of general adaptive systems proposed in pedagogical theory, with the emerging pre-training techniques in NLP, we try to conduct a practical exploration on applying pre-training to adaptive learning, to propose a unified framework based on data observation and learning style analysis, properly leveraging heterogeneous learning elements.
2021
CIKM 2021
MOOCCubeX: A Large Knowledge-centered Repository for Adaptive Learning in MOOCs (Best Resource Paper Nomination)
About this paper
We present MOOCCubeX, a large, knowledge-centered repository consisting of 4,216 courses, 230,263 videos, 358,265 exercises, 637,572 fine-grained concepts and over 296 million behavioral data of 3,330,294 students, for supporting the research topics on adaptive learning in MOOCs.
ACL 2021
Interpretable and Low-Resource Entity Matching via Decoupling Feature Learning from Decision Making
About this paper
We propose to decouple the representation learning stage and the decision making stage to fully utilize unlabeled data for entity matching task.
WISE 2021
Expertise-Aware Crowdsourcing Taxonomy Enrichment
About this paper
In this work, we propose a unified crowdsourcing framework to mitigate both challenges. It leverages the skill locality of workers with a Graph Gaussian Process model.
ICPCSEE 2021
Learning Behavior-Aware Cognitive Diagnosis for Online Education Systems
About this paper
In this paper, a learning behavior-aware cognitive diagnosis (LCD) framework is proposed for students’ cognitive modeling with both learning behavior records and exercising records.
2020
ACL 2020
MOOCCube: A Large-scale Data Repository for NLP Applications in MOOCs
About this paper
We present MOOCCube, a large-scale data repository of over 700 MOOC courses, 100k concepts, 8 million student behaviors with an external resource. Moreover, we conduct a prerequisite discovery task as an example application to show the potential of MOOCCube in facilitating relevant research.
AACL 2020
Expanrl: Hierarchical Reinforcement Learning for Course Concept Expansion in MOOCs
About this paper
We present ExpanRL, an end-to-end hierarchical reinforcement learning (HRL) model for concept expansion in MOOCs. Employing a two-level HRL mechanism of seed selection and concept expansion, ExpanRL is more feasible to adjust the expansion strategy to find new concepts based on the students’ feedback on expansion results.
AP-Web 2020
Geographical Information Enhanced POI Hierarchical Classification
About this paper
We propose an Ensemble POI Hierarchical Classification framework (EHC) consisting of three components: Textual and Geographic Feature Extraction, Hierarchical Classifier, and Soft Voting Ensemble Model.
2019
ACL 2019
Course Concept Expansion in MOOCs with External Knowledge and Interactive Game
About this paper
In this paper, we first build a novel boundary during searching for new concepts via external knowledge base and then utilize heterogeneous features to verify the high-quality results. In addition, to involve human efforts in our model, we design an interactive optimization mechanism based on a game.
2018
CCKS 2018
Predicting Concept-based Research Trends with Rhetorical Framing
About this paper
The existing researches mainly use topics extracted from literatures as objects to build predicting model. To get more accurate results, we use concepts instead of topics constructing a model to predict their rise and fall trends, considering the rhetorical characteristics of them.