July 2026 - This past year, we released a new dataset called FoundationalASSIST, developed through a collaboration between my lab at WPI, the ASSISTments Foundation, and Dr. Shashank Sonkar at UCF.

This dataset is, to our knowledge, the first large-scale educational dataset that includes full problem text, answer choices, and students’ exact responses (not just correctness labels). This matters for modern AI: instead of seeing “student 4521 got problem 892 wrong,” models can now see what the problem asked and what the student actually wrote—opening the door to modeling misconceptions, reasoning, and learning processes directly.

The dataset includes:

• 1.7M student interactions

• 5,000 students (each with 200+ problems)

• Middle school math problems from Illustrative Mathematics

• Rich metadata (skills, timing, hints, etc.)

We’ve also introduced benchmark tasks spanning knowledge tracing and pedagogical reasoning, showing that current LLMs still struggle with core challenges like predicting student behavior and identifying good assessment items.

Our paper was presented at the Festival of Learning at AIED, 2026: 

Worden, E., Heffernan, C., Heffernan, N., Sonkar, S. (2027). FoundationalASSIST: Dataset for Foundational Knowledge Tracing & Pedagogical Grounding of Large Language Models. In: Blanchard, E.G., Chen, G., Chi, M., Isotani, S. (eds) Artificial Intelligence in Education. AIED 2026. Lecture Notes in Computer Science(), vol 16582. Springer, Cham. https://doi.org/10.1007/978-3-032-29755-6_36  

You can also access a longer preprint version here:
https://arxiv.org/abs/2602.00070

You can request access to the dataset here:
https://huggingface.co/datasets/ASSISTments/FoundationalASSIST

We’d love to see how the community uses it and are happy to answer questions.

Neil Heffernan is an active researcher in the fields of (1) artificial intelligence and education, (2) educational data mining and (3) learning analytics. In order to support research in these fields, Dr. Heffernan created the E-TRIALS Testbed, a tool that allows ASSISTments to be used as a platform to do science and support evidence-based practice.

Most recently, the focus of my research is using large language models (or AI) to support math teachers on their students' open-ended answers. Open-ended answers represent a shift in math education and common core learning standards. By asking the student to describe their thinking in words, teachers can better gauge student understanding beyond a right or wrong answer. The problem is that teachers don't often have the time to read every single written answer or provide immediate feedback to their students. Much of my research is focused on validating AI scoring and feedback to students, and providing teachers with advanced insights into student thinking. 

For over a decade Neil Heffernan has been teaching one undergraduate AI class (Introduction to Artificial Intelligence: CS 4341) and one graduate course in his area (rotating between the three course listed below). 

He rotates through these three graduate classes, teaching one each year:

• Special Topics: Large Language Models for Education (CS 525) *New* To be taught 2026-27)

• Artificial Intelligence for Adaptive Educational Technology (CS 568). (Last taught Fall 2024, next expected to be taught in 2027-28) 

• User Modeling, which focuses on educational data mining (CS 565). (Last taught Fall 2025, next expected to be taught in 2028-29)

• Special Topics: Online Learning Infrastructure (CS 525) (Last taught 2023, next expected to be taught in Fall 2029-30) 

He teaches at the undergraduate level this class:

• Introduction to Artificial Intelligence (CS 4341) (last taught Fall 2019, A Term).