Papers for
education software developers
Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.
RubriX speeds up rubric-based grading with faster fine-tuned scoring
RubriX: A Fast, Fine-Tunable Non-Autoregressive Rubric Judge
Abstract: Rubric-based evaluation asks which level of a rubric a response meets. It is usually done by prompting a large language model (LLM) as a judge, which needs a generative call per judgment and reads every level description, so its cost grows with the rubric. We present RubriX, a rubric judge whose architecture is inspired by contrastive language models (CLM): a LoRA-adapted encoder embeds the input and each level description, and a softmax over scaled cosine similarities gives a probability for every level without text generation. Trained on public rubric data and proprietary synthetic data, RubriX reaches 75.4% exact accuracy on held-out rubrics, against 57.1% for TypeSafe's Jev and 49.8% for a Qwen3-8B judge. On eight human-labelled benchmark sets it is within 6 percentage points of Jev's accuracy on most of them, while Jev keeps a larger lead on two and ranks responses best on average; against the results reported in the G-Eval paper, RubriX outperforms G-Eval with GPT-3.5 on every SummEval dimension and both QAGS subsets. With level embeddings cached, RubriX's latency stays at about 28 ms per judgment as level descriptions grow from 10 to 3,000 tokens, while Jev's (approximate network time subtracted) rises from 44 ms to 211 ms and the Qwen3-8B judge's from 32 ms to 516 ms, making RubriX 7.7x faster than Jev and 18.8x faster than the Qwen3-8B judge at 3,000 tokens per level.
Complexity of academic degree planning depends on credit capacity and course rules
The Plan Language of a Curriculum: A Formal Model and the Complexity of Degree Planning
Abstract: We model an academic curriculum as a generator of a language of feasible study plans: prerequisites are monotone Boolean formulas in conjunctive normal form, degree requirements are credit-threshold covering constraints, and a study plan is a sequence of terms bounded by a per-term credit capacity. Within this model, we settle the complexity of the two natural planning objectives, the number of terms to a degree and the total credit load, and we isolate the structural commitment responsible for each source of hardness. Time to degree is polynomial whenever the per-term capacity is unbounded, for arbitrary disjunctive prerequisites and arbitrary electives, so disjunction never contributes to its hardness, yet it becomes strongly NP-hard as soon as capacity binds, even without any prerequisite. Load is complementary: disjunction and overlapping electives are each strongly NP-hard in isolation and their complexity does not depend on capacity, while load is polynomial on the conjunctive, mandatory fragment. The two objectives therefore have disjoint sources of hardness. We show that the delay-factor component of the standard curricular-complexity metric is a polynomially computable upper bound on time to degree, exact on the conjunctive fragment and loose elsewhere by a quantity we name the disjunctive slack, and we prove that program subsumption is coNP-complete and consensus prerequisite recovery is NP-complete. Instantiating the model on a corpus of twenty-two universities, we find that 88 percent of prerequisite-bearing courses are purely conjunctive and that capacity, not prerequisite logic, is the operative constraint on time to degree. The curriculum corpus is openly available (https://doi.org/10.5281/zenodo.22334674), and the analysis and figure-generation code accompany the paper.
Student answer timing and class divisions reveal learning insights
Insights on Student Learning from Live Classroom Polls: More Than Right or Wrong
Abstract: A live classroom poll is usually read through its answer key, as a count of participation and correctness. Yet the same record holds more before any key is fixed, namely who answers early and how the class divides across the options. Whether these key-free signals are stable, and what they reveal, has not been examined at the scale of a real course. We study them in a live deployment of 340,668 submissions across 603 polls in 47 sessions from roughly 2,400 learners, reading each poll for a learner's response order and for how much the class divides. A single further step brings in the author-designated key, on items with a settled answer, to interpret the divided ones. A learner's response order is a stable individual signature. It reproduces at a corrected split-half reliability of 0.92, holds at 0.87 within single sessions, and is unrelated to whether the learner is correct. Read without a key, about one poll in five does not reach a clear majority, and division marks the questions that split the class. On the settled-answer items the majority is itself wrong on 49 of 358. The key-free measure flags most of these, though division alone does not separate a collectively wrong answer from a merely hard one. The two readings are largely independent, meeting at one point where early responders anticipate the eventual majority only on questions the class has nearly settled. Together they map the learners and questions of a session without a key.