EdU-Jev improves automated education decisions with faster accurate models

EDU-Jev Training System-One Decision Models for Education

Computation and Language

Summary

Many educational tasks need automatic decisions from a fixed set of answers, like scoring or checking solutions. The authors developed EDU-Jev, a type of language model that picks answers directly instead of writing explanations, making it faster and more focused. Their biggest model is more accurate on education tests than popular commercial tools and works well across different subjects. EDU-Jev also runs quickly on modern GPUs, helping it handle many decisions efficiently.

What this means in practice

  • For education technology developers: Integrate EDU-Jev to assign scores or select answer options automatically in digital assessment platforms, improving accuracy and speed over current APIs.
  • For automated testing software builders: Use EDU-Jev models to rapidly classify student responses within fixed option sets without needing full text generation, reducing inference cost.

Authors

Jiazheng Li, Yulan He

Abstract

Many automated decisions in education involve choosing from a fixed set of options: assigning a rubric score, checking whether an answer meets a marking criterion, or identifying the first wrong step in a solution. Generative language models are increasingly used to make these decisions, but their responses require parsing, longer outputs increase inference costs, and confidence expressed in words does not necessarily reflect the probability of being correct. We present EDU-Jev, a family of single-pass ("System One") decision models at 0.8B, 2B and 4B parameters. Given an assessment context and a question with predefined answer options, the model returns a probability distribution over those options without generating text. A pointer head scores a marker for each option, ensuring the predictions are restricted to the allowed choices. Training uses a single supervised stage with soft cross-entropy against the available target label distributions; calibration is assessed empirically. Across twelve education benchmarks, EDU-Jev 4B achieves an average accuracy of 67.6%, exceeding the strongest of five commercial decision APIs by 9.4 percentage points and the best open 4B decision model by 18.7 points. On four cross-domain education sets, it matches the strongest API, which remains ahead on general decision-making and reasoning tasks. The 2B model processes about 150 decisions per second on a single H100 GPU.