RubriX speeds up rubric-based grading with faster fine-tuned scoring
RubriX: A Fast, Fine-Tunable Non-Autoregressive Rubric Judge
Computation and Language
Summary
Grading answers using detailed rubrics can be slow and costly when each part of the rubric needs to be read and judged separately by large language models. The authors present RubriX, a model that compares a student's answer directly with rubric levels using similarity scores without needing to generate text for each comparison, making it much faster. RubriX performs better on accuracy than some existing models and maintains speed even as rubric descriptions get longer. This makes it a promising tool for quickly and fairly grading student work or other responses according to complex guidelines.
What this means in practice
- •For education software developers: Integrate rapid rubric scoring to provide instant feedback on student assignments with long and detailed grading rubrics.
- •For content moderation teams: Automatically and quickly classify user text against detailed guidelines without expensive large model calls per criterion.
Authors
Balamurugan Paramasivam, Sudalai Rajkumar
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.