Tag
The paper proposes Ruspan, a rubric-conditioned Automatic Short Answer Scoring framework that serializes question, answer, and rubric levels for single-pass LLM scoring, plus a Rubric-Independent Mask variant that improves zero-shot transfer on multilingual ASAS benchmarks with rubric-structure shift.
IKS-Instruct is a multilingual dataset of 24,795 instruction-response pairs for teaching language models Indian Knowledge Systems, spanning seven Indian languages and covering 41 pedagogical techniques. Evaluation shows a fine-tuned 7B model performs competitively with larger general-purpose models on IKS-specific tasks.
This paper introduces SEFORA, a public corpus of instructor feedback on student essays, and UniMatch, a reference-based evaluation framework for assessing LLM-generated feedback. Experiments show that current LLMs struggle to match instructor feedback, achieving at most 0.4 F1.
This paper describes UOL@IDEM's closed-track submission to the BEA 2026 shared task on L1-aware vocabulary difficulty prediction, combining multilingual contextual representations with engineered features. The system achieves competitive RMSE scores for Spanish, German, and Chinese, with frequency being the most stable predictor.
This paper proposes a fully local AI cascade for de-identifying educational dialogue, combining a recall-first candidate proposer with a contextual Redact/Keep reviewer. The approach achieves high accuracy without sending data to external APIs, outperforming both smaller local models and commercial APIs on math tutoring transcripts.