Accepted
arXiv:2610.00664
Accepted · 11th IEEE Asia-Pacific Conference on Computer Science and Data Engineering (IEEE CSDE 2026)September 2026

PhysicsMate: A Curriculum-Grounded Bengali Benchmark for Secondary Physics QA with Small-Model Adaptation

Rashid Azraf JahinNorth South University
Saadman SajidNorth South University
Khan Raiyan Ibne RezaNorth South University
Sumaiya Tabassum NimiNorth South University

Abstract

PhysicsMate is a curriculum-grounded Bengali benchmark for secondary-school physics question answering, built from the NCTB Grade 9–10 syllabus. It contains 1,834 question-answer pairs grounded in a multi-relational knowledge graph with 1,760 nodes and 2,600 edges. Small-model adaptation across 0.6B, 1.7B, and 4B parameter scales improves closed-book accuracy by 5.5, 15.0, and 23.3 percentage points, respectively, supporting offline educational use in settings with limited connectivity and hardware.

Key Empirical Receipts & Quantitative Highlights

  • Curriculum-grounded Bengali physics QA benchmark with 1,834 question-answer pairs and a 1,760-node knowledge graph.
  • Small-model adaptation improves closed-book accuracy by 5.5, 15.0, and 23.3 percentage points at 0.6B, 1.7B, and 4B parameters.
  • The 4B adapted model is quantized for offline use in resource-constrained settings.

Problem Context & Real-World Motivation

The work studies curriculum-aligned physics question answering in Bengali and the use of compact adapted models where compute and connectivity are limited.

Principal Research Contributions

01Contribution

Introduces a Bengali secondary-physics QA benchmark grounded in the NCTB Grade 9–10 curriculum.

02Contribution

Structures curriculum knowledge as a multi-relational graph to support small-model adaptation.

03Contribution

Evaluates adaptation at 0.6B, 1.7B, and 4B parameter scales, including an offline quantized model.

Practical Significance & South Asian AI Implications

Shows a path toward curriculum-grounded physics support that can run offline on resource-constrained hardware.

BibTeX Citation

@misc{jahin2026physicsmate,
  title={PhysicsMate: A Curriculum-Grounded Bengali Benchmark for Secondary Physics QA with Small-Model Adaptation},
  author={Jahin, Rashid Azraf and Sajid, Saadman and Reza, Khan Raiyan Ibne and Nimi, Sumaiya Tabassum},
  year={2026},
  eprint={2610.00664},
  archivePrefix={arXiv}
}