Work

SHAPE: Unifying Safety, Helpfulness and Pedagogy for Educational LLMs

Year: 2026

Type: conference-paper

Abstract: Large Language Models (LLMs) have been widely explored in educational scenarios.We identify a critical vulnerability in current educational LLMs, pedagogical jailbreaks, where students use answer-inducing prompts to elicit solutions rather than scaffolded instructions.To enable systematic study, we unify and formalize safe, helpful, and pedagogical behaviors with a knowledge-mastery graph and introduce SHAPE, a benchmark of 9,087 studentquestion pairs for evaluating tutoring behavior under adversarial pressure.We propose a graph-augmented tutoring pipeline that infers prerequisite concepts from queries, identifies mastery gaps, and routes generation between instructing and problem-solving via explicit gating.Experiments across multiple LLMs show that our method yields significantly improved safety under two pedagogical jailbreak settings, while maintaining near-ceiling helpfulness under the same evaluation protocol.

Source: Proceedings of the conference - Association for Computational Linguistics. Meeting

DOI: 10.18653/v1/2026.acl-long.529

Authors: Sihang Zhao, Kangrui Yu, Youliang Yuan, Pinjia He, Hongyi Wen

Institutions: New York University Shanghai, Chinese University of Hong Kong, Chinese University of Hong Kong, Shenzhen

Language: English

Topic: Intelligent Tutoring Systems and Adaptive Learning

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

SDG: Quality education

Open Access status: hybrid

Funders: New York University Shanghai, Science and Technology Commission of Shanghai Municipality