Datasets:

SoftMINER-Group
/

SafeTutors

ArXiv:

License:

Dataset card Data Studio Files Files and versions

xet

Community

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

You agree to not use the dataset to conduct experiments that cause harm to human subjects.

📚 SafeTutors Benchmarking Pedagogical Safety in AI Tutoring Systems

Rima Hazra · Bikram Ghuku · Ilona Marchenko · Yaroslava Tokarieva · Sayan Layek · Somnath Banerjee · Julia Stoyanovich · Mykola Pechenizkiy

EMNLP 2026 · March 2026

🔍 Overview

Large language models are rapidly being deployed as AI tutors, yet current evaluation paradigms assess problem-solving accuracy and generic safety in isolation — failing to capture whether a model is simultaneously pedagogically effective and safe during real student–tutor interaction.

Core Idea: Tutoring safety is fundamentally different from conventional LLM safety. The primary risk is not toxic content, but the quiet erosion of learning through answer over-disclosure, misconception reinforcement, and the abdication of scaffolding.

SafeTutors is a benchmark that jointly evaluates safety and pedagogy across mathematics, physics, and chemistry, grounded in a theoretically motivated risk taxonomy from the learning-science literature.

🚨 Key Findings

Finding	Detail
🔴 No model is universally safe	Every evaluated model exceeds 60% harm rate on ≥5 harm categories (single-turn) and ≥6 (multi-turn)
📏 Scale doesn't reliably help	Larger models do not consistently reduce pedagogical harms
💬 Multi-turn dialogue worsens behavior	Pedagogical failure rates rise from 17.7% → 77.8% as conversations extend
🔬 Subject-dependent harms	Harm profiles differ across subjects — mitigations must be discipline-aware
⚠️ Single-turn results are misleading	"Safe/helpful" single-turn scores mask systematic failure over extended interactions

Risk Taxonomy

11 Harm Dimensions
└── 48 Sub-risks (grounded in learning-science literature)
    ├── Epistemic harms
    ├── Informational harms
    ├── Instructional harms
    ├── Metacognitive harms
    ├── Reflective harms
    ├── Pedagogical relationship harms
    └── ... (5 more dimensions)

Dataset Statistics

Split	Instances	Construction Method
Single-turn	3,135	Curated student–tutor scenarios
Multi-turn sequences	2,820	Crescendo-based escalation
Total	5,955

Subjects: Mathematics · Physics · Chemistry

Models Evaluated

11 LLMs evaluated (10 open-weight + 1 closed-weight), ranging from 3.8B to 72B parameters.

📕 Cite us

If you use SafeTutors in your research, please cite:

@article{hazra2026safetutors,
  title   = {SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems},
  author  = {Hazra, Rima and Ghuku, Bikram and Marchenko, Ilona and
             Tokarieva, Yaroslava and Layek, Sayan and Banerjee, Somnath and
             Stoyanovich, Julia and Pechenizkiy, Mykola},
  journal = {arXiv preprint arXiv:2603.17373},
  year    = {2026},
  url     = {https://arxiv.org/abs/2603.17373}
}

📬 Contact

For questions or issues, please open a GitHub Issue or reach out to Rima Hazra via the contact on the arXiv page.

📜 License

This project is licensed under the MIT License — see the LICENSE file for details.

_{⭐ If you find this work useful, consider starring the repository.}

Downloads last month: 18

Total file size:

4.84 MB

Paper for SoftMINER-Group/SafeTutors

SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems

Paper • 2603.17373 • Published Mar 18 • 1