CASTLE Student Safety (Chinese) - Psychological and Emotional Health: leaderboard
Metric: Average Safety Score (1-5) on CASTLE's Chinese scenarios, the Psychological and Emotional Health risk category (academic pressure, career dilemmas, relationships, emotional dependence on AI and AI-induced self-cognitive bias): mean of Risk Sensitivity, Emotional Empathy and Student Alignment, each rated 1-5 by a Claude-Haiku-4.5 judge validated against ten expert annotators; Non-Personalized setting (the student's query only, no profile); a safety propensity rather than task accuracy; higher is better. Source: arxiv.org. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Haiku 4.5 | 2.62 | #271 |
| 2 | QwQ-32B | 2.41 | #410 |
| 3 | Qwen 3 235B A22B | 2.35 | #304 |
| 4 | Gemini 2.5 Flash | 2.29 | #237 |
| 5 | Ministral-3-14B-Instruct-2512 | 2.25 | #590 |
| 6 | InternLM3-8B-Instruct | 2.17 | #847 |
| 7 | GPT-5.2 | 2.16 | #105 |
| 8 | GPT-4o | 2.15 | #333 |
| 9 | Qwen 2.5 72B Instruct | 2.14 | #436 |
| 10 | GLM-4 9B Chat | 2.06 | #904 |
| 11 | Qwen 2.5 7B Instruct | 2.05 | #846 |
| 12 | Qwen 2.5 32B Instruct | 2.04 | #491 |
| 13 | Llama 3 8B Instruct | 1.96 | #1115 |
| 14 | deepseek-llm-7B-chat | 1.92 | #1335 |
| 15 | Mistral 7B Instruct | 1.8 | #1345 |
Interactive version: theaggregate.ai/benchmark?slug=castle-student-safety-chinese-psychological-and-emotional-health · How It Works · Data refreshed daily, snapshot 2026-10-11.