
为什么要做这件事
方法
结果
污染意味着什么
点击“阅读原文”,查看附录中的完整测试结果、典型案例、提示词模板。
Golchin, S., & Surdeanu, M. (2024). Time Travel in LLMs: Tracing Data Contamination in Large Language Models. In Proceedings of the International Conference on Learning Representations (ICLR).
Ivanov, I. (2024). BioLP-bench: Measuring understanding of biological lab protocols by large language models. bioRxiv.
Rein, D., et al. (2023). GPQA: A Graduate-Level Google-Proof Q&A Benchmark.
Wang, Y., et al. (2024). MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark.
安远AI (2026). 关于:选择测评基准——平台的监测范围与评测域.
Anthropic (2025). Agentic Misalignment: How LLMs could be insider threats.
联系我们
合作咨询
通用咨询与合作
info@concordia-ai.com
媒体合作
media@concordia-ai.com
前沿AI风险监测平台咨询与合作
risk-monitor@concordia-ai.com
媒体矩阵
我们的官网
www.concordia-ai.com
公众号/视频号
安远AI
AI Safety in China英文简报
aisafetychina.substack.com
核心平台
前沿AI风险监测平台
< 平台官网 >
https://airiskmonitor.net/
< 测评基准数据库 >
https://airiskmonitor.net/benchmark/list/
State of AI Safety in China
< 系列官网 >
https://aisafetychina.com/
< China AI Safety Policy-Risk Matrix >
https://aisafetychina.com/matrix/
< Chinese Technical AI Safety Paper Database >
https://aisafetychina.com/research/
Credit:
作者:刘道宏、王金戈、王伟冰
排版:乌兰托雅

