The WMDP Benchmark: Measuring and Reducing Malicious Use with Unlearning

N/A

The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning Nathaniel Li∗1,2 , Alexander Pan∗2 , Anjali Gopal†3,4 , Summer Yue†5 , Daniel Berrios†5 , arXiv:2403.03218v7 [cs.LG] 15 May 2024 Alice Gatti , Justin D. Li‡1,6 , Ann-Kathrin Dombrowski‡1 , Shashwat Goel‡1,7 , Long Phan‡1 , ‡1…

related reading

• Claude Mythos Preview System Cardwww-cdn.anthropic.com
• Lapis Labslapis.rocks
• A Safe Path to Open Weights - Thinking Machines Labthinkingmachines.ai
• Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilitiesarxiv.org
• [2502.05209] Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilitiesarxiv.org
• Security incident disclosure — July 2026huggingface.co
• Bitter Lessons from Distillation Robustifies Unlearningbrucewlee.com
• [2312.06942] AI Control: Improving Safety Despite Intentional Subversionarxiv.org
• Claude Mythos Preview System Cardwww-cdn.anthropic.com
• gpt-4-system-card.pdfcdn.openai.com
• [2512.05648] Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMsarxiv.org
• 2312.06942arxiv.org