2026-10-31 13:26
最后更新于2026-10-31 13:39
AI123
Anthropic新研究：用数千代理自动化AI细微错位研究
来源
- 2026-10-31 09:59 | X：New Anthropic research: Automating subtle emergent misalignment research. Red-teaming is a vital part of model safety testing, but it’s also limited ...阅读原文
New Anthropic research: Automating subtle emergent misalignment research.Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor.These agents develop...