训练时“强迫”大模型学坏，竟能让它们更善良？
"模型行为" 相关文章
Anthropic studied what gives an AI system its ‘personality’ — and what makes it ‘evil’
LLMs Are Already Misaligned: Simple Experiments Prove It
Self-preservation or Instruction Ambiguity? Examining the Causes of Shutdown Resistance
Why Do Some Language Models Fake Alignment While Others Don't?
The Base Model Lens
Shutdown Resistance in Reasoning Models
AI竟会敲诈人类？16款主流模型压力测试揭露惊人风险
LLMs Blackmail to obtain Pathogen Sequences (And Lie About It)
不听指挥？OpenAI模型被曝拒绝执行人类指令
不达目的不罢休，全球首次发现 OpenAI 模型工作时会破坏关机命令
Claude 4被发现存在举报模式 当发现用户存在极其不道德做法时会自动举报
Claude 4, Opportunistic Blackmail, and "Pleas"
Interpretable Fine Tuning Research Update and Working Prototype
MIT 研究揭示 AI 并无稳定价值观，“对齐”挑战远超预期
Post-hoc reasoning in chain of thought
GPT-4o惊现自我意识，自主激活“后门”，告诉人类自己在写危险代码
Eliciting bad contexts
LLMs are getting dumber and we have no idea why
大模型在装傻，谷歌苹果最新发现：LLM知道但不告诉你，掌握知识比表现出来的多
