⚡Megatron LLM Hub
←Back to search
Twitter

改动不足十个神经元即可让 Agent 模型变成潜伏后门,且仍能通过全部评测

权重层面的后门可逃过现有评测流程,Agent 模型供应链信任需要重新评估。
ai_agentllm报告
RT @MsftSecIntel: New research from @MSFTResearch found that editing fewer than ten neurons could turn a capable agent model into a sleeper agent that passes all evaluations but runs an attacker’s command with the agent’s environment access when a trigger string appears in its context. https://t.co/PZRb8ulov5