Unverified50% confidenceFactExact time
Gary Klein提出的识别启动决策模型（RPD）描述专家在感知输入触发原型模式后直接激活行动脚本，跳过分析性比较阶段
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
UnverifiedRLHF的人类介入发生在训练阶段并固化为模型权重，属于隐式编码，而表征工程干预在推理时操控内部表征，具有即时性和可逆性69% similarUnverifiedRLHF遵循先用监督微调（本质是行为克隆）建立基础能力，再用PPO等算法优化人类偏好目标的范式68% similarUnverified指令微调和RLHF强化了模型的指令跟随倾向，使其在提升任务能力的同时更易被外部指令误导67% similarUnverifiedThe core technology of Builder mode adopts the 'ReAct' (Reasoning + Acting) paradigm, in which the model first performs reasoning at each step, then decides the next action, and then observes the execution result to form a closed-loop feedback.67% similarUnverifiedTraditional RPA is a rule-based deterministic system whose behavior is entirely determined by preset scripts, unlike Agentic AI, which possesses generalization capabilities.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/575732API
curl https://kongchang.com/api/v1/knowledge/claims/575732MCP
get_claim(id=575732)