Unverified50% confidenceSolutionExact time
经验丰富的研究者通常采用结构化rebuttal回复策略：先回应影响评分的核心质疑，再处理次要问题，最后用表格总结每个concern的解决状态
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
Related Claims
Unverified自适应访谈与固定问卷的关键区别在于后续问题基于前一个答案的内容动态生成，依赖大语言模型的上下文理解能力68% similarVerifiedReAct 智能体的核心是「推理-行动-观察」循环，LLM 分析问题后决定调用工具，观察返回结果，若不满足则继续推理直到给出最终回复67% similarUnverifiedSelf-reflection is achieved by having the model evaluate the quality of its own output and iteratively improve it; the Reflexion framework demonstrates that an agent can learn from failed attempts and adjust its strategy in the next iteration.66% similarUnverified作者认为决定最终优化效果的主要是提示词和使用的大模型型号，提示词越具体Agent执行越到位66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/701150API
curl https://kongchang.com/api/v1/knowledge/claims/701150MCP
get_claim(id=701150)