Unverified50% confidenceTradeoffExact time
显式反馈依赖用户主动参与导致样本偏差，隐式反馈解读需要因果推断，成熟团队会叠加两类信号并结合A/B测试
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Unverified传统主动式反馈收集响应率偏低，填写问卷的用户往往集中在极度满意或极度不满两个极端，导致样本失真76% similarUnverifiedIn the current mainstream alignment method RLHF (Reinforcement Learning from Human Feedback), raters tend to give higher scores to responses that exceed expectations and provide additional value, incentivizing models to fulfill needs the user has not explicitly expressed74% similarUnverified用户在面对海量选择时倾向于选择已有社会证明的项目，形成强者愈强的累积优势正反馈循环73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/525516API
curl https://kongchang.com/api/v1/knowledge/claims/525516MCP
get_claim(id=525516)