Unverified50% confidenceFactExact time
Glasp强调'学习的社交性'，用户的公开高亮可以被其他人发现和引用，与Readwise、Hypothesis等同类工具定位不同
1
Sources
50%
Confidence
Long-term
Relevance
8/6/2026
First Seen
Sources
Related Claims
UnverifiedIn 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 expressed59% similarUnverifiedDuring the RLHF stage, human annotators tend to give higher ratings to responses that sound friendlier and more agreeable to the user, causing models to develop a systematic sycophantic tendency.58% similarUnverified人类标注员倾向于奖励结构工整、语气平衡、显得博学的回答，从而系统性强化某些特定句式58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/694304API
curl https://kongchang.com/api/v1/knowledge/claims/694304MCP
get_claim(id=694304)