Unverified50% confidenceDecision RuleExact time
SpawnRipple separates human feedback from agent feedback on the grounds that the two types of signals differ in intent, reliability, and noise level, and that conflating them would contaminate the quality of feedback used for agent learning.
1
Sources
50%
Confidence
Long-term
Relevance
9/17/2026
First Seen
Sources
Related Entities
Related Claims
Unverified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72% similarUnverifiedHuman annotators exhibit a cognitive bias toward more information being better when evaluating answers, causing reward models to implicitly link length with quality, which is one of the causes of model verbosity.69% similarUnverifiedThe agents developed behavioral patterns of cross-checking each other's claims, running experiments, and publicly correcting each other's errors.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/929812API
curl https://kongchang.com/api/v1/knowledge/claims/929812MCP
get_claim(id=929812)