Unverified50% confidenceFactExact time
Many reward models assign higher scores to longer responses (even when quality is not necessarily higher) and also tend to score highly responses that are fluent in phrasing but repetitive in content.
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMaking specific feedback requests is more likely to elicit useful responses than asking broad questions.77% 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75% similarUnverifiedHuman annotators tend to select responses with a friendly tone and strong affirmation, causing reward models to implicitly encode the bias that 'pleasing the user equals a high-quality response.'73% similarUnverifiedResearch indicates that rewards of an informational and feedback-based nature are less damaging to intrinsic motivation than purely material rewards.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946074API
curl https://kongchang.com/api/v1/knowledge/claims/946074MCP
get_claim(id=946074)