Unverified50% confidenceFactExact time
SatisDive sets two hard constraints: a reward lower bound (the minimum reward threshold each candidate image must reach) and a diversity cutoff (the diversity threshold that the entire batch of images as a set must satisfy).
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn text-to-image generation, excessively pursuing reward scores causes generated results to trend toward homogeneity, while single-mindedly pursuing diversity may produce a large number of low-quality images.68% similarUnverified该方法依赖高质量可微的人脸相似度模型作为奖励来源，单步计算成本较高，对画风、构图等奖励信号不明确的属性未必适用63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/975678API
curl https://kongchang.com/api/v1/knowledge/claims/975678MCP
get_claim(id=975678)