Unverified50% confidenceCautionExact time
When models trained on game-rendered images are directly applied to real-world scenes, performance often degrades significantly due to domain differences.
1
Sources
50%
Confidence
Long-term
Relevance
8/17/2026
First Seen
Sources
Starfield Fauna数据集：2万张游戏生物图像的AI训练价值解析
Related Claims
Unverified过拟合是LoRA训练中最常见的失败模式，模型会死记硬背训练图像像素特征而非学习本质特征73% similarUnverifiedThe sim-to-real gap refers to the phenomenon where policies trained in simulation often suffer significant performance degradation when transferred to the real world, and domain randomization and system identification are commonly used mitigation techniques.70% similarUnverified对抗图案通常针对特定模型、特定版本训练，更换检测算法或升级模型权重后可能完全失效68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/763983API
curl https://kongchang.com/api/v1/knowledge/claims/763983MCP
get_claim(id=763983)