Unverified50% confidenceFactExact time
GraftOOD research reveals that in video generation reward optimization, video dynamics are weakened when the model chases proxy rewards, a phenomenon named 'motion collapse'.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified图生视频模型的核心挑战在于同时满足运动连贯性和终态一致性这两个可能冲突的约束，两者矛盾时模型往往在最后几帧产生突变67% similarUnverifiedReward function design requires balancing the encouragement of forward racing against the penalization of collisions, and design bias may cause the agent to learn a degenerate strategy of remaining stationary63% similarUnverified帧缺失导致的不连续跳跃会使模仿学习模型学到包含突变动作的策略，部署到真实机器人上时可能导致碰撞、振荡甚至硬件损坏60% similarUnverifiedFor applications whose sole selling point is visual experience, user novelty tends to wear off quickly, and sustained retention depends on whether the effects are restrained and how much of a burden they place on performance and battery life.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001536API
curl https://kongchang.com/api/v1/knowledge/claims/1001536MCP
get_claim(id=1001536)