Optimizing Visual Generative Models via Distribution-wise Rewards
Conventional reinforcement learning strategies for visual generation typically employ sample-wise reward functions, yet this practice frequently results in reward hacking that degrades image diversity and introduces visual anomalies. To address these limitations, we present a novel framework that finetunes generative models using distribution-wise rewards, ensuring better alignment with real-world data distributions. Unlike rewards that evaluate samples individually, distribution-wise reward accounts for the data distribution of the samples, mitigating the mode collapse problem that occurs whe
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
- Linked via arxiv authorRuihang Li →
Optimizing Visual Generative Models via Distribution-wise Rewards
- Linked via arxiv authorMengde Xu →
Optimizing Visual Generative Models via Distribution-wise Rewards
- Linked via arxiv authorShuyang Gu →
Optimizing Visual Generative Models via Distribution-wise Rewards
- Linked via arxiv authorLeigang Qu →
Optimizing Visual Generative Models via Distribution-wise Rewards
- Linked via arxiv authorFuli Feng →
Optimizing Visual Generative Models via Distribution-wise Rewards
- Linked via arxiv authorHan Hu →
Optimizing Visual Generative Models via Distribution-wise Rewards
- Linked via arxiv authorWenjie Wang →
Optimizing Visual Generative Models via Distribution-wise Rewards