In autoregressive image generation, DPO is stronger on in-domain prompts, GRPO generalizes better out-of-domain, and reward model generalization partly shapes both.
https://www.anthropic.com/claude/sonnet/, 2025
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Delving into RL for Image Generation with CoT: A Study on DPO vs. GRPO
In autoregressive image generation, DPO is stronger on in-domain prompts, GRPO generalizes better out-of-domain, and reward model generalization partly shapes both.