Research map / Image generation / Text-to-image diffusion
RMLer: Synthesizing Novel Objects Across Diverse Categories via Reinforcement Mixing Learning
Jun Li, Zikun Chen, Haibo Chen, Shuo Chen, Jian Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace standard PPO with critic-free PPO plus visual reward (ablation) | RMLer · CangJie-200 | VQAScore: 0.4155 -> 0.4287 | Improved |
| replace standard PPO with critic-free PPO plus visual reward (ablation) | RMLer · CangJie-200 | HPSv2: 0.2746 -> 0.2774 | Improved |
| increase reward balance factor alpha from 0 to 5 (ablation) | RMLer · ImageNet-200 (10 pairs) | VQAScore: 0.1773 -> 0.2148 | Improved |
| increase reward balance factor alpha from 0 to 5 (ablation) | RMLer · ImageNet-200 (10 pairs) | HPSv2: 0.2753 -> 0.2748 | No clear effect |
| use the RL-trained policy rather than direct reward optimization (ablation) | RMLer · qualitative (Fig. 6c) | fusion quality | Improved |
| use stochastic sampling rather than deterministic sampling (ablation) | RMLer · qualitative (Fig. 6c) | fusion quality | Improved |
| RMLer versus SDXL-Turbo on image similarity | RMLer · ImageNet-200 | Avg. Sim (I->I): 0.7647 -> 0.7324 | Hurt |
| RMLer versus BASS on balance | RMLer · ImageNet-200 | Balance (I->I): 0.0918 -> 0.0080 | Improved |
| different exemplar image sets for reward (ablation) | RMLer · qualitative (Fig. 6b) | visual appearance | No clear effect |
Closest papers by meaning
- HiCoGen: Hierarchical Compositional Text-to-Image Generation in Diffusion Models via Reinforcement Learning (CVPR 2026 · similarity 0.88)
- Non-confusing Generation of Customized Concepts in Diffusion Models (ICML 2024 · similarity 0.86)
- Sketch Then Paint: Hierarchical Reinforcement Learning for Diffusion Multi-Modal Large Language Models (NeurIPS 2026 · similarity 0.86)
- Progressive Text-to-Image Diffusion with Soft Latent Direction (AAAI 2024 · similarity 0.86)
- Magnet: We Never Know How Text-to-Image Diffusion Models Work, Until We Learn How Vision-Language Models Function (NeurIPS 2024 · similarity 0.86)
- DreamDistribution: Learning Prompt Distribution for Diverse In-distribution Generation (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
