Research map / Image generation / Text-to-image diffusion
MapReduce LoRA: Advancing the Pareto Front in Multi-Preference Optimization for Generative Models
Chieh-Yun Chen, Zhonghao Wang, Qi Chen, Zhifan Ye, Min Shi, Yue Zhao, Yinan Zhao, Hui Qu, Wei-An Lin, Yiru Shen, Ajinkya Kale, Irfan Essa and 1 more
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 |
|---|---|---|---|
| increase merging iterations from k=1 to k=10 under fixed budget (ablation) | MapReduce LoRA (SD 3.5 M) · GenEval | GenEval: +11.84% | Improved |
| increase merging iterations from k=4 to k=10 (ablation) | MapReduce LoRA (SD 3.5 M) · GenEval | GenEval: +1.12% | Improved |
| increase merging iterations from k=1 to k=10 (ablation) | MapReduce LoRA (SD 3.5 M) · PickScore | PickScore: +1.56% | Improved |
| increase merging iterations from k=4 to k=10 (ablation) | MapReduce LoRA (SD 3.5 M) · PickScore | PickScore: +0.57% | Improved |
| increase merging iterations from k=1 to k=10 (ablation) | MapReduce LoRA (SD 3.5 M) · OCR | OCR: +10.72% | Improved |
| increase merging iterations from k=4 to k=10 (ablation) | MapReduce LoRA (SD 3.5 M) · OCR | OCR: +0.03% | No clear effect |
| add RaTE token on top of MapReduce LoRA | MapReduce LoRA (SD 3.5 M) · GenEval | GenEval: 0.90 -> 0.92 | Improved |
| add RaTE token on top of MapReduce LoRA | MapReduce LoRA (SD 3.5 M) · PickScore | PickScore: 22.552 -> 22.777 | Improved |
| append a single GenEval RaTE token to the prompt | SD 3.5 M · GenEval | GenEval: 0.68 -> 0.76 | Improved |
| append more RaTE tokens (x1 to x3) (ablation) | SD 3.5 M + RaTE · PickScore | PickScore: 22.052 -> 22.021 | Hurt |
| append more RaTE tokens (x1 to x3) (ablation) | SD 3.5 M + RaTE · OCR | OCR: 0.623 -> 0.635 | Improved |
| train multi-reward with mixed data and mixed reward (MORL-DR) instead of MapReduce LoRA | SD 3.5 M · GenEval | GenEval: 0.90 -> 0.67 | Hurt |
Closest papers by meaning
- Multimodal Large Language Models Make Text-to-Image Generative Models Align Better (NeurIPS 2024 · similarity 0.90)
- DreamAlign: Dynamic Text-to-3D Optimization with Human Preference Alignment (AAAI 2025 · similarity 0.89)
- Auto-Rubric as Reward: From Implicit Preference to Explicit Generative Criteria (NeurIPS 2026 · similarity 0.88)
- Personalized Preference Fine-tuning of Diffusion Models (CVPR 2025 · similarity 0.88)
- RoVRM: A Robust Visual Reward Model Optimized via Auxiliary Textual Preference Data (AAAI 2025 · similarity 0.88)
- Data with High and Consistent Preference Difference Are Better for Reward Model (AAAI 2025 · similarity 0.88)
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.
