Research map / Image generation / Text-to-image diffusion
BideDPO: Conditional Image Generation with Simultaneous Text and Condition Alignment
Dewei Zhou, Mingwei Li, Zongxin Yang, Yu Lu, Yunqiu Xu, Zhizhong Wang, Zeyi Huang, Yi 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 |
|---|---|---|---|
| add adaptive loss balancing (ablation) | BideDPO on Union-Pro2 · DualAlign depth | success rate: 0.78 -> 0.84 | Improved |
| add adaptive loss balancing (ablation) | BideDPO on Union-Pro2 · DualAlign depth | MSE: 157.729 -> 163.968 | Hurt |
| add condition preference pairs to text-only pairs (ablation) | BideDPO on Union-Pro2 · DualAlign depth | MSE: 258.954 -> 163.968 | Improved |
| add text preference pairs to condition-only pairs (ablation) | BideDPO on Union-Pro2 · DualAlign depth | success rate: 0.59 -> 0.84 | Improved |
| iterate optimization to 3 rounds (ablation) | BideDPO on Union-Pro2 · DualAlign depth | success rate: 0.84 -> 0.88 | Improved |
| 4th iteration (ablation) | BideDPO on Union-Pro2 · DualAlign depth | success rate: 0.88 -> 0.86 | Hurt |
| decoupled BideDPO instead of naive DPO | Union-Pro2 · DualAlign depth | success rate: 0.71 -> 0.84 | Improved |
| decoupled BideDPO instead of naive DPO | FLUX-Canny · DualAlign canny | success rate: 0.55 -> 0.73 | Improved |
| decoupled BideDPO instead of SFT on same positives | FLUX-Depth · DualAlign depth | MSE: 162.2 -> 145.9 | Improved |
| apply BideDPO | Union-Pro2 · DualAlign depth | success rate: 0.49 -> 0.84 | Improved |
| apply BideDPO | FLUX-Canny · DualAlign canny | success rate: 0.33 -> 0.73 | Improved |
| apply BideDPO to IP-Adapter | FLUX IP-Adapter · Style-conditioned | style score: 6.50 -> 6.26 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ControlNet++ — ControlNet++: Improving Conditional Controls with Efficient Consistency Feedback (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ImagenHub: Standardizing the evaluation of conditional image generation models (ICLR 2024 · similarity 0.89)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.89)
- Discriminative Probing and Tuning for Text-to-Image Generation (CVPR 2024 · similarity 0.88)
- Referee Can Play: An Alternative Approach to Conditional Generation via Model Inversion (ICML 2024 · similarity 0.88)
- Generalizing Alignment Paradigm of Text-to-Image Generation with Preferences Through f-Divergence Minimization (AAAI 2025 · similarity 0.88)
- CHATS: Combining Human-Aligned Optimization and Test-Time Sampling for Text-to-Image Generation (ICML 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.
