Research map / Image generation / Autoregressive and latent generation
UFC-BERT: Unifying Multi-Modal Controls for Conditional Image Synthesis
Zhu Zhang, Jianxin Ma, Chang Zhou, Rui Men, Zhikang Li, Ming Ding, Jie Tang, Jingren Zhou, Hongxia 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 relevance estimator (ablation) | PNAG · M2C-Fashion | FID: 12.17 -> 11.53 | Improved |
| add fidelity estimator (ablation) | PNAG · M2C-Fashion | FID: 13.14 -> 11.53 | Improved |
| add relevance estimator (ablation) | PNAG · Multi-Modal CelebA-HQ | FID: 68.90 -> 66.72 | Improved |
| add fidelity estimator (ablation) | PNAG · Multi-Modal CelebA-HQ | FID: 70.32 -> 66.72 | Improved |
| use PNAG with both estimators instead of Mask-Predict (ablation) | UFC-BERT · M2C-Fashion | FID: 14.77 -> 11.53 | Improved |
| use PNAG instead of Mask-Predict (ablation) | UFC-BERT · Multi-Modal CelebA-HQ | FID: 72.04 -> 66.72 | Improved |
| increase parallel decoding number B from 1 to 5 (ablation) | PNAG · M2C-Fashion | FID: 12.72 -> 11.53 | Improved |
| increase B from 5 to 10 (ablation) | PNAG · Multi-Modal CelebA-HQ | FID: 66.72 -> 65.30 | Improved |
| replace autoregressive VQGAN with non-autoregressive UFC-BERT | two-stage synthesis · M2C-Fashion | FID: 12.48 -> 11.53 | Improved |
| replace autoregressive VQGAN with non-autoregressive UFC-BERT | two-stage synthesis · Multi-Modal CelebA-HQ | FID: 52.63 -> 66.72 | Hurt |
| replace autoregressive VQGAN with non-autoregressive UFC-BERT | two-stage synthesis · M2C-Fashion | inference time: 8.73 sec/sample -> 0.81 sec/sample | Improved |
| use UFC-BERT instead of TediGAN | text-to-image · Multi-Modal CelebA-HQ | FID: 106.37 -> 66.72 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TediGAN — TediGAN: Text-Guided Diverse Face Image Generation and Manipulation (CVPR 2021) · better on 2 of 2 shared results
- AttnGAN — AttnGAN: Fine-Grained Text to Image Generation With Attentional Generative Adversarial Networks (CVPR 2018) · better on 1 of 1 shared result
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
- EditAR: Unified Conditional Generation with Autoregressive Models (CVPR 2025 · similarity 0.87)
- ControlAR: Controllable Image Generation with Autoregressive Models (ICLR 2025 · similarity 0.86)
- ENAT: Rethinking Spatial-temporal Interactions in Token-based Image Synthesis (NeurIPS 2024 · similarity 0.85)
- UNIC-Adapter: Unified Image-instruction Adapter with Multi-modal Transformer for Image Generation (CVPR 2025 · similarity 0.85)
- MMAR: Towards Lossless Multi-Modal Auto-Regressive Probabilistic Modeling (CVPR 2025 · similarity 0.85)
- CtrlSynth: Controllable Image Text Synthesis for Data-Efficient Multimodal Learning (ICML 2025 · similarity 0.85)
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.
