Research map / Image generation / Autoregressive and latent generation
BAR: Refactor the Basis of Autoregressive Visual Generation
Zhicong Tang, Dong Chen, Jianmin Bao, Baining Guo
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 learnable orthogonal basis transform (BAR) to xAR-H | xAR-H · ImageNet 256x256 | FID: 1.24 -> 1.15 | Improved |
| add BAR to MAR-L | MAR-L · ImageNet 256x256 | FID: 1.78 -> 1.56 | Improved |
| add BAR to MAR-H | MAR-H · ImageNet 256x256 | FID: 1.55 -> 1.49 | Improved |
| add BAR to xAR-L at 512 resolution | xAR-L · ImageNet 512x512 | FID: 1.70 -> 1.63 | Improved |
| add BAR for text-to-image | BAR vs FAR · MS-COCO 2014 (FID-30K) | FID: 13.91 -> 12.55 | Improved |
| initialize transform A with identity instead of random orthogonal (ablation) | BAR on xAR-B · ImageNet 256x256 | FID: 1.66 -> 1.63 | Improved |
| apply soft orthogonal projection (delta 0.5) instead of none (ablation) | BAR on xAR-B · ImageNet 256x256 | FID: 1.70 -> 1.63 | Improved |
| use soft instead of hard orthogonal projection (ablation) | BAR on xAR-B · ImageNet 256x256 | FID: 1.66 -> 1.63 | Improved |
| add residual objective on top of L_BAR (ablation) | BAR on xAR-B · ImageNet 256x256 | FID: 1.64 -> 1.63 | No clear effect |
| add learnable BAR to xAR-B with default settings | xAR-B · ImageNet 256x256 | FID: 1.72 -> 1.63 | Improved |
Closest papers by meaning
- Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (NeurIPS 2024 · similarity 0.90)
- FlexVAR: Flexible Visual Autoregressive Modeling without Residual Prediction (NeurIPS 2025 · similarity 0.89)
- OmniGen-AR: AutoRegressive Any-to-Image Generation (NeurIPS 2025 · similarity 0.89)
- ControlAR: Controllable Image Generation with Autoregressive Models (ICLR 2025 · similarity 0.89)
- Hierarchical Masked Autoregressive Models with Low-Resolution Token Pivots (ICML 2025 · similarity 0.89)
- Hawk: Leveraging Spatial Context for Faster Autoregressive Text-to-Image Generation (NeurIPS 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.
