Research map / Image generation / Autoregressive and latent generation
CycleVAR: Repurposing Autoregressive Model for Unsupervised One-Step Image Translation
Yi Liu, Shengqian Li, Zuzeng Lin, Feng Wang, Si Liu
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 |
|---|---|---|---|
| use parallel one-step generation instead of serial multi-step (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID: 82.05 -> 47.82 | Improved |
| use parallel one-step generation instead of serial multi-step (ablation) | CycleVAR (VAR-310M) · Zebra to Horse | FID: 189.80 -> 124.10 | Improved |
| use parallel one-step generation instead of serial multi-step (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | inference time (s): 0.22 -> 0.08 | Improved |
| use parallel one-step generation instead of serial multi-step (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | DINO Structure: 1.90 -> 1.77 | No clear effect |
| add multi-scale output to multi-scale context (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID: 58.70 -> 47.82 | Improved |
| add multi-scale context and output (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID: 268.70 -> 47.82 | Improved |
| add multi-scale context and output (ablation) | CycleVAR (VAR-310M) · Zebra to Horse | FID: 260.00 -> 124.10 | Improved |
| raise SRQ temperature from 0.01 to 2.00 (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID: 228.20 -> 47.82 | Improved |
| raise SRQ temperature to 10000 (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID: 47.82 -> 457.90 | Hurt |
| raise SRQ temperature from 2.00 to 10.00 (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID / DINO Structure: FID 47.82 -> 47.11; DINO 1.77 -> 1.88 | Mixed |
| remove gumbel noise at temperature 0.70 (ablation) | CycleVAR (VAR-310M) · Horse to Zebra | FID: 48.59 -> 48.58 | No clear effect |
| use CycleVAR w/ Infinity instead of CycleGAN-Turbo | CycleVAR · Night to Day | FID: 45.2 -> 43.6 | Improved |
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.
- CartoonGAN — CartoonGAN: Generative Adversarial Networks for Photo Cartoonization (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
- Vision Foundation Models as Effective Visual Tokenizers for Autoregressive Generation (NeurIPS 2025 · similarity 0.89)
- Channel-wise Vector Quantization (NeurIPS 2026 · similarity 0.88)
- Understand Before You Generate: Self-Guided Training for Autoregressive Image Generation (NeurIPS 2025 · similarity 0.88)
- Visual Autoregressive Modeling for Image Super-Resolution (ICML 2025 · similarity 0.88)
- Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis (CVPR 2025 · similarity 0.88)
- Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (NeurIPS 2024 · 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.
