Research map / Image generation / Editing and style transfer
Visual Autoregressive Modeling for Instruction-Guided Image Editing
Qingyang Mao, Qi Cai, Yehao Li, Yingwei Pan, Mingyue Cheng, Ting Yao, Qi Liu, Tao Mei
This paper tried a common change: Add attention module
Add SAR in first self-attention layer · PIE-Bench, VAREdit-2B 256 · GPT-Balance: 6.588 -> 6.684 Improved
What 492 accepted papers report for this change
- 444 improved
- 20 no clear effect
- 4 mixed
- 24 hurt
- Multi-Scale Wavelet Transformers for Operator Learning of Dynamical Systems ICML 2026 · add wavelet attention block · CKF step 64 · Rel L2: 1.09 -> 0.35Improved
- DA-DFGAS:Differentiable Federated Graph Neural Architecture Search with Distribution-Aware Attentive Aggregation AAAI 2026 · add attention mechanism to aggregation · Table 4 datasets · parameters: about 5-10% more parametersHurt
- Caracal: Causal Architecture via Spectral Mixing ICML 2026 · add sliding window attention layers to MHF stack · Avg accuracy · avg accuracy (%): 48.22 -> 49.01Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 Scale-Aligned Reference (SAR) in first self-attention layer (ablation) | finest-scale conditioning · EMU-Edit, VAREdit-2B 256 | GPT-Balance: 5.248 -> 5.565 | Improved |
| Add SAR in first self-attention layer (ablation) | finest-scale conditioning · PIE-Bench, VAREdit-2B 256 | GPT-Balance: 6.588 -> 6.684 | Improved |
| Add SAR (ablation) | finest-scale conditioning · PIE-Bench, VAREdit-2B 256 | GPT-Success: 7.291 -> 7.234 | No clear effect |
| Use finest-scale source only instead of all source scales (ablation) | full-scale conditioning · EMU-Edit, VAREdit-2B 256 | GPT-Balance: 4.967 -> 5.248 | Improved |
| Use finest-scale source only instead of all source scales (ablation) | full-scale conditioning · VAREdit-2B 256 | inference time: 0.8s -> 0.5s | Improved |
| Add SAR vs full-scale conditioning (ablation) | full-scale conditioning · PIE-Bench, VAREdit-2B 256 | GPT-Balance: 6.423 -> 6.684 | Improved |
| Scale VAREdit from 2B to 8B | VAREdit-2B · EMU-Edit | GPT-Balance: 5.662 -> 7.892 | Improved |
| Use VAREdit-8B instead of strongest open basic baseline | ICEdit · EMU-Edit | GPT-Balance: 4.785 -> 7.892 | Improved |
| Use VAREdit-8B instead of UltraEdit | UltraEdit · PIE-Bench | GPT-Balance: 5.580 -> 8.105 | Improved |
| VAREdit-8B vs Qwen-Image-Edit (20B) | Qwen-Image-Edit · EMU-Edit | GPT-Balance: 8.550 -> 7.892 | Hurt |
| VAREdit-8B vs UltraEdit latency | UltraEdit · 512x512 editing | time: 2.6s -> 1.2s | Improved |
| VAREdit-8B vs GPT-4o-Image on ImgEdit-Bench Extract | GPT-4o-Image · ImgEdit-Bench Extract | score: 2.96 vs 1.52 | 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.
- OmniGen — OmniGen: Unified Image Generation (CVPR 2025) · better on 4 of 6 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
- Edit the Bits, Diff the Codes: Bitwise Residual Editing for Visual Autoregressive Models (NeurIPS 2026 · similarity 0.92)
- Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (NeurIPS 2024 · similarity 0.91)
- EditAR: Unified Conditional Generation with Autoregressive Models (CVPR 2025 · similarity 0.90)
- FireEdit: Fine-grained Instruction-based Image Editing via Region-aware Vision Language Model (CVPR 2025 · similarity 0.89)
- NEP: Autoregressive Image Editing via Next Editing Token Prediction (NeurIPS 2025 · similarity 0.89)
- The Promise of RL for Autoregressive Image Editing (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.
