Research map / Image generation / Text-to-image diffusion
Text-driven synthesis and semantic guidance
Control text-to-image models for size, semantics, and LLM-guided correction.
137 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2023–2026) · an approach in Text-to-image diffusion, keyword cluster “synthesis · layout · object”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
60 papers in 2023–24 and 77 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 15.3% to 9.0%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2023 · Magic3D: High-Resolution Text-to-3D Content Creation (CVPR 2023 · 813 citations)
- 2024 · Scaling Rectified Flow Transformers for High-Resolution Image Synthesis (ICML 2024 · 92 citations)
- 2025 · SALAD: Skeleton-aware Latent Diffusion for Text-driven Motion Generation and Editing (CVPR 2025 · 13 citations)
- 2026 · MagicPaint: Operate Anything for Image Inpainting with Diffusion Model (AAAI 2026 · 1 citation)
Earliest papers here
- Magic3D: High-Resolution Text-to-3D Content Creation (CVPR 2023 · 813 citations)
- Align Your Latents: High-Resolution Video Synthesis With Latent Diffusion Models (CVPR 2023 · 679 citations)
- NULL-Text Inversion for Editing Real Images Using Guided Diffusion Models (CVPR 2023 · 678 citations)
What has been tried
108 papers here report 1,167 measured changes (a component added, removed or swapped, compared with a baseline): helped 71%, no clear effect 5%, hurt 20%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use a temporal Transformer instead of 1D temporal convolution | Video motion modeling · Motion incorporation | Helped | AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning (ICLR 2024 · 85 citations) | |
| Use SeeCoder-PA instead of SeeCoder | Image-Variation, no structural conditioning · image quality | Helped | Prompt-Free Diffusion: Taking "Text" out of Text-to-Image Diffusion Models (CVPR 2024 · 52 citations) | |
| add reward feedback | Six styles · FID | 84.04 -> 83.21 | Helped | SVGDreamer: Text Guided SVG Generation with Diffusion Model (CVPR 2024 · 41 citations) |
| Remove optimization and directly inject Gaussian-falloff activations into directed-word and trailing attention maps | Object placement and scene generation · placement and scene quality | Mixed | Directed Diffusion: Direct Control of Object Placement through Attention Guidance (AAAI 2024 · 40 citations) | |
| Add one round of SLD self-correction | LMD benchmark — Average · accuracy | + 26.0 | Helped | Self-correcting LLM-controlled Diffusion Models (CVPR 2024 · 38 citations) |
| add nonlinear mapping network F | Emotional image generation · FID | 57.54 -> 41.60 | Helped | EmoGen: Emotional Image Content Generation with Text-to-Image Diffusion Models (CVPR 2024 · 36 citations) |
| Add latent scaling to the encoded image | Pascal VOC2012 segmentation · mIoUss | 82.34 -> 83.06 | Helped | Text-Image Alignment for Diffusion-Based Perception (CVPR 2024 · 33 citations) |
| Add global structure guidance via SPM to RM, predicting image x | TII-ST · accuracy | 69.40 -> 71.73 | Helped | Text Image Inpainting via Global Structure-Guided Diffusion Models (AAAI 2024 · 29 citations) |
| Add InToken interaction tokens | HICO-DET test, Default, Full · mAP | 25.23 -> 28.73 | Helped | InteractDiffusion: Interaction Control in Text-to-Image Diffusion Models (CVPR 2024 · 26 citations) |
| Add the LoRA teacher during student training | COCO 2014 5K subset · FID | Helped | SwiftBrush: One-Step Text-to-Image Diffusion Model with Variational Score Distillation (CVPR 2024 · 25 citations) | |
| use explicit tile overlap instead of no overlap | MM-CelebA-HQ · PSNR | 26.89 -> 27.49 | Helped | Any-Size-Diffusion: Toward Efficient Text-Driven Synthesis for Any-Size HD Images (AAAI 2024 · 17 citations) |
| Disable unconditional guidance training | CXR generation · FID_XRV | 10.0 -> 16.4 | Hurt | Trade-Offs in Fine-Tuned Diffusion Models between Accuracy and Interpretability (AAAI 2024 · 14 citations) |
What reviewers will expect
Datasets and baselines named in the 52 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- T2I-CompBench · 4 of 52 papers
- LAION-5B · 3 of 52 papers
- LAION · 3 of 52 papers
- DreamBench · 3 of 52 papers
- MultiGen-20M · 3 of 52 papers
- WikiArt · 3 of 52 papers
- MS-COCO · 3 of 52 papers
- JourneyDB · 3 of 52 papers
Baselines
- ControlNet · 7 of 52 papers
- Textual Inversion · 7 of 52 papers
- SDXL · 5 of 52 papers
- DreamBooth · 4 of 52 papers
- IP-Adapter · 4 of 52 papers
- LoRA · 3 of 52 papers
- InstructPix2Pix · 3 of 52 papers
- BLIP-Diffusion · 3 of 52 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 137 papers; from OpenAlex and the papers' affiliations).
- Google · 8 papers
- Peking University · 6 papers
- ByteDance · 6 papers
- University of California, Berkeley · 5 papers
- NVIDIA · 4 papers
- Korea Advanced Institute of Science and Technology · 4 papers
- Tsinghua University · 4 papers
- The Chinese University of Hong Kong · 4 papers
- Snap Inc · 4 papers
- Stanford University · 4 papers
Use it
Download BibTeX of 32 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Compositional text-to-image guidance (194 papers)
- Personalized and region-controlled generation (136 papers)
- Layout-to-image control (164 papers)
- Multi-subject personalization (82 papers)
- Text generation and structured prediction (65 papers)
Elsewhere in the topic
- T2I evaluation and controllable adapters (149 papers)
- Inference-time T2I improvement (117 papers)
- Preference optimization for diffusion (82 papers)
- Concept erasure in diffusion models (76 papers)
- Multi-concept composition (56 papers)
- Fairness and bias in generative models (52 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).