Research map / Image generation / Text-to-image diffusion
Text generation and structured prediction
Generate text or images with structured prediction and controlled decoding.
65 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · 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.
13 papers in 2023–24 and 26 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 3.3% to 3.1%.
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
- 2017 · Phrase-Based Presentation Slides Generation for Academic Papers (AAAI 2017 · 18 citations)
- 2018 · Faithful to the Original: Fact Aware Neural Abstractive Summarization (AAAI 2018 · 294 citations)
- 2019 · Non-Monotonic Sequential Text Generation (ICML 2019 · 62 citations)
- 2020 · Unsupervised Text Generation by Learning from Search (NeurIPS 2020 · 36 citations)
- 2021 · BARTScore: Evaluating Generated Text as Text Generation (NeurIPS 2021 · 314 citations)
- 2022 · CoNT: Contrastive Neural Text Generation (NeurIPS 2022 · 22 citations)
- 2023 · Being Comes From Not-Being: Open-Vocabulary Text-to-Motion Generation With Wordless Training (CVPR 2023 · 33 citations)
- 2024 · Audio Generation with Multiple Conditional Diffusion Model (AAAI 2024 · 26 citations)
- 2025 · Zero-Shot Styled Text Image Generation, but Make It Autoregressive (CVPR 2025 · 9 citations)
- 2026 · Infinite-Story: A Training-Free Consistent Text-to-Image Generation (AAAI 2026 · 1 citation)
Earliest papers here
- Phrase-Based Presentation Slides Generation for Academic Papers (AAAI 2017 · 18 citations)
- Faithful to the Original: Fact Aware Neural Abstractive Summarization (AAAI 2018 · 294 citations)
- Table-to-Text Generation by Structure-Aware Seq2seq Learning (AAAI 2018 · 237 citations)
What has been tried
52 papers here report 534 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 4%, hurt 20%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| train one model on multiple control conditions instead of separate single-condition models | AudioCondition test · MAE | 0.267 -> 0.200 | Helped | Audio Generation with Multiple Conditional Diffusion Model (AAAI 2024 · 26 citations) |
| add large-scale motion pre-training | Mixamo test set (zero-shot) · FID | Helped | OMG: Towards Open-vocabulary Motion Generation via Mixture of Controllers (CVPR 2024 · 24 citations) | |
| Remove Module Reorganization from QAB | Generated QR images · error level | Hurt | Text2QR: Harmonizing Aesthetic Customization and Scanning Robustness for Text-Guided QR Code Generation (CVPR 2024 · 10 citations) | |
| Add contrastive loss | MIMIC-CXR · BLEU-4 | 0.111 -> 0.113 | Helped | DART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation (CVPR 2025 · 6 citations) |
| Generate from more-occluded parts to less-occluded parts rather than in inverse order | Complex-attributes 3D generation · BLIP-VQA | 0.5023 -> 0.7295 | Helped | Apply Hierarchical-Chain-of-Generation to Complex Attributes Text-to-3D Generation (CVPR 2025 · 5 citations) |
| Add T2I-generated images with a simple class prompt | PACS (SDG, Avg.) · accuracy | 46.60 -> 58.78 | Helped | Adversarial Domain Prompt Tuning and Generation for Single Domain Generalization (CVPR 2025 · 4 citations) |
| Increase training data from Infographics-65K to Infographics-650K | BizEval infographics, visual text spelling precision for ≤10 layers · visual text spelling precision | 86.47 -> 92.98 | Helped | BizGen: Advancing Article-level Visual Text Rendering for Infographics Generation (CVPR 2025 · 4 citations) |
| Use noun typography rather than other word types | VMT-IGMs-FM · CLIP Score increase | 5.28; 1.19 and 1.76 | Helped | Not Just Text: Uncovering Vision Modality Typographic Threats in Image Generation Models (CVPR 2025 · 1 citation) |
| add SVR+ to PCon | Extended ConsiStory benchmark ablation · CLIP-T | Helped | One-Prompt-One-Story: Free-Lunch Consistent Text-to-Image Generation Using a Single Prompt (ICLR 2025 · 1 citation) | |
| Increase the number of POS sequences changed (L) | TuringBench · obfuscation success | Helped | ALISON: Fast and Effective Stylometric Authorship Obfuscation (AAAI 2024 · 1 citation) | |
| Replace dynamic noise with static noise σ=0.016 | ShapeNet55 with BLIP-2 captions · FID | 26.94 -> 27.64 | Hurt | TextField3D: Towards Enhancing Open-Vocabulary 3D Generation with Noisy Text Fields (ICLR 2024 · 1 citation) |
| Use Fooocus-Inpainting rather than SD-Inpainting, ControlNet inpainting, or Paint By Example | Unsafe content injection dataset · ASR | Helped | ColJailBreak: Collaborative Generation and Editing for Jailbreaking Text-to-Image Deep Generation (NeurIPS 2024 · 1 citation) |
What reviewers will expect
Datasets and baselines named in the 17 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- The Chosen One · 2 of 17 papers
- PhotoMaker · 2 of 17 papers
- IP-Adapter · 2 of 17 papers
- ConsiStory · 2 of 17 papers
- StoryDiffusion · 2 of 17 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 39 papers; from OpenAlex and the papers' affiliations).
- Microsoft Research Asia · 3 papers
- Microsoft · 3 papers
- Chinese Academy of Sciences · 2 papers
- Huawei · 2 papers
- Google · 2 papers
- The Chinese University of Hong Kong · 2 papers
- Zhejiang University · 2 papers
- University of Science and Technology of China · 2 papers
- Harbin Institute of Technology · 2 papers
Use it
Download BibTeX of 38 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)
- Text-driven synthesis and semantic guidance (137 papers)
- Personalized and region-controlled generation (136 papers)
- Layout-to-image control (164 papers)
- Multi-subject personalization (82 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).