Research map / Vision-language and multimodal / Multimodal LLMs
Interleaved image-text generation with MLLMs
Uses multimodal LLMs to jointly understand and generate images and text in context.
153 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2023–2026) · an approach in Multimodal LLMs, keyword cluster “mllm · instruction · llms”
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 140 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 6.3% to 6.5%.
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 · LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day (NeurIPS 2023 · 232 citations)
- 2024 · MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models (ICLR 2024 · 480 citations)
- 2025 · Towards General Visual-Linguistic Face Forgery Detection (CVPR 2025 · 20 citations)
- 2026 · Patho-AgenticRAG: Towards Multimodal Agentic Retrieval-Augmented Generation for Pathology VLMs via Reinforcement Learning (AAAI 2026 · 4 citations)
Earliest papers here
- LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day (NeurIPS 2023 · 232 citations)
- Repetition In Repetition Out: Towards Understanding Neural Text Degeneration from the Data Perspective (NeurIPS 2023 · 3 citations)
- MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models (ICLR 2024 · 480 citations)
What has been tried
49 papers here report 540 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 |
|---|---|---|---|---|
| add second-stage finetuning with curated detailed image descriptions | Detailed caption generation · Failure rate | 35% -> 2% | Helped | MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models (ICLR 2024 · 480 citations) |
| Add news-domain pre-training to InstructBLIP | NewsCLIPpings Merged/Balance test · All accuracy | 47.4 -> 49.3 | Helped | SNIFFER: Multimodal Large Language Model for Explainable Out-of-Context Misinformation Detection (CVPR 2024 · 78 citations) |
| Add multimodal learning to the CLIP image encoder without PFIG | Celeb-DF · AUC | 77.62 -> 80.79 | Helped | Towards General Visual-Linguistic Face Forgery Detection (CVPR 2025 · 20 citations) |
| Combine pre-training and test-image expectation normalization instead of no normalization | All datasets (Avg) · accuracy | 66.92 -> 67.44 | Helped | Visual Fact Checker: Enabling High-Fidelity Detailed Caption Generation (CVPR 2024 · 16 citations) |
| Add meditative tokens | PubTab1M-Det · precision | 93.5 -> 98.5 | Helped | TabPedia: Towards Comprehensive Visual Table Understanding with Concept Synergy (NeurIPS 2024 · 16 citations) |
| Add the Vague-to-Specific prompt strategy | Q-Eval-Image long prompts · SRCC | 0.591 -> 0.620 | Helped | Q-Eval-100K: Evaluating Visual Quality and Alignment Level for Text-to-Vision Content (CVPR 2025 · 12 citations) |
| add semi-supervised learning | RealSnow10K · BRISQUE | 22.52 -> 17.55 | Helped | SnowMaster: Comprehensive Real-world Image Desnowing via MLLM with Multi-Model Feedback Optimization (CVPR 2025 · 12 citations) |
| Use multi-task training rather than training on the corresponding task only | Pose comprehension, generation, estimation, and editing · R-Precision Top-1 / MPJPE | 18.1 -> 85.6; 8.4 -> 67.9; 22.4 -> 24.5; 13.0 -> 13.5; 342.7 -> 308.6; 97.4 -> 94.7; 310.8 -> 270.3 | Helped | UniPose: A Unified Multimodal Framework for Human Pose Comprehension, Generation and Editing (CVPR 2025 · 10 citations) |
| Increase model scale from 1B to 8B | Crello · Mjudge Win Rate (%) | 2.43% to 17.19% | Helped | AesthetiQ: Enhancing Graphic Layout Design via Aesthetic-Aware Preference Alignment of Multi-modal Large Language Models (CVPR 2025 · 8 citations) |
| initialize UniTok with pretrained CLIP weights rather than random initialization | ImageNet · zero-shot accuracy | 70.8 -> 78.6 | Helped | UniTok: a Unified Tokenizer for Visual Generation and Understanding (NeurIPS 2025 · 8 citations) |
| Add MedTrinity-25M alignment pretraining | VQA-RAD · average accuracy | 73.0 -> 75.0 (+2.0) | Helped | MedTrinity-25M: A Large-scale Multimodal Dataset with Multigranular Annotations for Medicine (ICLR 2025 · 8 citations) |
| Add AlignerNet (CLIP-supervised alignment) before end-to-end training versus direct end-to-end fine-tuning | MS-COCO FID · FID | Failed -> 11.30 | Helped | Kosmos-G: Generating Images in Context with Multimodal Large Language Models (ICLR 2024 · 6 citations) |
What reviewers will expect
Datasets and baselines named in the 140 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ImageNet · 5 of 140 papers
- GenEval · 5 of 140 papers
- SPAQ · 4 of 140 papers
- ShareGPT4V · 4 of 140 papers
- DocVQA · 4 of 140 papers
- ChartQA · 4 of 140 papers
- COCO · 4 of 140 papers
- FFHQ · 4 of 140 papers
Baselines
- GPT-4o · 47 of 140 papers
- Gemini 2.5 Pro · 14 of 140 papers
- BAGEL · 13 of 140 papers
- Qwen2.5-VL-7B · 12 of 140 papers
- Qwen2.5-VL-72B · 10 of 140 papers
- GPT-5 · 9 of 140 papers
- OmniGen2 · 8 of 140 papers
- Janus-Pro · 8 of 140 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 153 papers; from OpenAlex and the papers' affiliations).
- ByteDance · 5 papers
- Peking University · 5 papers
- University of Science and Technology of China · 5 papers
- NVIDIA · 4 papers
- Hong Kong University of Science and Technology · 4 papers
- Nanyang Technological University · 4 papers
- Tencent · 4 papers
- Zhejiang University · 4 papers
- Alibaba Group · 4 papers
- Beijing University of Posts and Telecommunications · 3 papers
Use it
Download BibTeX of 21 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
- Instruction-tuned multimodal assistants (199 papers)
- Benchmarks and analysis of MLLMs (181 papers)
- Unified any-modality language models (152 papers)
- Visual instruction tuning (68 papers)
Elsewhere in the topic
- Chain-of-thought multimodal reasoning (332 papers)
- VLM capability analysis (264 papers)
- Multimodal LLM benchmarks (252 papers)
- Spatial reasoning in VLMs (207 papers)
- Reinforcement learning for visual reasoning (163 papers)
- Attention-based visual token pruning (117 papers)
- Decoding-time hallucination mitigation (91 papers)
- Training-free token reduction for MLLMs (90 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).