Research map / Vision-language and multimodal / Human and embodied modeling
Multimodal retrieval-augmented modeling
Uses cross-modal alignment, retrieval, and relation propagation for multimodal dialog and language modeling.
85 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Human and embodied modeling, keyword cluster “retrieval · scene · captioning”
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.
23 papers in 2023–24 and 37 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 4.3% to 5.5%.
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
- 2016 · SentiCap: Generating Image Descriptions with Sentiments (AAAI 2016 · 214 citations)
- 2017 · Words or Characters? Fine-grained Gating for Reading Comprehension (ICLR 2017 · 43 citations)
- 2018 · Adaptive Co-attention Network for Named Entity Recognition in Tweets (AAAI 2018 · 389 citations)
- 2019 · Multi-Interactive Memory Network for Aspect Based Multimodal Sentiment Analysis (AAAI 2019 · 254 citations)
- 2020 · Aspect-Aware Multimodal Summarization for Chinese E-Commerce Products (AAAI 2020 · 66 citations)
- 2021 · Multi-modal Graph Fusion for Named Entity Recognition with Targeted Visual Guidance (AAAI 2021 · 210 citations)
- 2022 · Open-Domain, Content-Based, Multi-Modal Fact-Checking of Out-of-Context Images via Online Resources (CVPR 2022 · 86 citations)
- 2023 · BERT-ERC: Fine-Tuning BERT Is Enough for Emotion Recognition in Conversation (AAAI 2023 · 38 citations)
- 2024 · UFineBench: Towards Text-based Person Retrieval with Ultra-fine Granularity (CVPR 2024 · 68 citations)
- 2025 · Human-centered Interactive Learning via MLLMs for Text-to-Image Person Re-identification (CVPR 2025 · 22 citations)
- 2026 · KPDM: Key Phrase Dynamic Masking for Robust Text-to-Image Person Retrieval (AAAI 2026 · 2 citations)
Earliest papers here
- SentiCap: Generating Image Descriptions with Sentiments (AAAI 2016 · 214 citations)
- Words or Characters? Fine-grained Gating for Reading Comprehension (ICLR 2017 · 43 citations)
- Adaptive Co-attention Network for Named Entity Recognition in Tweets (AAAI 2018 · 389 citations)
What has been tried
64 papers here report 712 measured changes (a component added, removed or swapped, compared with a baseline): helped 76%, no clear effect 6%, hurt 16%, 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 global alignment loss | CUHK-PEDES · R@1 | 68.45 -> 70.42 | Helped | UFineBench: Towards Text-based Person Retrieval with Ultra-fine Granularity (CVPR 2024 · 68 citations) |
| add image and knowledge graph inputs | PolitiFact, GossipCop, Pheme · accuracy | 0.848 -> 0.919; 0.806 -> 0.856; 0.802 -> 0.858 | Helped | Reinforced Adaptive Knowledge Learning for Multimodal Fake News Detection (AAAI 2024 · 56 citations) |
| add coarse and separate attribute words | Market1501 · mAP | 89.60 -> 89.40 | Hurt | Multi-Prompts Learning with Cross-Modal Alignment for Attribute-Based Person Re-identification (AAAI 2024 · 39 citations) |
| add patch selection | Flickr30K 1K · R@1 | 69.2 -> 74.0 | Helped | Linguistic-Aware Patch Slimming Framework for Fine-grained Cross-Modal Alignment (CVPR 2024 · 39 citations) |
| add object features | OKVQA-GS112K · P@5 | 46.9 -> 49.1 | Helped | Generative Multi-Modal Knowledge Retrieval with Large Language Models (AAAI 2024 · 24 citations) |
| add LoRA fine-tuning | CUHK-PEDES · Rank-1 | 76.38 -> 77.91 | Helped | Human-centered Interactive Learning via MLLMs for Text-to-Image Person Re-identification (CVPR 2025 · 22 citations) |
| Add point-cloud input to the camera inputs | MM-OR test · macro F1 | 0.484 -> 0.494 | Helped | MM-OR: A Large Multimodal Operating Room Dataset for Semantic Understanding of High-Intensity Surgical Environments (CVPR 2025 · 21 citations) |
| Use bottleneck size 8 instead of 4 | HatefulDiscussions · F1 | 0.877 -> 0.863 | Hurt | Multi-Modal Discussion Transformer: Integrating Text, Images and Graph Transformers to Detect Hate Speech on Social Media (AAAI 2024 · 18 citations) |
| remove instruction-tuning finetuning | Four-task average, Mistral-7b · performance | 44.58 -> 37.87 | Hurt | xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Token (NeurIPS 2024 · 18 citations) |
| Add hierarchical attention instead of conventional attention | MixATIS · overall accuracy | 51.8 -> 55.6 | Helped | Towards Multi-Intent Spoken Language Understanding via Hierarchical Attention and Optimal Transport (AAAI 2024 · 16 citations) |
| Add contextual semantic adaption | SA · W-F1 | 27.82 -> 31.54 | Helped | CAMEL: Capturing Metaphorical Alignment with Context Disentangling for Multimodal Emotion Recognition (AAAI 2024 · 15 citations) |
| Add retrieval | ICIP · MSE | 1.5614 -> 0.9662 | Helped | Improving Multimodal Social Media Popularity Prediction via Selective Retrieval Knowledge Augmentation (AAAI 2025 · 14 citations) |
What reviewers will expect
Datasets and baselines named in the 21 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CUHK-PEDES · 3 of 21 papers
- ICFG-PEDES · 3 of 21 papers
- RSTPReid · 3 of 21 papers
- HotpotQA · 2 of 21 papers
Baselines
- IRRA · 3 of 21 papers
- RDE · 3 of 21 papers
- APTM · 3 of 21 papers
- RoBERTa · 2 of 21 papers
- UMT · 2 of 21 papers
- VGG-16 · 2 of 21 papers
- RaSa · 2 of 21 papers
- AUL · 2 of 21 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 60 papers; from OpenAlex and the papers' affiliations).
- Peking University · 3 papers
- National University of Singapore · 3 papers
- Sichuan University · 3 papers
- Harbin Institute of Technology · 2 papers
- Chinese Academy of Sciences · 2 papers
- Xidian University · 2 papers
- Microsoft · 2 papers
- Tencent · 2 papers
- Tsinghua University · 2 papers
- Hubei Engineering Technology Research Center of Agricultural Big Data · 2 papers
Use it
Download BibTeX of 36 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
- Scene graph generation (126 papers)
- Language-grounded scene text and structure (124 papers)
- Composed and text-based image retrieval (78 papers)
- Concept-guided image captioning (69 papers)
- Neural image captioning (72 papers)
Elsewhere in the topic
- Vision-language pretraining (182 papers)
- Medical vision-language models (130 papers)
- fMRI visual decoding (106 papers)
- Human-like visual cognition in VLMs (97 papers)
- Multimodal and affective interaction modeling (87 papers)
- Language-grounded visual concept learning (83 papers)
- Layout-aware document understanding (73 papers)
- Human perception modeling (71 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).