Research map / Vision-language and multimodal / Human and embodied modeling
Concept-guided image captioning
Improves captions by injecting semantic concepts, context guidance, and scene-graph evaluation.
69 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.
14 papers in 2023–24 and 18 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 2.6% to 2.7%.
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 · Image Captioning With Semantic Attention (CVPR 2016 · 1,739 citations)
- 2017 · Knowing When to Look: Adaptive Attention via a Visual Sentinel for Image Captioning (CVPR 2017 · 1,734 citations)
- 2018 · SemStyle: Learning to Generate Stylised Image Captions Using Unaligned Text (CVPR 2018 · 131 citations)
- 2019 · Look Back and Predict Forward in Image Captioning (CVPR 2019 · 155 citations)
- 2020 · More Grounded Image Captioning by Distilling Image-Text Matching Model (CVPR 2020 · 139 citations)
- 2021 · RSTNet: Captioning With Adaptive Attention on Visual and Non-Visual Words (CVPR 2021 · 279 citations)
- 2022 · End-to-End Transformer Based Model for Image Captioning (AAAI 2022 · 154 citations)
- 2023 · Positive-Augmented Contrastive Learning for Image and Video Captioning Evaluation (CVPR 2023 · 43 citations)
- 2024 · SECap: Speech Emotion Captioning with Large Language Model (AAAI 2024 · 38 citations)
- 2025 · Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · 6 citations)
- 2026 · BioCAP: Exploiting Synthetic Captions Beyond Labels in Biological Foundation Models (ICLR 2026 · 0 citations)
Earliest papers here
- Image Captioning With Semantic Attention (CVPR 2016 · 1,739 citations)
- Knowing When to Look: Adaptive Attention via a Visual Sentinel for Image Captioning (CVPR 2017 · 1,734 citations)
- Semantic Compositional Networks for Visual Captioning (CVPR 2017 · 487 citations)
What has been tried
59 papers here report 656 measured changes (a component added, removed or swapped, compared with a baseline): helped 78%, no clear effect 5%, hurt 15%, 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 |
|---|---|---|---|---|
| use HuBERT instead of HTSAT | EMOSpeech test · SIM1 | 7.26% | Helped | SECap: Speech Emotion Captioning with Large Language Model (AAAI 2024 · 38 citations) |
| add prediction entropy evaluation | MS-COCO uniform noise 80% · CIDEr | 87.0 -> 88.3 | Helped | Noise-Aware Image Captioning with Progressively Exploring Mismatched Words (AAAI 2024 · 10 citations) |
| Add global difference extraction and contextual interaction instead of direct feature subtraction | CLEVR-Change · CIDEr | 120.4 -> 125.2 | Helped | Region-aware Difference Distilling with Attribute-guided Contrastive Regularization for Change Captioning (AAAI 2025 · 5 citations) |
| add relationship learning and relationship-to-sentence modules | COCO-Shutterstock · CIDEr | 27.2 -> 38.5 | Helped | Relational Distant Supervision for Image Captioning without Image-Text Pairs (AAAI 2024 · 5 citations) |
| add visual inputs to the text-only model | LSMDC validation · Class accuracy | 64.45 -> 69.14 | Helped | MICap: A Unified Model for Identity-Aware Movie Descriptions (CVPR 2024 · 5 citations) |
| augment CLIPdir with source attributes (DS) | EditVal · s2AFC | 0.697 -> 0.629 | Hurt | Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing (CVPR 2025 · 3 citations) |
| Add Global-Part Optimal Transport | CLEVR-Change · BLEU-4 | 56.7 -> 58.7 | Helped | Revisiting Change Captioning from Self-supervised Global-Part Alignment (AAAI 2025 · 2 citations) |
| add sentence-level alignment loss (Ls) to VQ-GAN baseline | CUB-200 · FID | 5.51 -> 4.89 | Helped | Towards Improved Text-Aligned Codebook Learning: Multi-Hierarchical Codebook-Text Alignment with Long Text (CVPR 2025 · 2 citations) |
| Add CLIP matching to Grounding DINO | BACON benchmark · mIOU | 68.42 -> 80.34 | Helped | BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs (CVPR 2025 · 1 citation) |
| add image conditioning to tweet-text-only conditioning | Twitter alt-text test set · CIDEr | 1.738 -> 7.661 | Helped | Alt-Text with Context: Improving Accessibility for Images on Twitter (ICLR 2024 · 1 citation) |
| Remove the tau weighting terms | Label error detection datasets · Label error detection performance | about 1% | Hurt | LEMoN: Label Error Detection using Multimodal Neighbors (ICML 2025 · 0 citations) |
| add panoptic mask-pooled features | COCONut-PanCap val · CAPTURE | 58.7 -> 61.4 | Helped | COCONut-PanCap: Joint Panoptic Segmentation and Grounded Captions for Fine-Grained Understanding and Generation (NeurIPS 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 11 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- MS COCO · 3 of 11 papers
- CLEVR-Change · 2 of 11 papers
- Spot-the-Diff · 2 of 11 papers
- Visual Genome · 2 of 11 papers
Baselines
- DUDA · 2 of 11 papers
- NCT · 2 of 11 papers
- SCORER+CBR · 2 of 11 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 32 papers; from OpenAlex and the papers' affiliations).
- Amazon · 2 papers
- Chinese Academy of Sciences · 2 papers
- University of Science and Technology of China · 2 papers
- Northeastern University · 2 papers
- Ant Group · 2 papers
- Zhejiang University · 2 papers
Use it
Download BibTeX of 40 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)
- Multimodal retrieval-augmented modeling (85 papers)
- Composed and text-based image retrieval (78 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).