Research map / Vision-language and multimodal / Human and embodied modeling
Referring expression comprehension
Resolves referring expressions with one-step transformers, iterative scanning, and adaptive inference.
28 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Human and embodied modeling, keyword cluster “grounding · referring · expression”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 1 tracked benchmark.
8 papers in 2023–24 and 6 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 1.5% to 0.9%.
Open questions the data points to
- RefCOCO val (mIoU): best went from 77.43 (up to 2024) to 81.37 in 2025 — Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints (AAAI 2025 · 14 citations)
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 · Modeling Relationships in Referential Expressions With Compositional Modular Networks (CVPR 2017 · 410 citations)
- 2018 · MAttNet: Modular Attention Network for Referring Expression Comprehension (CVPR 2018 · 869 citations)
- 2019 · Improving Referring Expression Grounding With Cross-Modal Attention-Guided Erasing (CVPR 2019 · 204 citations)
- 2020 · Multi-Task Collaborative Network for Joint Referring Expression Comprehension and Segmentation (CVPR 2020 · 314 citations)
- 2021 · Referring Transformer: A One-step Approach to Multi-task Visual Grounding (NeurIPS 2021 · 72 citations)
- 2022 · LAVT: Language-Aware Vision Transformer for Referring Image Segmentation (CVPR 2022 · 397 citations)
- 2023 · Referring Expression Comprehension Using Language Adaptive Inference (AAAI 2023 · 23 citations)
- 2024 · Mask Grounding for Referring Image Segmentation (CVPR 2024 · 41 citations)
- 2025 · Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints (AAAI 2025 · 14 citations)
- 2026 · PC-CrossDiff: Point-Cluster Dual-Level Cross-Modal Differential Attention for Unified 3D Referring and Segmentation (AAAI 2026 · 0 citations)
Earliest papers here
- Modeling Relationships in Referential Expressions With Compositional Modular Networks (CVPR 2017 · 410 citations)
- A Joint Speaker-Listener-Reinforcer Model for Referring Expressions (CVPR 2017 · 284 citations)
- MAttNet: Modular Attention Network for Referring Expression Comprehension (CVPR 2018 · 869 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
RefCOCO val · mIoU (higher is better)
4 papers, 2021–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 66.75 | Referring Transformer: A One-step Approach to Multi-task Visual Grounding (NeurIPS 2021 · 72 citations) |
| 2024 | 77.43 | Mask Grounding for Referring Image Segmentation (CVPR 2024 · 41 citations) |
| 2025 | 81.37 | Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints (AAAI 2025 · 14 citations) |
What has been tried
26 papers here report 306 measured changes (a component added, removed or swapped, compared with a baseline): helped 75%, no clear effect 2%, hurt 21%, 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 Mask Grounding to LAVT | RefCOCO · oIoU | +1.44 | Helped | Mask Grounding for Referring Image Segmentation (CVPR 2024 · 41 citations) |
| Add score-based learning | GRES val · oIoU | 62.75 -> 64.98 | Helped | LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · 16 citations) |
| Add multimodal fusion pretraining (BEiT-3) | RefCOCO testB · Acc (REC) | +5.11 | Helped | Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints (AAAI 2025 · 14 citations) |
| Add referring-aware image masking instead of random masking, keeping vanilla MIM and MLM | RefCOCOg test · accuracy | 81.11 -> 82.16 | Helped | OneRef: Unified One-tower Expression Grounding and Segmentation with Mask Referring Modeling (NeurIPS 2024 · 11 citations) |
| Use the finer image-pyramid scales after the coarsest scale | RefCOCOg val · IoU | Helped | ScanFormer: Referring Expression Comprehension by Iteratively Scanning (CVPR 2024 · 10 citations) | |
| Add hierarchical word-object, phrase-object, and text-image alignments | gRefCOCO · Pr | 65.2 -> 67.8 | Helped | Hierarchical Alignment-enhanced Adaptive Grounding Network for Generalized Referring Expression Comprehension (AAAI 2025 · 8 citations) |
| add attribute features | C-RefCOCO testB · Acc-Box (%) | 75.23 -> 78.05 | Helped | Revisiting Counterfactual Problems in Referring Expression Comprehension (CVPR 2024 · 6 citations) |
| add vision-conditioned dynamic routing | RefCOCO+ testA · IoU@0.5 accuracy | 39.99 -> 56.84 | Helped | DViN: Dynamic Visual Routing Network for Weakly Supervised Referring Expression Comprehension (CVPR 2025 · 4 citations) |
| Use the density-map estimate to select the number of queries | REC-8K val · MAE | 4.83 -> 4.23 | Helped | Exploring Contextual Attribute Density in Referring Expression Counting (CVPR 2025 · 2 citations) |
| Use all annotation sentences instead of only sentences for the selected subject | HC-RefLoCo subject-specific evaluation · mAcc | Helped | A Large-Scale Human-Centric Benchmark for Referring Expression Comprehension in the LMM Era (NeurIPS 2024 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 10 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- RefCOCO · 8 of 10 papers
- RefCOCOg · 7 of 10 papers
- ReferItGame · 3 of 10 papers
- Flickr30K Entities · 2 of 10 papers
- MS-COCO · 2 of 10 papers
Baselines
- TransVG++ · 5 of 10 papers
- MCN · 4 of 10 papers
- MDETR · 4 of 10 papers
- LAVT · 4 of 10 papers
- ReLA · 4 of 10 papers
- CRIS · 4 of 10 papers
- OFA · 4 of 10 papers
- VLT · 3 of 10 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 14 papers; from OpenAlex and the papers' affiliations).
- Zhejiang University · 3 papers
Use it
Download BibTeX of 26 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
- Visual grounding transformers (63 papers)
- Region-level vision-language alignment (32 papers)
Elsewhere in the topic
- Vision-language pretraining (182 papers)
- Medical vision-language models (130 papers)
- Scene graph generation (126 papers)
- Language-grounded scene text and structure (124 papers)
- fMRI visual decoding (106 papers)
- Human-like visual cognition in VLMs (97 papers)
- Multimodal and affective interaction modeling (87 papers)
- Multimodal retrieval-augmented modeling (85 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).