AI Paper Atlas

Research map / Benchmarks / RefCOCO

RefCOCO leaderboard

148 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on RefCOCO. Best accuracy (val): 94.8 (Phi3V-dist, ICLR 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

RefCOCO · accuracy (val)

55 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
194.8Phi3V-dist
Phi-3.5-vision
Teaching Metric Distance to Discrete Autoregressive Language Models
ICLR 2026 · Table 1
Language, safety and interpretability
294CofactGVRCofactGVR: Counterfactual Intervention for Grounded Visual Reasoning
ICML 2026 · Table 3
Multimodal LLMs
393.4Florence-2-LFlorence-2: Advancing a Unified Representation for a Variety of Vision Tasks
CVPR 2024 · Table 5
CLIP and zero-shot
492.76CogVLM-Grounding
Vicuna-7B
CogVLM: Visual Expert for Pretrained Language Models
NeurIPS 2024 · Grounding table
Human and embodied modeling
592.73SSRVG-B*
Swin-B
Look Around Before Locating: Considering Content and Structure Information for Visual Grounding
AAAI 2025 · Table 1
Human and embodied modeling
692.34CogCoM
CogVLM
CogCoM: A Visual Language Model with Chain-of-Manipulations Reasoning
ICLR 2025 · Table: Results on VG benchmark
Multimodal LLMs
791.5RoboInter-Qwen-7BRoboInter: A Holistic Intermediate Representation Suite Towards Robotic Manipulation
ICLR 2026 · Table 2
Robot action and navigation
891.42ForeSight (7B)
Qwen2.5-VL-7B
See Further, Think Deeper: Advancing VLM's Reasoning Ability with Low-level Visual Cues and Reflection
CVPR 2026 · Table 6
Multimodal LLMs
991.3Real-Model* (with RefCOCO/g/+)
Swin-B
Re-Aligning Language to Visual Objects with an Agentic Workflow
ICLR 2025 · Table 3
Multimodal LLMs
1091.1VideoChat-R1.5-7B
Qwen2.5-VL-7B
VideoChat-R1.5: Visual Test-Time Scaling to Reinforce Multimodal Reasoning by Iterative Perception
NeurIPS 2025 · spatial grounding table
Multimodal LLMs
45 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

RefCOCO · cIoU (val)

42 papers, each with its best reported number for its own method; higher is better.

#cIoU ↑MethodPaperTopic
186.4VIRSTVIRST: Video-Instructed Reasoning Assistant for SpatioTemporal Segmentation
CVPR 2026 · Table 3
Video segmentation and tracking
286.3SegCompass-13BSegCompass: Exploring Interpretable Alignment with Sparse Autoencoders for Enhanced Reasoning Segmentation
CVPR 2026 · Table 1
Multimodal LLMs
386.3ARGenSeg (ft)
InternVL2.5-8B
ARGenSeg: Image Segmentation with Autoregressive Image Generation Model
NeurIPS 2025 · Table 1
Multimodal LLMs
485.8FlowSegFlowSeg: Dynamic Semantic Guidance for LLM-Conditioned Segmentation
ICML 2026 · Table 1
Language, safety and interpretability
585.8ALToLLM-8B (AL)
InternVL-2.5-8B
ALTo: Adaptive-Length Tokenizer for Autoregressive Mask Generation
NeurIPS 2025 · Table: Performance comparison
Multimodal LLMs
685.8InstructSeg
Mipha-3B
InstructSeg: Unifying Instructed Visual Segmentation with Multi-modal Large Language Models
ICCV 2025 · Table 2
Video-language understanding
785.1X-SAM
Phi-3-3.8B
X-SAM: From Segment Anything to Any Segmentation
AAAI 2026 · Table 3
Open-set and few-shot detection
883.7RSAgent
Qwen2.5-VL-7B
RSAgent: Learning to Reason and Act via Multi-Turn Tool Invocations for Text-Guided Segmentation
ICML 2026 · Table 1
Multimodal LLMs
983.6PSALM
Phi-1.5 (1.3B)
PSALM: Pixelwise Segmentation with Large Multi-modal Model
ECCV 2024 · Table 7
Open-set and few-shot detection
1082.6FCLM-7B
Qwen2.5-VL
Hugging Visual Prompt and Segmentation Tokens: Consistency Learning for Fine-Grained Visual Understanding in MLLMs
CVPR 2026 · Table 1
Multimodal LLMs
32 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

RefCOCO · mIoU (val)

24 papers, each with its best reported number for its own method; higher is better.

#mIoU ↑MethodPaperTopic
185.72DeRIS-LDeRIS: Decoupling Perception and Cognition for Enhanced Referring Image Segmentation through Loopback Synergy
ICCV 2025 · Table 1
Open-set and few-shot detection
281.84UniRef++-L
Swin-L
Segment Every Reference Object in Spatial and Temporal Spaces
ICCV 2023 · Table 1
Video segmentation and tracking
381.37C³VG
BEiT3-ViT-B
Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints
AAAI 2025 · Main Results, RIS table
Human and embodied modeling
480.2DeRVOS
BEiT3-B
DeRVOS: Decoupling Consistent Trajectory Generation and Multimodal Understanding for Referring Video Object Segmentation
CVPR 2026 · Table 3
Video segmentation and tracking
579.2UniRES
CLIP
Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation
CVPR 2024 · Table 4
Open-set and few-shot detection
678.2Ours-LWIMFRIS: WIndow Mamba Fusion and Parameter Efficient Tuning for Referring Image Segmentation
ICLR 2026 · Table 2
CNN features
778.1Prompt-RIS
CLIP ViT-B/16 + SAM ViT-B/16
Prompt-Driven Referring Image Segmentation with Instance Contrasting
CVPR 2024 · Table 1
CLIP and zero-shot
877.43MagNet
Swin-B
Mask Grounding for Referring Image Segmentation
CVPR 2024 · Table 2
Human and embodied modeling
977.12TCRT
Swin-B
Task-aware Cross-modal Feature Refinement Transformer with Large Language Models for Visual Grounding
CVPR 2025 · Table 2
Human and embodied modeling
1076.94PolyFormer-L
Swin-L, BERT-base
PolyFormer: Referring Image Segmentation As Sequential Polygon Generation
CVPR 2023 · Table 1
Open-set and few-shot detection
14 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

RefCOCO · Acc@0.5 (val)

18 papers, each with its best reported number for its own method; higher is better.

8 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

RefCOCO · oIoU (val)

15 papers, each with its best reported number for its own method; higher is better.

5 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.