Research map / Vision-language and multimodal / Human and embodied modeling
Cops-Ref: A New Dataset and Task on Compositional Referring Expression Comprehension
Zhenfang Chen, Peng Wang, Lin Ma, Kwan-Yee K. Wong, Qi Wu
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add modular hard-negative mining (ablation) | MattNet · Cops-Ref Full | accuracy: 26.3 -> 33.8 | Improved |
| add modular hard-negative mining (ablation) | MattNet · Cops-Ref Cat | accuracy: 45.2 -> 54.4 | Improved |
| add modular hard-negative mining (ablation) | MattNet · Cops-Ref Cat&attr | accuracy: 42.5 -> 46.8 | Improved |
| add modular hard-negative mining (ablation) | MattNet · Cops-Ref Cat&cat | accuracy: 45.8 -> 52.0 | Improved |
| add modular hard-negative mining (ablation) | MattNet · Cops-Ref DiffCat | accuracy: 69.1 -> 70.5 | Improved |
| add modular hard-negative mining (ablation) | MattNet · Cops-Ref WithoutDist | accuracy: 77.9 -> 78.4 | Improved |
| use random negatives from the whole dataset (ablation) | MattNet · Cops-Ref Full | accuracy: 26.3 -> 27.6 | Improved |
| restrict randomly selected negatives to the same object category (ablation) | MattNet · Cops-Ref Cat | accuracy: 45.2 -> 53.2 | Improved |
| sample negatives by global sentence-feature similarity instead of class-aware random sampling (ablation) | MattNet · Cops-Ref Full | accuracy: 32.2 -> 32.3 | No clear effect |
| sample negatives by modular phrase similarity instead of global sentence-feature similarity (ablation) | MattNet · Cops-Ref Cat&cat | accuracy: 51.2 -> 52.0 | Improved |
| add distractor images | MattNet · Cops-Ref Full versus WithoutDist | accuracy: 77.9 -> 26.3 | Hurt |
| mask out the language input (ablation) | GroundeR · Cops-Ref Full | accuracy: 19.1 -> 2.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MattNet — MAttNet: Modular Attention Network for Referring Expression Comprehension (CVPR 2018) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Tasks (ICLR 2026 · similarity 0.94)
- RefDetector: A Simple Yet Effective Matching-based Method for Referring Expression Comprehension (AAAI 2025 · similarity 0.91)
- Rex-Thinker: Grounded Object Referring via Chain-of-Thought Reasoning (ICLR 2026 · similarity 0.90)
- ScanFormer: Referring Expression Comprehension by Iteratively Scanning (CVPR 2024 · similarity 0.90)
- Revisiting Counterfactual Problems in Referring Expression Comprehension (CVPR 2024 · similarity 0.89)
- A Large-Scale Human-Centric Benchmark for Referring Expression Comprehension in the LMM Era (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
