Research map / Detection and segmentation / Open-set and few-shot detection
RIO: A Benchmark for Reasoning Intention-Oriented Objects in Open Environments
Mengxue Qu, Yu Wu, Wu Liu, Xiaodan Liang, Jingkuan Song, Yao Zhao, Yunchao Wei
Read the paperCheck 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 pretraining on noun grounding data (ablation) | MDETR · RIO detection | AP@0.5: 2.57 -> 48.16 | Improved |
| put confidence only on the word 'something' instead of whole sentence (ablation) | MDETR · RIO detection | AP@0.5: 27.6 -> 48.16 | Improved |
| add object noun [obj] into the sentence with confidence on [obj] (ablation) | MDETR · RIO detection | AP@0.5: 27.6 -> 32.64 | Improved |
| use [obj] sentence form instead of 'something' form (ablation) | MDETR · RIO detection | AP@0.5: 48.16 -> 32.64 | Hurt |
| add noun-pronoun distillation from [obj] to something (ablation) | TOIST vs MDETR · RIO detection | AP@0.5: 48.16 -> 49.00 | Improved |
| noun-pronoun distillation (TOIST vs MDETR), common set | TOIST · RIO common set | AP@0.5: 48.61 -> 49.05 | Improved |
| noun-pronoun distillation (TOIST vs MDETR), uncommon set | TOIST · RIO uncommon set | AP@0.5: 24.20 -> 21.96 | Hurt |
| replace COCO category names with synonyms (ablation) | GroundingDINO · COCO2014 val | AP: 48.4 -> 35.6 | Hurt |
| replace COCO category names with synonyms (ablation) | GroundingDINO · COCO2014 val | AP50: 64.6 -> 47.5 | Hurt |
| use end-to-end MDETR instead of detector-reasoner D-R* | MDETR · RIO common set | AP@0.5: 45.15 -> 48.61 | Improved |
| fine-tune GroundingDINO on COCO in detector-reasoner | D-R* · RIO uncommon set | Top1-Acc: 33.07 -> 41.98 | Improved |
| evaluate on uncommon intentions instead of common | MDETR · RIO | AP@0.5: 48.61 -> 24.20 | Hurt |
Closest papers by meaning
- Intent3D: 3D Object Detection in RGB-D Scans Based on Human Intention (ICLR 2025 · similarity 0.87)
- GREAT: Geometry-Intention Collaborative Inference for Open-Vocabulary 3D Object Affordance Grounding (CVPR 2025 · similarity 0.85)
- Seeing the Unseen: Visual Common Sense for Semantic Placement (CVPR 2024 · similarity 0.84)
- Rex-Thinker: Grounded Object Referring via Chain-of-Thought Reasoning (ICLR 2026 · similarity 0.84)
- NL-Eye: Abductive NLI For Images (ICLR 2025 · similarity 0.83)
- What’s in Common? Multimodal Models Hallucinate When Reasoning Across Scenes (NeurIPS 2025 · similarity 0.83)
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.
