Research map / Detection and segmentation / Open-set and few-shot detection
Zero-shot HOI Detection with MLLM-based Detector-agnostic Interaction Recognition
Shiyu Xuan, Dongkai Wang, Zechao Li, Jinhui Tang
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 spatial-aware pooling (SAP) (ablation) | baseline · HICO-DET UO Full | mAP: 39.24 -> 42.31 | Improved |
| add deterministic matching (DM) (ablation) | baseline · HICO-DET UO Full | mAP: 39.24 -> 40.50 | Improved |
| add pairwise spatial encoding to SAP (ablation) | Ours · HICO-DET UO Full | mAP: 41.98 -> 43.60 | Improved |
| add cross attention to SAP (ablation) | Ours · HICO-DET UO Full | mAP: 41.37 -> 43.60 | Improved |
| use SAP instead of UPT (ablation) | Ours · HICO-DET UO Full | mAP: 41.76 -> 43.60 | Improved |
| deterministic generation instead of simple prompt (ablation) | training-free MLLM · HICO-DET Full | mAP: 14.23 -> 31.50 | Improved |
| deterministic generation instead of multiple-choice prompt (ablation) | training-free MLLM · HICO-DET Full | mAP: 18.91 -> 31.50 | Improved |
| deterministic generation instead of in-context prompt (ablation) | training-free MLLM · HICO-DET Full | mAP: 19.57 -> 31.50 | Improved |
| use conditional likelihood as confidence for simple prompt (ablation) | training-free MLLM · HICO-DET Full | mAP: 14.23 -> 16.19 | Improved |
| add deterministic generation to SFT (ablation) | SFT MLLM · HICO-DET Full | mAP: 31.61 -> 39.87 | Improved |
| scale MLLM from 0.5B to 3B (ablation) | Ours · HICO-DET UO Full | mAP: 42.00 -> 43.60 | Improved |
| scale MLLM from 3B to 7B (ablation) | Ours · HICO-DET UO Full | mAP: 43.60 -> 45.99 | Improved |
Closest papers by meaning
- Locality-Aware Zero-Shot Human-Object Interaction Detection (CVPR 2025 · similarity 0.91)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.90)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.90)
- EZ-HOI: VLM Adaptation via Guided Prompt Learning for Zero-Shot HOI Detection (NeurIPS 2024 · similarity 0.90)
- Toward Open-Set Human Object Interaction Detection (AAAI 2024 · similarity 0.89)
- HORP: Human-Object Relation Priors Guided HOI Detection (CVPR 2025 · 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.
