Research map / Vision-language and multimodal / CLIP and zero-shot
AgentDet: A Shared-Blackboard Multi-Agent Framework for Zero-/Few-Shot Object Detection
Haolin Li, Yaohua Wang, Ze Yan, Lijie Wen, Biqing Huang
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 Q-Former fine-tuning (ablation) | AgentDet-Qwen2.5-7b · PASCAL VOC 10-shot | mAP: 8.2% drop when removed | Improved |
| add knowledge base integration (ablation) | AgentDet-Qwen2.5-7b · PASCAL VOC 10-shot | mAP: 19.1% decrease when removed | Improved |
| add LLM fine-tuning (LoRA) (ablation) | AgentDet-Qwen2.5-7b · PASCAL VOC 10-shot | mAP: drops to 0% when removed | Improved |
| add multi-scale segmentation at query stage (ablation) | AgentDet · PASCAL VOC 10-shot | mAP: 56.9 -> 65.1 | Improved |
| add multi-scale segmentation at training stage (ablation) | AgentDet · PASCAL VOC 10-shot | mAP: 55.4 -> 65.1 | Improved |
| add minimum enclosing rectangle (ablation) | AgentDet · PASCAL VOC 10-shot | mAP: 49.7 -> 65.1 | Improved |
| add filtering mechanism (ablation) | AgentDet · PASCAL VOC 10-shot | mAP: 61.3 -> 65.1 | Improved |
| use Qwen3-8B instead of Llama3.1-8B as LLM backbone | AgentDet · PASCAL VOC Novel Split 1, 10-shot | nAP50: 64.3 -> 69.6 | Improved |
| use AgentDet pipeline instead of plain fine-tuned Qwen2.5-VL-7b | Qwen2.5-VL-7B · PASCAL VOC zero-shot Split 1 | nAP50: 27.4 -> 33.8 | Improved |
| use AgentDet-Qwen3-8b vs LLMdet at COCO 30-shot | AgentDet-Qwen3-8b · COCO 30-shot | mAP: 37.8 -> 37.5 | No clear effect |
| use Llama3.1-8B backbone at COCO 30-shot | AgentDet · COCO 30-shot | mAP: 37.0 -> 31.7 | Hurt |
Closest papers by meaning
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- OVOD-Agent: A Markov-Bandit Framework for Proactive Visual Reasoning and Self-Evolving Detection (CVPR 2026 · similarity 0.88)
- GCD: Advancing Vision-Language Models for Incremental Object Detection via Global Alignment and Correspondence Distillation (AAAI 2025 · similarity 0.88)
- Zero-shot Generalizable Incremental Learning for Vision-Language Object Detection (NeurIPS 2024 · similarity 0.88)
- Agent-X: Evaluating Deep Multimodal Reasoning in Vision-Centric Agentic Tasks (ICLR 2026 · similarity 0.86)
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.
