Research map / Detection and segmentation / Open-set and few-shot detection
LMC: Large Model Collaboration with Cross-assessment for Training-Free Open-Set Object Recognition
Haoxuan Qu, Xiaofei Hui, Yujun Cai, Jun Liu
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 |
|---|---|---|---|
| Use LMC instead of the strongest listed baseline, PMAL | Open-set recognition · CIFAR10 | AUROC: 95.1 ± - -> 96.6 ± 0.3 | Improved |
| Use LMC instead of the strongest listed baseline, MLS | Open-set recognition · CIFAR+10 | AUROC: 97.9 ± - -> 98.9 ± 0.7 | Improved |
| Use LMC instead of ZOC | Open-set recognition · CIFAR+50 | AUROC: 97.6 ± 0.0 -> 98.5 ± 0.4 | Improved |
| Use LMC instead of ZOC | Open-set recognition · TinyImageNet | AUROC: 84.6 ± 1.0 -> 86.7 ± 1.4 | Improved |
| Use LMC instead of ARPL+CS | Open-set recognition · CIFAR10 | OSCR: 87.9 ± 1.5 -> 93.6 ± 1.5 | Improved |
| Use LMC instead of ARPL+CS | Open-set recognition · CIFAR+10 | OSCR: 94.7 ± 0.7 -> 96.8 ± 0.7 | Improved |
| Use LMC instead of ARPL+CS | Open-set recognition · CIFAR+50 | OSCR: 92.9 ± 0.3 -> 96.4 ± 0.4 | Improved |
| Use LMC instead of ARPL+CS | Open-set recognition · TinyImageNet | OSCR: 65.9 ± 3.8 -> 80.6 ± 3.4 | Improved |
| Add CLIP alignment with class names (ablation) | LMC on TinyImageNet · TinyImageNet | AUROC: 84.4 -> 86.7 | Improved |
| Add DINO alignment with generated class images (ablation) | LMC on TinyImageNet · TinyImageNet | AUROC: 82.1 -> 86.7 | Improved |
| Add intermediate reasoning and self-checking to class-name prompts (ablation) | LMC on TinyImageNet · TinyImageNet | AUROC: 84.1 -> 86.7 | Improved |
| Add intermediate reasoning to class-name prompts (ablation) | LMC on TinyImageNet · TinyImageNet | AUROC: 85.6 -> 86.7 | Improved |
Closest papers by meaning
- Towards Open-Ended Visual Recognition with Large Language Models (ECCV 2024 · similarity 0.87)
- Open-Set Recognition in the Age of Vision-Language Models (ECCV 2024 · similarity 0.85)
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.85)
- VL-SAM-V2: Open-World Object Detection with General and Specific Query Fusion (NeurIPS 2025 · similarity 0.84)
- Improving Open Set Recognition via Visual Prompts Distilled from Common-Sense Knowledge (AAAI 2024 · similarity 0.83)
- All Beings Are Equal in Open Set Recognition (AAAI 2024 · 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.
