Research map / Large language models / Agents, code and math
Universal Guideline-Driven Image Clustering via a Hybrid LLM Agent
Wenliang Zhong, Rob Barton, Lucas Goncalves, Kushal Kumar, Feng Jiang, Hehuan Ma, Yuzhi Guo, Vidit Bansal, Karim Bouyarmane, Junzhou 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 MST-based LLM Traversal merging after HDBSCAN (ablation) | GCPM-G · ImageNet-10 | ARI: 0.3 -> 72.1 | Improved |
| add MST Traversal after HDBSCAN on long-tail e-commerce (ablation) | GCPM-E · ABO-LC | ARI: 28.2 -> 51.5 | Improved |
| add MST Traversal on fine-grained clustering (ablation) | GCPM · fine-grained datasets | NMI: 1.0 to 3.2 NMI gain | Improved |
| add MST Traversal: recall vs precision (ablation) | HDBSCAN · ImageNet-10 | BCubed recall: 19.9 -> 62.3 recall, 99.7 -> 93.5 precision | Mixed |
| use GCPM guideline captions instead of image only (ablation) | GME-Qwen embedder · Card-Number | NMI: 71.9 -> 82.0 | Improved |
| use GCPM guideline captions instead of image only (ablation) | GME-Qwen embedder · Stanford Cars | NMI: 61.5 -> 86.2 | Improved |
| use GCPM guideline captions instead of standard captions (ablation) | GME-Qwen embedder · ImageNet-10 | NMI: 93.7 -> 96.7 | Improved |
| use MLLM embedder (GCPM-G) instead of text instruction encoder (GCPM-I) | GCPM · CIFAR-10 | ACC: 75.6 -> 94.1 | Improved |
| use LLM embedder E5-Mistral instead of GME on Card | GCPM · Card-Number | NMI: 82.0 -> 91.1 | Mixed |
| K-Means with known cluster count instead of HDBSCAN alone | GCPM-G · CIFAR-10 | ACC/NMI: 42.0 -> 94.1 | Improved |
| prompt optimization of merge guideline with PromptAgent (ablation) | MST Traversal · Oxford Flowers | B-Prec.: 90.1 -> 95.7 | Improved |
Closest papers by meaning
- Image Clustering with External Guidance (ICML 2024 · similarity 0.88)
- Multi-Modal Proxy Learning Towards Personalized Visual Multiple Clustering (CVPR 2024 · similarity 0.86)
- Language-Assisted Image Clustering Guided by Discriminative Relational Signals and Adaptive Semantic Centers (NeurIPS 2026 · similarity 0.86)
- Image Clustering Conditioned on Text Criteria (ICLR 2024 · similarity 0.85)
- Image Clustering via the Principle of Rate Reduction in the Age of Pretrained Models (ICLR 2024 · similarity 0.85)
- Customized Multiple Clustering via Multi-Modal Subspace Proxy Learning (NeurIPS 2024 · similarity 0.84)
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.
