Research map / Image generation / Fast diffusion sampling
Cones: Concept Neurons in Diffusion Models for Customized Generation
Zhiheng Liu, Ruili Feng, Kai Zhu, Yifei Zhang, Kecheng Zheng, Yu Liu, Deli Zhao, Jingren Zhou, Yang Cao
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 gradient-identified, shut concept neurons instead of Custom Diffusion | Customized generation · Single Subject | Text-alignment: 0.352 -> 0.361 | Improved |
| Use gradient-identified, shut concept neurons instead of Custom Diffusion | Customized generation · Single Subject | Image-alignment: 0.722 -> 0.725; DreamBooth 0.731; Textual Inversion 0.744 | Mixed |
| Concatenate subject-specific concept-neuron clusters rather than use Custom Diffusion | Multi-subject customized generation · Two Subjects | Text-alignment: 0.314 -> 0.337 | Improved |
| Concatenate subject-specific concept-neuron clusters rather than use Custom Diffusion | Multi-subject customized generation · Two Subjects | Image-alignment: 0.685 -> 0.698 | Improved |
| Concatenate subject-specific concept-neuron clusters rather than use Custom Diffusion | Multi-subject customized generation · Three Subjects | Text-alignment: 0.289 -> 0.301 | Improved |
| Concatenate subject-specific concept-neuron clusters rather than use Custom Diffusion | Multi-subject customized generation · Three Subjects | Image-alignment: 0.669 -> 0.685 | Improved |
| Concatenate subject-specific concept-neuron clusters rather than use Custom Diffusion | Multi-subject customized generation · Four Subjects | Text-alignment: 0.269 -> 0.285 | Improved |
| Concatenate subject-specific concept-neuron clusters rather than use Custom Diffusion | Multi-subject customized generation · Four Subjects | Image-alignment: 0.632 -> 0.653 | Improved |
| Use binary shutting rather than float32, float16, or quaternary concept-neuron updates (ablation) | Cones · Customized subject generation | Generation performance | No clear effect |
| Fine-tune concatenated concept neurons rather than learn multi-subject neurons from scratch (ablation) | Cones · Multi-subject generation | Generation performance | Improved |
| Concatenate concept-neuron clusters rather than use Custom Diffusion constraint optimization | Tuning-free multi-subject generation · Visual quality and subject-generation accuracy | Generation quality | Improved |
Later papers that compared with this method
6 later accepted papers report Cones2 as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- Concept Conductor: Orchestrating Multiple Personalized Concepts in Text-to-Image Synthesis (AAAI 2025 · better on 6 of 6 shared results)
- MultiBooth: Towards Generating All Your Concepts in an Image from Text (AAAI 2025 · better on 3 of 3 shared results)
- LaTexBlend: Scaling Multi-concept Customized Generation with Latent Textual Blending (CVPR 2025 · better on 3 of 3 shared results)
- ConceptSplit: Decoupled Multi-Concept Personalization of Diffusion Models via Token-wise Adaptation and Attention Disentanglement (ICCV 2025 · better on 2 of 2 shared results)
- MC^2: Multi-concept Guidance for Customized Multi-concept Generation (CVPR 2025 · better on 1 of 1 shared result)
- Customizable Image Synthesis with Multiple Subjects (NeurIPS 2023 · better on 7 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LINE: LLM-based Iterative Neuron Explanations for Vision Models (NeurIPS 2026 · similarity 0.85)
- Knowledge-Aware Neuron Interpretation for Scene Classification (AAAI 2024 · similarity 0.85)
- Automated Natural Language Explanation of Deep Visual Neurons with Large Models (Student Abstract) (AAAI 2024 · similarity 0.84)
- Understanding Distributed Representations of Concepts in Deep Neural Networks without Supervision (AAAI 2024 · similarity 0.84)
- Explainable Concept Generation through Vision-Language Preference Learning for Understanding Neural Networks' Internal Representations (ICML 2025 · similarity 0.84)
- The Hidden Language of Diffusion Models (ICLR 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.
