Research map / Probabilistic and scientific modeling / Molecules and proteins
Vibe Spaces for Creatively Connecting and Expressing Visual Concepts
Huzheng Yang, Katherine Xu, Andrew Lu, Michael D. Grossberg, Yutong Bai, Jianbo Shi
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 Vibe Space instead of GPT Image 1 for blending | Vibe Blending · Totally Looks Like (high difficulty) | human preference rate: 20.0% -> 60.0% | Improved |
| use Vibe Space instead of GPT Image 1 for blending | Vibe Blending · Totally Looks Like (medium difficulty) | human preference rate: 21.4% -> 50.0% | Improved |
| use Vibe Space instead of GPT Image 1 for blending | Vibe Blending · Totally Looks Like (low difficulty) | human preference rate: 40.0% -> 26.7% | Hurt |
| use Vibe Space instead of CLIP Avg | Vibe Blending · Architecture | human preference rate: 39.0% -> 42.0% | Improved |
| use Vibe Space instead of GPT Image 1 as judged by GPT-5 | Vibe Blending · Totally Looks Like (low difficulty) | LLM preference rate: 66.7% -> 20.0% | Hurt |
| use Vibe Space instead of GPT Image 1 as judged by GPT-5 | Vibe Blending · Totally Looks Like (medium difficulty) | LLM preference rate: 50.0% -> 35.7% | Hurt |
| use Vibe Space instead of GPT Image 1 as judged by GPT-5 | Vibe Blending · Architecture | LLM preference rate: 30.3% -> 39.7% | Improved |
| use Vibe Space instead of GPT Image 1 as judged by GPT-5 | Vibe Blending · Totally Looks Like (high difficulty) | LLM preference rate: 40.0% -> 40.0% | No clear effect |
| use path nonlinearity score to predict blend difficulty | blend difficulty estimation · Totally Looks Like pairs | agreement with humans: 80.0% | Improved |
| expand human reference from top-1 to top-2 when comparing with LLM judge | GPT-5 judge · Totally Looks Like | agreement: 35.7 -> 55.1 | Improved |
Closest papers by meaning
- Bridging Cognitive Gap: Hierarchical Description Learning for Artistic Image Aesthetics Assessment (AAAI 2026 · similarity 0.84)
- An analytic theory of creativity in convolutional diffusion models (ICML 2025 · similarity 0.84)
- Learning to Infer Generative Template Programs for Visual Concepts (ICML 2024 · similarity 0.84)
- ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.84)
- The Hidden Language of Diffusion Models (ICLR 2024 · similarity 0.84)
- Bridging the gap to real-world language-grounded visual concept learning (NeurIPS 2025 · 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.
