Research map / Image generation / Text-to-image diffusion
Breaking Semantic Boundaries: Distribution-Guided Semantic Exploration for Creative Generation
Fu Feng, Yucheng Xie, Ruixiao Shi, Xu Yang, Jing Wang, Xin Geng
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 distribution consistency supervision (VLM-based) (ablation) | DisTok · CangJie class distributions | KL divergence (lower is better): 0.0732 -> 0.0602 | Improved |
| use DisTok instead of BASS and CreTok for pair fusion | Class-pair creative generation · CangJie text-pair (TP2O) | VQAScore: 0.667 / 0.695 -> 0.840 | Improved |
| use DisTok instead of BASS and CreTok | Class-pair creative generation · CangJie text-pair (TP2O) | ImageReward: 0.387 / 1.018 -> 1.168 | Improved |
| use DisTok instead of FLUX for class-distribution generation | Distribution-conditional generation · GPT-4o creativity | Originality: 8.0 -> 9.8 | Improved |
| single-pass token decoding replaces iterative sampling and filtering | DisTok vs BASS · Class-pair generation | time per concept: ~40s -> ~3s | Improved |
| replace gradient-based token search with encoder-decoder sampling | DisTok vs ConceptLab · Unconditional exploration | time per concept: ~120s -> ~3s | Improved |
| latent-space regularization enabling sampling from Gaussian, Laplace or Cauchy | DisTok · Unconditional exploration | diversity (qualitative) | Improved |
| use DisTok tokens versus GPT-4o-derived text prompts | Kandinsky · Prompt engineering comparison | semantic coherence (qualitative) | Improved |
| user study: DisTok vs baselines on creativity votes | DisTok · User study, class distribution (vs FLUX) | votes: 396:104 vs FLUX | Improved |
Closest papers by meaning
- Enhancing Creative Generation on Stable Diffusion-based Models (CVPR 2025 · similarity 0.89)
- Redefining <Creative> in Dictionary: Towards an Enhanced Semantic Understanding of Creative Generation (CVPR 2025 · similarity 0.89)
- DreamDistribution: Learning Prompt Distribution for Diverse In-distribution Generation (ICLR 2025 · similarity 0.88)
- Seeking the Unfamiliar but Memorable: Conceptual Creativity as Meta-Learning (NeurIPS 2026 · similarity 0.88)
- Progressive Text-to-Image Diffusion with Soft Latent Direction (AAAI 2024 · similarity 0.88)
- The Hidden Language of Diffusion Models (ICLR 2024 · similarity 0.88)
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.
