Research map / Probabilistic and scientific modeling / Variational and generative sampling
MixPoet: Diverse Poetry Generation via Learning Controllable Mixed Latent Space
Xiaoyuan Yi, Ruoyu Li, Cheng Yang, Wenhao Li, Maosong Sun
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 factor mixtures in the latent space rather than an unsupervised learned style | MixPoet-AUS · Chinese poetry, intra-topic diversity | intra-JS: 5.01% -> 3.73% | Improved |
| Use factor mixtures to distinguish poems generated from different keywords | MixPoet-AUS · Chinese poetry, inter-topic diversity | inter-JS: 1.89% -> 1.39%; MRL: 1.28% | Mixed |
| Use MixPoet-AUS rather than MRL | Chinese poetry · Human evaluation, Set 1 | Overall Quality: 3.55 -> 3.98$^{**}$ | Improved |
| Use MixPoet-AUS rather than fBasic with specified factor mixtures | Chinese poetry · Human evaluation, Set 2 | Relevance: 2.36 -> 2.92$^{**}$ | Improved |
| Specify factor mixtures while generating poems | MixPoet · Human evaluation, Set 1 versus Set 2 | Relevance: 3.39 -> 2.92$^{**}$ | Hurt |
| Replace adversarial universal latent subspaces with isotropic Gaussian latent subspaces (ablation) | MixPoet · Chinese poetry, intra-topic diversity | intra-JS: 3.73% -> 8.35% | Hurt |
| Replace adversarial universal latent subspaces with isotropic Gaussian latent subspaces (ablation) | MixPoet · Chinese poetry, inter-topic diversity | inter-JS: 1.39% -> 1.55% | Hurt |
| Replace adversarial universal latent subspaces with isotropic Gaussian latent subspaces (ablation) | MixPoet · Chinese poetry | LMS: 0.39 -> 0.37 | Hurt |
| Use reinforcement learning to penalize frequent words | MRL · Chinese poetry | inter-JS: 1.39% -> 1.28% | Improved |
| Increase TF-IDF words to promote diversity | MRL · Chinese poetry | Topic Relevance: 3.39 -> 3.19 | Hurt |
Closest papers by meaning
- PoetryDiffusion: Towards Joint Semantic and Metrical Manipulation in Poetry Generation (AAAI 2024 · similarity 0.85)
- DiverseFlow: Sample-Efficient Diverse Mode Coverage in Flows (CVPR 2025 · similarity 0.81)
- Alchemist: Turning Public Text-to-Image Data into Generative Gold (NeurIPS 2025 · similarity 0.81)
- NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers (ICLR 2024 · similarity 0.81)
- Quality-Diversity through AI Feedback (ICLR 2024 · similarity 0.81)
- Compositional Generalization via Forced Rendering of Disentangled Latents (ICML 2025 · similarity 0.80)
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.
