Research map / Image generation / Autoregressive and latent generation
Toward Controlled Generation of Text
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 attribute discriminators rather than conditional reconstruction alone | Text generation with controlled sentiment · SST-full | sentiment generation accuracy: 0.822 -> 0.851 | Helped |
| Use attribute discriminators rather than conditional reconstruction alone | Text generation with controlled sentiment · SST-small | sentiment generation accuracy: 0.679 -> 0.707 | Helped |
| Use attribute discriminators rather than conditional reconstruction alone | Text generation with controlled sentiment · Lexicon | sentiment generation accuracy: 0.660 -> 0.701 | Helped |
| Augment classifier training with generated sentences using minimum entropy regularization | Sentiment classifier trained on Lexicon · Lexicon | classifier accuracy: 0.701 -> 0.733 | Helped |
| Add the minimum entropy regularization to generated training sentences (ablation) | H-reg sentiment classifier augmentation · Four sentiment datasets | classifier accuracy | Helped |
| Add conditioned sentiment codes to the generated classifier-training samples (ablation) | Sentiment classifier augmentation · Four sentiment datasets | classifier accuracy | Helped |
| Add the independency constraint so unmodeled attributes remain encoded in the unstructured code (ablation) | Controlled text generation · SST-full | attribute disentanglement | Helped |
Closest papers by meaning
- Educating Text Autoencoders: Latent Representation Guidance via Denoising (ICML 2020 · similarity 0.89)
- Plug and Play Language Models: A Simple Approach to Controlled Text Generation (ICLR 2020 · similarity 0.89)
- Content preserving text generation with attribute controls (NeurIPS 2018 · similarity 0.88)
- Dispersed Exponential Family Mixture VAEs for Interpretable Text Generation (ICML 2020 · similarity 0.88)
- Improved Variational Autoencoders for Text Modeling using Dilated Convolutions (ICML 2017 · similarity 0.87)
- Compressed and Smooth Latent Space for Text Diffusion Modeling (NeurIPS 2025 · similarity 0.85)
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.