Research map / Probabilistic and scientific modeling / Variational and generative sampling
Differentiable Perturb-and-Parse: Semi-Supervised Parsing with a Structured Variational Autoencoder
Caio Corro, Ivan Titov
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 semi-supervised VAE training with unlabeled sentences | Dependency parser · English test | UAS: ~88.79 -> 89.39 | Improved |
| Use semi-supervised VAE training with unlabeled sentences | Dependency parser · French test | UAS: 84.09 -> 84.43 | Improved |
| Use semi-supervised VAE training with unlabeled sentences | Dependency parser · Swedish test | UAS: 86.59 -> 86.92 | Improved |
| Remove latent sentence embedding z (ablation) | VAE w. z · English test | UAS: 89.39 -> 89.50 | Improved |
| Remove latent sentence embedding z (ablation) | VAE w. z · French test | LAS: 77.89 -> 78.49 | Improved |
| Remove latent sentence embedding z (ablation) | VAE w. z · Swedish test | LAS: 80.01 -> 79.80 | Hurt |
| Remove or heavily downweight the dependency-tree KL prior term (ablation) | Semi-supervised VAE · English, French, and Swedish | UAS | Improved |
| Use bi-affine attention in the parser (ablation) | Semi-supervised VAE · Low-resource parsing | UAS | No clear effect |
Closest papers by meaning
- Dependency Parsing is More Parameter-Efficient with Normalization (NeurIPS 2025 · similarity 0.86)
- Improving Unsupervised Constituency Parsing via Maximizing Semantic Information (ICLR 2025 · similarity 0.86)
- Discriminative Finetuning of Generative Large Language Models without Reward Models and Human Preference Data (ICML 2025 · similarity 0.86)
- Dynamic Syntactic Feature Filtering and Injecting Networks for Cross-lingual Dependency Parsing (AAAI 2025 · similarity 0.85)
- Nexus: Same Pretraining Loss, Better Downstream Generalization via Common Minima (NeurIPS 2026 · similarity 0.84)
- Latent Paraphrasing: Perturbation on Layers Improves Knowledge Injection in Language Models (NeurIPS 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.
