Research map / Architectures and efficiency / RNNs and normalization
What the Vec? Towards Probabilistically Grounded Embeddings
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 a linear least-squares projection of PMI vectors instead of the W2V loss | Word embeddings trained on text8 · WordSim353 relatedness | Spearman correlation: .628 -> .727 | Helped |
| Use a linear least-squares projection of PMI vectors instead of the W2V loss | Word embeddings trained on text8 · WordSim353 similarity | Spearman correlation: .703 -> .791 | Helped |
| Use a linear least-squares projection of PMI vectors instead of the W2V loss | Word embeddings trained on text8 · Google analogy dataset | accuracy: .283 -> .425 | Helped |
| Tie the W and C embedding matrices (ablation) | LSQ word embeddings · WordSim353 relatedness | Spearman correlation: .727 -> .721 | Hurt |
| Tie the W and C embedding matrices (ablation) | LSQ word embeddings · WordSim353 similarity | Spearman correlation: .791 -> .786 | Hurt |
| Tie the W and C embedding matrices (ablation) | LSQ word embeddings · Google analogy dataset | accuracy: .425 -> .411 | Hurt |
Closest papers by meaning
- Word Embedding as Maximum A Posteriori Estimation (AAAI 2019 · similarity 0.90)
- Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models (NeurIPS 2025 · similarity 0.88)
- On the Emergence of Linear Analogies in Word Embeddings (NeurIPS 2025 · similarity 0.87)
- Analogies Explained: Towards Understanding Word Embeddings (ICML 2019 · similarity 0.87)
- Poincare Glove: Hyperbolic Word Embeddings (ICLR 2019 · similarity 0.86)
- By Tying Embeddings You Are Assuming the Distributional Hypothesis (ICML 2024 · 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.