Research map / Learning theory and algorithms / Prediction and decision losses
Numerically Accurate Hyperbolic Embeddings Using Tiling-Based Models
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 L-tiling model instead of Lorentz model | 2-dim hyperbolic embedding · Mammals | MAP: 0.855±0.013 -> 0.930±0.005 | Helped |
| use L-tiling-RSGD instead of Poincare | 2-dim embedding · WordNet Nouns | MAP: 0.124±0.001 -> 0.413±0.007 | Helped |
| use RSGD instead of SGD in L-tiling model (ablation) | L-tiling · Mammals | MAP: 0.892±0.031 -> 0.930±0.005 | Helped |
| use L-tiling-RSGD instead of Lorentz | 2-dim embedding · WordNet Verbs | MAP: 0.750±0.004 -> 0.746±0.004 | No clear effect |
| use group-based L-tiling instead of non-group H-tiling (ablation) | 2-dim tiling model · WordNet Nouns | MAP: 0.390±0.002 -> 0.413±0.007 | Helped |
| product of 2-dim L-tiling copies instead of single high-dim Lorentz | embedding · Gr-QC | MAP: 0.944±0.007 -> 0.953±0.002 | Mixed |
| store embedding in L-tiling model instead of Poincare at same size | Bio-yeast compression · Bio-yeast | MAP: 0.310 -> 0.873 | Helped |
| reduce fundamental domain floats from f64 to f16 (ablation) | L-tiling compression · Bio-yeast | MAP: 0.873 -> 0.873 | No clear effect |
Closest papers by meaning
- Representation Tradeoffs for Hyperbolic Embeddings (ICML 2018 · similarity 0.92)
- Hyperbolic Fine-Tuning for Large Language Models (NeurIPS 2025 · similarity 0.87)
- Low-distortion and GPU-compatible Tree Embeddings in Hyperbolic Space (ICML 2025 · similarity 0.87)
- Hyperbolic Embeddings of Supervised Models (NeurIPS 2024 · similarity 0.86)
- HELM: Hyperbolic Large Language Models via Mixture-of-Curvature Experts (NeurIPS 2025 · similarity 0.86)
- Weighted Embeddings for Low-Dimensional Graph Representation (AAAI 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.