Research map / Probabilistic and scientific modeling / Variational and generative sampling
Model Comparison for Semantic Grouping
Francisco Vargas, Kamen Brestnichki, Nils Hammerla
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 |
|---|---|---|---|
| Model embedding magnitude with a diagonal Gaussian and AIC rather than a unit-length vMF with TIC (ablation) | Semantic grouping similarity · STS12–STS16 | Spearman correlation: 0.5762 -> 0.6764 (FastText); 0.5859 -> 0.6564 (GloVe); 0.6115 -> 0.6618 (Word2Vec GN) | Improved |
| Use diagonal instead of spherical Gaussian covariance (ablation) | Gaussian+AIC similarity · STS12–STS16 | Spearman correlation: 0.6312 -> 0.6764 (FastText); 0.6073 -> 0.6564 (GloVe); 0.6368 -> 0.6618 (Word2Vec GN) | Improved |
| Use AIC rather than BIC (ablation) | Gaussian model with GloVe embeddings · STS-12 to STS-16 | Spearman correlation: STS-12: 0.5009 -> 0.6031; STS-13: 0.6011 -> 0.6132; STS-14: 0.5926 -> 0.6445; STS-15: 0.6625 -> 0.7171; STS-16: 0.5826 - | Mixed |
| Use AIC rather than Bayes Factor (ablation) | Gaussian model with GloVe embeddings · STS-12 to STS-16 | Spearman correlation: 0.4592 -> 0.6031; 0.5687 -> 0.6132; 0.5829 -> 0.6445; 0.6627 -> 0.7171; 0.5389 -> 0.7346 | Improved |
| Use AIC rather than TIC for a diagonal Gaussian (ablation) | Semantic grouping similarity · STS12–STS16 | Spearman correlation: 0.6632 -> 0.6764 (FastText); 0.6536 -> 0.6564 (GloVe); 0.6363 -> 0.6618 (Word2Vec GN) | Improved |
| Use AIC rather than TIC for vMF (ablation) | Semantic grouping similarity · STS12–STS16 | Spearman correlation: 0.5762 vs 0.5726 (FastText); 0.5859 vs 0.5802 (GloVe); 0.6115 vs 0.6094 (Word2Vec GN) | No clear effect |
| Use Diag+AIC instead of SIF | Semantic grouping similarity · STS12–STS16 average | Spearman correlation: 0.6764 vs 0.6821 (FastText); 0.6564 vs 0.6255 (GloVe); 0.6618 vs 0.6588 (Word2Vec GN) | Mixed |
| Use Diag+AIC instead of MWV | Semantic grouping similarity · STS12–STS16 average | Spearman correlation: 0.6542 -> 0.6764 (FastText); 0.5784 -> 0.6564 (GloVe); 0.6451 -> 0.6618 (Word2Vec GN) | Improved |
| Use Diag+AIC instead of WMD | Semantic grouping similarity · STS12–STS16 average | Spearman correlation: 0.5997 -> 0.6764 (FastText); 0.5896 -> 0.6564 (GloVe); 0.6034 -> 0.6618 (Word2Vec GN) | Improved |
| Use Diag+AIC instead of SIF+PCA | Semantic grouping similarity · STS12–STS16 average | Spearman correlation: 0.6764 vs 0.6843 (FastText); 0.6564 vs 0.6589 (GloVe); 0.6618 vs 0.6639 (Word2Vec GN) | Mixed |
Closest papers by meaning
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · similarity 0.89)
- Semantic-KG: Using Knowledge Graphs to Construct Benchmarks for Measuring Semantic Similarity (NeurIPS 2025 · similarity 0.87)
- Analyzing Similarity Metrics for Data Selection for Language Model Pretraining (NeurIPS 2025 · similarity 0.87)
- Bayesian Test-Time Inference of Task-Aligned Similarity from Weak Interactive Feedback (NeurIPS 2026 · similarity 0.87)
- On the Emergence of Linear Analogies in Word Embeddings (NeurIPS 2025 · similarity 0.86)
- Narrowing the Gap between Supervised and Unsupervised Sentence Representation Learning with Large Language Model (AAAI 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.
