Research map / Probabilistic and scientific modeling / Variational and generative sampling
Context Selection for Embedding 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 |
|---|---|---|---|
| Context selection (CS-EFE, learned binary subset of context) instead of conditioning on all context elements | exponential family embeddings · MovieLens-100K, eBird-PA, Market-Basket | test log-likelihood | Helped |
| Context selection with β=20 (expected context size control) | exponential family embeddings · MovieLens-100K | test log-likelihood: -1.06 to -1.00 (K=10); -1.06 to -0.97 (K=50) | Helped |
| Context selection, larger improvement on eBird-PA | exponential family embeddings · eBird-PA | test log-likelihood: -1.74 to about -1.33/-1.34 at best β | Helped |
| Context selection with β=2 to 10 on Market-Basket | exponential family embeddings · Market-Basket | test log-likelihood: -0.632 to -0.623 (K=50) | Helped |
| Context selection (CS-EFE, β=20) on MovieLens embeddings | embedding quality vs genre similarity · MovieLens-100K | Spearman's correlation with genre similarity and mean-sim@5: 0.066 to 0.108 (Spearman); 0.272 to 0.328 (mean-sim@5) | Helped |
| Larger prior β (fewer expected selected context elements, approaching full EFE) (ablation) | CS-EFE · MovieLens-100K | test log-likelihood | Mixed |
Closest papers by meaning
- Dynamic Word Embeddings (ICML 2017 · similarity 0.89)
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · similarity 0.88)
- By Tying Embeddings You Are Assuming the Distributional Hypothesis (ICML 2024 · similarity 0.86)
- When is an Embedding Model More Promising than Another? (NeurIPS 2024 · similarity 0.86)
- EmbedLLM: Learning Compact Representations of Large Language Models (ICLR 2025 · similarity 0.86)
- IDEAL: Influence-Driven Selective Annotations Empower In-Context Learners in Large Language Models (ICLR 2024 · similarity 0.86)
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.