Epicure corpus resources provide the canonical vocabulary and validation results for three ingredient-embedding models. The dataset includes cross-modal validation against external USDA and FlavorDB labels, WEAT and Procrustes robustness checks, and a full SLERP direction-arithmetic result table. It was created by Kaikaku and last updated on May 27, 2026.
Use Cases
- Evaluating embedding model quality based on supervised direction-quality results
- Assessing model robustness based on WEAT and Procrustes checks
- Validating cross-modal alignment based on external USDA and FlavorDB labels
- Performing direction arithmetic operations based on the SLERP result table
- Analyzing factor alignment based on unsupervised factor-alignment tables
Strengths
- Includes cross-modal validation against external USDA and FlavorDB labels
- Contains a full SLERP direction-arithmetic result table
- Provides per-model GMM mode atlases
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- Kaikaku via Hugging Face
- Collection Method
- Companion dataset for three Epicure ingredient-embedding model repositories
- Freshness
- Last updated 2026-05-27 13:44:31