Epicure: Navigating the Emergent Geometry of Food Ingredient Embeddings
2026/05/21 by Jakub Radzikowski, Josef Chen · 27 voices
#cs.AI #cs.CL #cs.CY
paper · pdf
Abstract
We present Epicure, a family of three sibling skip-gram ingredient embeddings retrained from scratch on a multilingual recipe corpus. We aggregate 4.14M recipes from 11 sources spanning seven languages, English, Chinese, Russian, Vietnamese, Spanish, Turkish, Indonesian, German, and Indian-English, and normalise the raw ingredient strings to 1,790 canonical entries via an LLM-augmented pipeline. A 203,508-edge ingredient-ingredient NPMI graph and an 80,019-edge typed FlavorDB ingredient-compound graph, 2,247 typed compound nodes across 15 categories, seed three Metapath2Vec variants that share architecture and hyperparameters and differ only in the random-walk schema: Cooc walks the co-occurrence graph only, Chem walks the typed compound metapaths only, and Core blends both via injected ingredient-ingredient walks at controlled mixing, placing each model at a distinct point on the chemistry-vs-recipe-context spectrum.
Citations
Discussions
- All of human cooking compressed into 2 megabytes [hn, 444 points, 174 comments]
- Is there a link that others could use? Slightly related: arxiv.org/abs/2605.22391 [bsky, 5 points, 1 comments]
- From Josef Chen on the other site: "we trained the largest multilingual food model ever built. 4.1M recipes. 7 languages. 1,790 ingredients. 300 dimensions. All of human cooking compressed into 2 mega [bsky, 3 points, 1 comments]
- I've always said that the one thing neural networks will be good at is finding the gaps between pieces of existing knowledge where human brains, for whatever reason, haven't explored. A new neural net [bsky, 2 points, 1 comments]
- why limit just to breakfast-space? arxiv.org/abs/2605.22391 [bsky, 2 points, 0 comments]
- Imagine compressing the essence of human cooking into just 2 megabytes! This groundbreaking research explores the intersection of culinary art and data science, showcasing how far we've come in unders [bsky, 1 points, 0 comments]
- All of human cooking compressed into 2 megabytes [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes #HackerNews https://arxiv.org/abs/2605.22391 [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 https://news.ycombinator.com/item?id=48291225 [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 https://news.ycombinator.com/item?id=48291225 [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 comments #arxiv.org [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes View Article | Join the HN Conversation Summary of HN discussion 🧵👇 [bsky, 0 points, 1 comments]
- Epicure: Navigating the Emergent Geometry of Food Ingredient Embeddings arxiv.org/abs/2605.22391 [bsky, 0 points, 0 comments]
- 📰 All of human cooking compressed into 2 megabytes 🔗 https://arxiv.org/abs/2605.22391 💬 Discuss on HN [bsky, 0 points, 0 comments]
- [5/30] 361 Upvotes, 145 Comments, 2 Posts, arXiv:2605.22391 🆕Epicure: Navigating the Emergent Geometry of Food Ingredient Embeddings Jakub Radzikowski, Josef Chen [bsky, 0 points, 1 comments]
- e pra quem é dodói da cabeça igual eu: arxiv.org/pdf/2605.22391 [bsky, 0 points, 0 comments]
- https://bsky.app/profile/news.ycombinator.com.web.brid.gy/post/3mmtisgqbjmh2 [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 (http://news.ycombinator.com/item?id=48291225) [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 (http://news.ycombinator.com/item?id=48291225) [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https:// arxiv.org/abs/2605.22391 # arxiv [mastodon, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 [comments] [292 points] [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 (https://news.ycombinator.com/item?id=48291225) [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 (https://news.ycombinator.com/item?id=48291225) [bsky, 0 points, 0 comments]
- All of human cooking compressed into 2 megabytes https://arxiv.org/abs/2605.22391 (https://news.ycombinator.com/item?id=48291225) [bsky, 0 points, 0 comments]
- [2605.22391] Epicure: Navigating the Emergent Geometry of Food Ingredient Embeddings — Interesting example of how embedding geometry can surface structure in messy, real-world domains like ingredients [bsky, 0 points, 0 comments]
- Epicure: Navigating the Emergent Geometry of Food Ingredient Embeddings [bsky, 0 points, 0 comments]
Related