generative recommendation
In plain wordsA newer approach to recommendation that uses language-model-style AI to predict the next item a person will want, one step at a time, much as a chatbot predicts the next word.
An approach to recommender systemsrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on. that casts recommendation as sequence generation: a transformertransformer, n.: The design behind today's chatbots and much of modern AI. It handles text by letting every word look at every other word and decide which of them matter for understanding it. trained on users' interaction histories, treated as token sequences, predicts or generates the next item or action, instead of scoring candidates with separately engineered retrieval and ranking models. Two main lines exist. Generative retrieval gives each item a Semantic ID, a short sequence of discrete codes learned from item content (for example by residual vector quantizationvector quantization, n.: Compressing data by replacing each item with the nearest entry from a small catalog of representative examples, so that only the catalog number needs to be stored or sent.), and decodes the ID of the next item autoregressively in place of approximate nearest neighbor searchapproximate nearest neighbor search, n.: A family of shortcuts for finding the most similar items in a huge collection without checking every one — accepting an occasional miss in exchange for answers thousands of times faster. over embeddings; because similar items share code prefixes, new items can be recommended more easily (cold start problemcold start problem, n.: The trouble a recommendation system has with brand-new users or items: with no history to learn from, it has little basis for good suggestions.). Generative recommenders in the sense of Meta's HSTU architecture treat retrieval and ranking as sequential transduction over long, high-cardinality streams of user actions and report quality improving as a power lawpower law, n.: A rule linking two quantities in which scaling one up by a given factor always changes the other by the same multiple, wherever you start. Plotted on logarithmic axes, it is a straight line. in training compute, which conventional deep recommendation models had failed to show. Serving cost, latency and fast-changing catalogs remain practical obstacles.
Our pilot decodes Semantic IDs for the next three videos instead of scoring a million candidates.
Origin
Shashank Rajput and colleagues at Google introduced generative retrieval with Semantic IDs for recommendation in a May 2023 preprint, 'Recommender Systems with Generative Retrieval'. Jiaqi Zhai and colleagues at Meta named their framework 'Generative Recommenders' in 'Actions Speak Louder than Words' (February 2024; ICML 2024), reporting a 1.5-trillion-parameter model deployed in production. Other large platforms, among them Kuaishou with OneRec, reported similar systems in 2025.
Related
- recommender systemrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on.
- two-tower modeltwo-tower model, n.: A design for recommendation and search in which one neural network turns users or queries into lists of numbers and another does the same for items, so good matches can be found by quick comparison.
- transformertransformer, n.: The design behind today's chatbots and much of modern AI. It handles text by letting every word look at every other word and decide which of them matter for understanding it.
- next-token predictionnext-token prediction, n.: The simple task most AI chatbots were trained on: given some text, guess the next small piece of it. Repeating that guess, piece by piece, produces whole answers.
- vector quantizationvector quantization, n.: Compressing data by replacing each item with the nearest entry from a small catalog of representative examples, so that only the catalog number needs to be stored or sent.
- scaling lawsscaling laws, n.: Measured patterns showing that an AI model's error shrinks smoothly and predictably as the model is made bigger and given more data and computing power, so labs can forecast a giant model's performance from small, cheap experiments.
- cold start problemcold start problem, n.: The trouble a recommendation system has with brand-new users or items: with no history to learn from, it has little basis for good suggestions.
Sources
- Rajput, S. et al. (2023). Recommender Systems with Generative Retrieval. Advances in Neural Information Processing Systems 36 (NeurIPS 2023). arXiv:2305.05065. arxiv.org/abs/2305.05065
- Zhai, J. et al. (2024). Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations. Proc. ICML 2024. arXiv:2402.17152. arxiv.org/abs/2402.17152
Mentioned in
click-through rate predictionclick-through rate prediction, n.: Estimating how likely a particular person is to click on a particular ad, search result or recommended item: a prediction made billions of times a day to decide what each user is shown., vector quantizationvector quantization, n.: Compressing data by replacing each item with the nearest entry from a small catalog of representative examples, so that only the catalog number needs to be stored or sent.