netflix·Jul 30, 2026·4 min readCurated summary

GenRec: Towards LLM-Native Recommendation at Netflix

GenRec is Netflix’s LLM-backed recommendation ranker, designed to reduce dependence on thousands of hand-engineered features. It verbalizes user history, item metadata, and context, then post-trains a Netflix-adapted foundation model for catalog-aware ranking and long-term member value. In large-scale A/B testing, it reportedly improved both short- and long-term metrics while using far fewer labeled examples and input signals than an established production ranker. ## Motivation for an LLM-Native Recommender - Netflix’s existing recommendation stack supports many content types and product surfaces but is costly to extend. - New use cases can require substantial feature engineering, architectural changes, infrastructure work, and experimentation. - LLMs offer: - Shared semantic representations for users, items, and interactions - Natural-language steering through prompts - Rich understanding of content and user histories - General-purpose LLMs are not production-ready on their own because they may: - Over-recommend globally popular titles - Hallucinate items outside the catalog - Ignore business constraints - Provide limited personalization ## Ranking Problem and Long-Term Utility - GenRec ranks the full Netflix catalog, or a provided candidate set. - It uses the user, interaction history, current context, and time to produce a personalized ranking. - The optimization target is expected long-term member utility—a proxy for satisfaction and retention—rather than only immediate clicks or plays. ## Two-Phase Training ### Netflix-Adapted Foundation Model - Netflix first adapts an open-source LLM using proprietary Netflix data. - The model learns: - Netflix content and metadata - Member behavior and preference patterns - General language understanding and generation - This relatively stable foundation can support multiple Netflix applications. ### GenRec Post-Training - A second training phase specializes the foundation model for recommendation. - It focuses on ranking quality, steering, reward alignment, and serving-cost constraints. - This phase is refreshed more frequently to reflect new content and changing member preferences. ## Interaction Data as Conversations - Hundreds of billions of events—including views, play duration, feedback, add-to-list actions, and abandons—are converted into single- or multi-turn recommender conversations. - Each user message includes verbalized: - Context and profile - Interaction history - Item metadata - The recommendation task - Assistant messages represent actual member behavior, such as titles played, viewing duration, or feedback. - During inference, GenRec uses the verbalized request and a catalog-aware scoring head; it does not generate conversational responses. - The conversational format mainly supports language-model training and comprehension of rich textual inputs. ## Context Engineering Instead of Feature Engineering - GenRec represents histories and context as natural language rather than relying primarily on dense, manually designed features. - The token budget becomes the new feature budget, so histories are selectively compressed: - Retain detailed, high-signal events such as long plays and thumbs-up - Remove low-signal events such as brief plays and hovers - Summarize repetitive behavior such as binge-watching - Add detail for important or cold-start items, including new releases - Recent and high-value interactions receive priority, while older information is compressed or dropped. - Prompt structure is also optimized for shared prefixes and prefix caching, reducing serving costs. ## Ranking, Language, and Reward Objectives - GenRec combines several training objectives: - **Catalog-aware ranking:** Uses high-value engagements as positive labels, applies denoising and thresholds, and trains the model with cross-entropy over the catalog or candidate set. - **Language modeling:** Preserves understanding of natural-language histories and metadata and supports potential future features such as recommendation explanations. - **Reward-weighted alignment:** Incorporates business requirements and longer-term satisfaction into training. - Reward signals can help balance content categories such as movies, series, games, live content, and podcasts instead of optimizing only immediate engagement. ## Serving and Results - GenRec runs in prefill-only mode on Netflix’s LLM serving infrastructure, avoiding the cost of decoding generated text. - A catalog-aware scoring head converts the model’s representations into item rankings. - Compared with a mature production ranker, GenRec achieved statistically significant improvements in short- and long-term online metrics. - It did so with a small fraction of the labeled data and input signals used by the existing system. GenRec suggests that recommendation systems can shift from extensive manual feature construction toward careful context engineering, LLM post-training, and reward alignment. The approach is most promising when paired with catalog constraints, efficient serving, and objectives that reflect long-term member value rather than raw engagement alone.

Read original(opens in new tab)

Recommendation SystemsLarge Language ModelsNatural Language ProcessingReinforcement Learning+2

netflix·Oct 25, 2025Original article

Post-Training Generative Recommenders with Advantage-Weighted Supervised Finetuning | by Netflix Technology Blog | Netflix TechBlog (opens in new tab)

Netflix is evolving its recommendation systems by moving beyond simple behavior imitation toward generative recommenders that better align with true user preferences. While generative models like HSTU and OneRec effectively capture sequential user patterns, they often struggle to distinguish between habitual clicks and genuine satisfaction. To bridge this gap, Netflix developed Advantage-Weighted Supervised Fine-tuning (A-SFT), a post-training method that leverages noisy reward signals to refine model performance without the need for complex counterfactual data. ### The Shift to Generative Recommenders * Modern generative recommenders (GRs), such as HSTU and OneRec, utilize transformer architectures to treat recommendation as a sequential transduction task. * The models are typically trained using next-item prediction, where the system learns to imitate the chronological sequence of a user’s activities. * A significant drawback of this "behavior cloning" approach is that it captures external trends and noise rather than long-term user satisfaction, potentially recommending content the user finished but did not actually enjoy. ### Barriers to Reinforcement Learning in RecSys * Traditional post-training methods used in Large Language Models, such as Proximal Policy Optimization (PPO) or Direct Preference Optimization (DPO), require counterfactual feedback that is difficult to obtain in recommendation contexts. * Because user sequences span weeks or years, it is impractical to generate and test hypothetical, counterfactual experiences for real-time user validation. * Reward signals in recommendation systems are inherently noisy; for instance, high watch time might indicate interest, but it can also be a result of external circumstances, making it an unreliable metric for optimization. ### Advantage-Weighted Supervised Fine-tuning (A-SFT) * A-SFT is a hybrid approach that sits between offline reinforcement learning and standard supervised fine-tuning. * The algorithm incorporates an advantage function to weight training examples, allowing the model to prioritize actions that lead to higher rewards while filtering out noise from the reward model. * This method is specifically designed to handle high-variance reward signals, using them as directional guides rather than absolute truth, which prevents the model from over-exploiting inaccurate data. * Benchmarks against other representative methods show that A-SFT achieves superior alignment between the generative recommendation policy and the underlying reward model. For organizations managing large-scale recommendation engines, A-SFT offers a practical path to implementing post-training improvements. By focusing on advantage-weighted signals, developers can improve recommendation quality using existing implicit feedback—like watch time and clicks—without the infrastructure hurdles of online reinforcement learning.

Recommendation SystemsAIMachine LearningTransformer+4

netflix·Oct 21, 2025Original article

Behind the Streams: Real-Time Recommendations for Live Events Part 3 | by Netflix Technology Blog | Netflix TechBlog (opens in new tab)

Netflix manages the massive surge of concurrent users during live events by utilizing a hybrid strategy of prefetching and real-time broadcasting to deliver synchronized recommendations. By decoupling data delivery from the live trigger, the system avoids the "thundering herd" effect that would otherwise overwhelm cloud infrastructure during record-breaking broadcasts. This architecture ensures that millions of global devices receive timely updates and visual cues without requiring linear, inefficient scaling of compute resources. ### The Constraint Optimization Problem To maintain a seamless experience, Netflix engineers balance three primary technical constraints: time to update, request throughput, and compute cardinality. * **Time:** The specific duration required to coordinate and push a recommendation update to the entire global fleet. * **Throughput:** The maximum capacity of cloud services to handle incoming requests without service degradation. * **Cardinality:** The variety and complexity of unique requests necessary to serve personalized updates to different user segments. ### Two-Phase Recommendation Delivery The system splits the delivery process into two distinct stages to smooth out traffic spikes and ensure high availability. * **Prefetching Phase:** While members browse the app normally before an event, the system downloads materialized recommendations, metadata, and artwork into the device's local cache. * **Broadcasting Phase:** When the event begins, a low-cardinality "at least once" message is broadcast to all connected devices, triggering them to display the already-cached content instantaneously. * **Traffic Smoothing:** This approach eliminates the need for massive, real-time data fetches at the moment of kickoff, distributing the heavy lifting of data transfer over a longer period. ### Live State Management and UI Synchronization A dedicated Live State Management (LSM) system tracks event schedules in real time to ensure the user interface stays perfectly in sync with the production. * **Dynamic Adjustments:** If a live event is delayed or ends early, the LSM adjusts the broadcast triggers to preserve accuracy and prevent "spoilers" or dead links. * **Visual Cues:** The UI utilizes "Live" badging and dynamic artwork transitions to signal urgency and guide users toward the stream. * **Frictionless Playback:** For members already on a title’s detail page, the system can trigger an automatic transition into the live player the moment the broadcast begins, reducing navigation latency. To support global-scale live events, technical teams should prioritize edge-heavy strategies that pre-position assets on client devices. By shifting from a reactive request-response model to a proactive prefetch-and-trigger model, platforms can maintain high performance and reliability even during the most significant traffic peaks.

Recommendation SystemsDistributed SystemsCachingScalability+4