GenRec: Towards LLM-Native Recommendation at Netflix
GenRec, Netflix's LLM-backed recommendation ranker, improves personalization by verbalizing user data and aligning with long-term goals, outperforming traditional models with fewer labeled examples and input signals.
MAIN POINTS
- GenRec uses large language models (LLMs) to verbalize user histories and item metadata for recommendations.
- The model aligns with long-term member satisfaction and business goals using reward-weighted objectives.
- GenRec outperforms traditional models in A/B tests with fewer labeled examples and input signals.
- It shifts focus from feature engineering to context engineering for efficient and effective recommendations.
TAKEAWAYS
- GenRec demonstrates the potential of LLMs in enhancing recommendation systems by reducing reliance on hand-crafted features.
- The model's architecture supports efficient scoring and cost-effective serving on Netflix's LLM infrastructure.
- Context engineering is crucial for maintaining recommendation quality while managing serving costs.
- LLM-native recommendation systems may become central to Netflix's personalization strategy, leveraging shared foundation backbones.