Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify
Authors: Edoardo D'Amico, Marco De Nadai, Praveen Chandar, Divita Vohra, Shawn Lin, Max Lefarov, Paul Gigioli, Gustavo Penha, Ilya Kopysitsky, Ivo Joel Senese, Darren Mei, Francesco Fabbri, Oguz Semerci, Yu Zhao, Vincent Tang, Brian St. Thomas, Alexandra Ranieri, Matthew N. K. Smith, Aaron Bernkopf, Bryan Leung, Ghazal Fazelnia, Mark VanMiddlesworth, Timothy Christopher Heath, Petter Pehrson Skiden, Alice Y. Wang, Doug J. Cole, Andreas Damianou, Maya Hristakeva, Reid Wilbur, Tarun Chillara, Vladan Radosavljevic, Pooja Chitkara, Sainath Adapa, Juan Elenter, Bernd Huber, Jacqueline Wood, Saaketh Vedantam, Jan Stypka, Sandeep Ghael, Martin D. Gould, David Murgatroyd, Yves Raimond, Mounia Lalmas, Paul N. Bennett
Podcast listening is often grounded in a set of favorite shows, while listener intent can evolve over time. This combination of stable preferences and changing intent motivates recommendation approaches that support both familiarity and exploration. Traditional recommender systems typically emphasize long-term interaction patterns, and are less explicitly designed to incorporate rich contextual signals or flexible, intent-aware discovery objectives. In this setting, models that can jointly reason over semantics, context, and user state offer a promising direction. Large Language Models (LLMs) provide strong semantic reasoning and contextual conditioning for discovery-oriented recommendation, but deploying them in production introduces challenges in catalog grounding, user-level personalization, and latency-critical serving. We address these challenges with GLIDE, a production-scale generative recommender for podcast discovery at Spotify. GLIDE formulates recommendation as an instruction-following task over a discretized catalog using Semantic IDs, enabling grounded generation over a large inventory. The model conditions on recent listening history and lightweight user context, while injecting long-term user embeddings as soft prompts to capture stable preferences under strict inference constraints. We evaluate GLIDE using offline retrieval metrics, human judgments, and LLM-based evaluation, and validate its impact through large-scale online A/B testing. Across experiments involving millions of users, GLIDE increases non-habitual podcast streaming on Spotify home surface by up to 5.4% and new-show discovery by up to 14.3%, while meeting production cost and latency constraints.
Figures & tables
Figure 1. We extend the vocabulary of a pre-trained LLM with SID tokens. Then, we undergo a two phase training. First, we ground the model to speak SIDs, along with the original languages. Then, we add a soft prompt user vector projection to add individual-level personalization and train the model for language-steerable podcast recommendations.
Figure 2. Structure of the GLIDE prompt. The model is conditioned on a hybrid of continuous soft prompts and discrete tokens.
Method
Recall@30
NDCG@30
Overall
LLM (SID only) (Baseline)
-
-
LLM (SID + Text)
+25.0%
+28.2%
GLIDE (Ours)
+29.9%
+31.2%
Non-Habitual but Familiar
LLM (SID only) (Baseline)
-
-
Table 1. Offline evaluation. Relative improvement (%) over the baseline across content familiarity segments.
SemID Space
HR@30
Intra-Bucket Similarity
RQ-VAE (Baseline)
–
0.657
R-LFQ
+4.76%
0.713
R-KMeans
+9.52%
0.856
Table 2. Ablation across SID quantization methods. R-KMeans achieves better accuracy and higher embedding cosine similarity within each group (sharing the same SID) where there is at least one collision.
Decoding Strategy
Scope
Δ Recall@30
Beam Search
Full Sequence (4 tokens)
–
Sampling
Full Sequence (4 tokens)
-27.07%
Sampling
Prefix Ceiling (3 tokens)
-26.17%
Sampling
Prefix Ceiling (2 tokens)
-12.53%
Table 3. Impact of different decoding strategies. We report the relative degradation when moving from Beam Search to Greedy Decoding. To analyze the source of the degradation, we provide the Prefix Ceiling (the maximum achievable Recall given the first N tokens of the generated SIDs).
Figure 3. Interest alignment scores from the LLM-judge across multiple model development iterations.
Generative recommendation models user histories and item catalogs to generate personalized items, lists, or queries rather than only scoring candidates.
Modern music streaming platforms face a persistent tradeoff: exploiting familiar content versus driving the exploration of novel items. While users frequently desire discovery, they hesitate to select unknown artists over proven favorites. Providing transparent, natural language rationales that explain why an unexplored item is recommended lowers this barrier. However, while Large Language Models (LLMs) excel at this nuanced explainability, their real-time deployment is severely bottlenecked by prohibitive inference costs and computational overhead. In this paper, we present an industry case study of a decoupled recommendation architecture that successfully scales exploration without compromising latency. Our system isolates LLM inference asynchronously offline, pre-computing personalized candidate pools of undiscovered artists alongside tailored rationales. Large-scale online A/B experiments validate our design. We demonstrate that combining LLM-backed recommendations with these explanatory rationales significantly reduces the trust barrier for new content, yielding statistically significant improvements in both user exploration and overall engagement on the discovery surfaces.
Recent advances in generative recommendation have leveraged pretrained LLMs by formulating sequential recommendation as autoregressive generation over a unified token space comprising language tokens and itemic identifiers, where each item is represented by a compact sequence of discrete tokens, namely Semantic IDs (SIDs). This SID-based formulation enables efficient decoding over large-scale item corpora and provides a natural interface for LLM-based recommenders to leverage rich world knowledge. Meanwhile, breakthroughs in LLM reasoning motivate reasoning-enhanced recommendation, yet effective reasoning over SIDs remains underexplored and challenging. Itemic tokens are not natively meaningful to LLMs; moreover, recommendation-oriented SID reasoning is hard to evaluate, making high-quality supervision scarce. To address these challenges, we propose SIDReasoner, a two-stage framework that elicits reasoning over SIDs by strengthening SID--language alignment to unlock transferable LLM reasoning, rather than relying on large amounts of recommendation-specific reasoning traces. Concretely, SIDReasoner first enhances SID-language alignment via multi-task training on an enriched SID-centered corpus synthesized by a stronger teacher model, grounding itemic tokens in diverse semantic and behavioral contexts. Building on this enhanced alignment, SIDReasoner further improves recommendation reasoning through outcome-driven reinforced optimization, which guides the model toward effective reasoning trajectories without requiring explicit reasoning annotations. Extensive experiments on three real-world datasets demonstrate the effectiveness of our reasoning-augmented SID-based generative recommendation. Beyond accuracy, the results highlight the broader potential of large reasoning models for generative recommendation, including improved interpretability and cross-domain generalization.
Yingzhi He, Yan Sun, Junfei Tan +6
National University of Singapore Singapore, Singapore · University of Science and Technology of China Hefei, China · Tencent Inc. Shenzhen, China
We introduce DeepInterestGR, a novel framework that integrates deep interest mining into the generative recommendation pipeline. This addresses the "Shallow Interest" problem - existing generative methods rely on surface-level textual features and fail to capture latent user motivations, limiting personalization depth and recommendation interpretability. Our approach leverages Multi-LLM Interest Mining (MLIM) via structured reasoning prompting, Reward-Labeled Deep Interest (RLDI) for quality control, and Interest-Enhanced Item Discretization (IEID) via RQ-VAE, combined with a two-stage SFT-GRPO training pipeline guided by an Interest-Aware Reward. We validate DeepInterestGR on three Amazon Review benchmarks (Beauty, Sports, Instruments), comparing against 14 state-of-the-art baselines including SASRec, BERT4Rec, TIGER, LC-Rec, and S-DPO. Our method achieves 5.8%-8.3% relative improvements on HR@10 and 7.7%-9.9% on NDCG@10 over the strongest baseline, with cross-domain generalization gains of +24.8%. These results provide evidence that incorporating deep semantic interests can effectively improve SID-based generative recommendation.