# SA-RSQ: A Versatile Sparse Representation Framework for Multi-modal Recommender Systems

arXiv: https://arxiv.org/abs/2608.22979 · HTML: https://mlsanity.com/paper/2608.22979/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 40 references that are in the map.
- First posted on arXiv 24 Aug 2026.

## Builds on (most cited first, 27 of 27)

- [SGDR: Stochastic Gradient Descent with Warm Restarts](https://mlsanity.com/paper/1608.03983/) (2016)
- [Neural Discrete Representation Learning](https://mlsanity.com/paper/1711.00937/) (2017)
- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation](https://mlsanity.com/paper/1308.3432/) (2013)
- [Generating Diverse High-Fidelity Images with VQ-VAE-2](https://mlsanity.com/paper/1906.00446/) (2019)
- [Autoregressive Image Generation using Residual Quantization](https://mlsanity.com/paper/2203.01941/) (2022)
- [Deep Interest Network for Click-Through Rate Prediction](https://mlsanity.com/paper/1706.06978/) (2017)
- [Deep Learning Recommendation Model for Personalization and Recommendation Systems](https://mlsanity.com/paper/1906.00091/) (2019)
- [Self-supervised Graph Learning for Recommendation](https://mlsanity.com/paper/2010.10783/) (2020)
- [Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations](https://mlsanity.com/paper/2402.17152/) (2024)
- [OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignment](https://mlsanity.com/paper/2502.18965/) (2025)
- [Sparse Sequence-to-Sequence Models](https://mlsanity.com/paper/1905.05702/) (2019)
- [Learning Discrete Representations via Information Maximizing Self-Augmented Training](https://mlsanity.com/paper/1702.08720/) (2017)
- [Contrastive Learning for Representation Degeneration Problem in Sequential Recommendation](https://mlsanity.com/paper/2110.05730/) (2021)
- [Learnable Item Tokenization for Generative Recommendation](https://mlsanity.com/paper/2405.07314/) (2024)
- [Controllable Multi-Interest Framework for Recommendation](https://mlsanity.com/paper/2005.09347/) (2020)
- [SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizer](https://mlsanity.com/paper/2412.10958/) (2024)
- [OneRec-Think: In-Text Reasoning for Generative Recommendation](https://mlsanity.com/paper/2510.11639/) (2025)
- [TokenRec: Learning to Tokenize ID for LLM-based Generative Recommendation](https://mlsanity.com/paper/2406.10450/) (2024)
- [Generative Recommender with End-to-End Learnable Item Tokenization](https://mlsanity.com/paper/2409.05546/) (2024)
- [PLUM: Adapting Pre-trained Language Models for Industrial-scale Generative Recommendations](https://mlsanity.com/paper/2510.07784/) (2025)
- [Eider: Empowering Document-level Relation Extraction with Efficient Evidence Extraction and Inference-stage Fusion](https://mlsanity.com/paper/2106.08657/) (2021)
- [QARM: Quantitative Alignment Multi-Modal Recommendation at Kuaishou](https://mlsanity.com/paper/2411.11739/) (2024)
- [Generative Recommendation with Semantic IDs: A Practitioner's Handbook](https://mlsanity.com/paper/2507.22224/) (2025)
- [FORGE: Forming Semantic Identifiers for Generative Retrieval in Industrial Datasets](https://mlsanity.com/paper/2509.20904/) (2025)
- [Time-based Sequence Model for Personalization and Recommendation Systems](https://mlsanity.com/paper/2008.11922/) (2020)

## Abstract (arXiv)

Deploying high-dimensional multimodal features in industrial recommender systems incurs substantial storage and latency overhead. Hard quantization is compact but introduces boundary distortion, whereas dense soft quantization couples representation quality to the limited storage budget. We propose Sparse Activation-based Residual Soft Quantization (SA-RSQ), which uses Top-K sparse routing and softmax weights to store compact (Index, Probability) tuples. The stored tuples decouple per-item storage from codebook dimensionality; for a fixed selected support, gradients propagate through the routing weights and weighted reconstruction without relying on a straight-through estimator. Experiments on a proprietary food-delivery advertising dataset show favorable reconstruction-performance and CTR trade-offs across storage budgets of 8-48 bytes per item. A preliminary Next-Distribution Prediction study and a one-week online A/B test further demonstrate the practical potential of SA-RSQ, with relative lifts of +2.51% in CTR and +3.66% in CPM.