URecJPQ: Memory-efficient Multimodal Recommendation Models through RecJPQ in Large-Scale Scenarios

• URL: http://arxiv.org/abs/2606.23291v1
• Date: Mon, 22 Jun 2026 13:05:03 GMT
• Title: URecJPQ: Memory-efficient Multimodal Recommendation Models through RecJPQ in Large-Scale Scenarios
• Abstract summary: We introduce URecJPQ, a Joint Product Quantization method specifically designed for large-scale and multimodal top-k recommendation tasks.<n>The core idea is to represent each user/item not as a fully learned, unique embedding, but rather as a concatenation of shared learned sub-embeddings.<n>Experiments on three widely-used datasets show that URecJPQ can be effectively applied to multimodal recommendation settings.
• Score: 52.726756906540636
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Training state-of-the-art recommendation models on large-scale industrial datasets can be a challenging task due to the high number of users and items which are typically represented through ID embeddings. Such embeddings typically require a large amount of memory resources, which are not always available. This problem is further exacerbated in multimodal recommendation, in which multimodal item features generally improve recommendation performance, but require more resources to encode. In this paper, we introduce URecJPQ, a Joint Product Quantization method specifically designed for large-scale and multimodal top-k recommendation tasks, in which the vast number of users and items, combined with the available modalities, further increases the memory demands for the computation. The core idea is to represent each user/item not as a fully learned, unique embedding, but rather as a concatenation of shared learned sub-embeddings, thereby significantly reducing the total number of trainable parameters. Our experiments on three widely-used datasets across different domains (movies, baby and sports products) show that URecJPQ can be effectively applied to multimodal recommendation settings. In large scale scenarios, we observe a substantial reduction in checkpoint sizes and the number of trainable parameters (ranging from 86% to 98%, and 98% to 99%, respectively), with only a marginal decrease in accuracy (8.5% on recall and 16% on NDCG, on average), and, in some cases, even performance improvements (up to 85%), as in the baby products domain. Our codebase is available at https://anonymous.4open.science/r/large_mmrecjpq-839B/README.md.

Related papers

• Exemplar Masking for Multimodal Incremental Learning [47.18796033633918]
  Multimodal incremental learning needs to digest the information from multiple modalities while concurrently learning new knowledge.<n>In this paper, we propose the exemplar masking framework to efficiently replay old knowledge.<n>We show that our exemplar masking framework is more efficient and robust to catastrophic forgetting under the same limited memory buffer.
  arXiv  Detail & Related papers  (2024-12-12T18:40:20Z)
• Mastering Collaborative Multi-modal Data Selection: A Focus on Informativeness, Uniqueness, and Representativeness [63.484378941471114]
  We propose a collaborative framework, DataTailor, which leverages three key principles--informativeness, uniqueness, and representativeness--for effective data selection.<n>Experiments on various benchmarks demonstrate that DataTailor achieves 101.3% of the performance of full-data fine-tuning with only 15% of the data.
  arXiv  Detail & Related papers  (2024-12-09T08:36:10Z)
• Train Once, Deploy Anywhere: Matryoshka Representation Learning for Multimodal Recommendation [27.243116376164906]
  We introduce a lightweight framework called full-scale Matryoshka representation learning for multimodal recommendation (fMRLRec) Our fMRLRec captures item features at different granularities, learning informative representations for efficient recommendation across multiple dimensions. We demonstrate the effectiveness and efficiency of fMRLRec on multiple benchmark datasets.
  arXiv  Detail & Related papers  (2024-09-25T05:12:07Z)
• Efficient Inference of Sub-Item Id-based Sequential Recommendation Models with Millions of Items [63.117573355917465]
  We show that it is possible to improve RecJPQ-based models' inference efficiency using the PQTopK algorithm. We speed up RecJPQ-enhanced SASRec by a factor of 4.5 x compared to the original SASRec's inference method and by a factor of 1.56 x compared to the method implemented in RecJPQ code.
  arXiv  Detail & Related papers  (2024-08-19T13:43:48Z)
• Scalable Dynamic Embedding Size Search for Streaming Recommendation [47.69328280479569]
  Real-world recommender systems often operate in streaming recommendation scenarios.<n>Number of users and items continues to grow, leading to substantial storage resource consumption.<n>We learn Lightweight Embeddings for streaming recommendation, called SCALL, which can adaptively adjust the embedding sizes of users/items.
  arXiv  Detail & Related papers  (2024-07-22T06:37:24Z)
• Budgeted Embedding Table For Recommender Systems [39.882452493920724]
  Budgeted Embedding Table (BET) is a novel method that generates table-level actions that are guaranteed to meet memory budgets.<n>BET is paired with three popular recommender models under different memory budgets.
  arXiv  Detail & Related papers  (2023-10-23T12:53:22Z)
• Parameter-efficient Tuning of Large-scale Multimodal Foundation Model [68.24510810095802]
  We propose A graceful prompt framework for cross-modal transfer (Aurora) to overcome these challenges. Considering the redundancy in existing architectures, we first utilize the mode approximation to generate 0.1M trainable parameters to implement the multimodal prompt tuning. A thorough evaluation on six cross-modal benchmarks shows that it not only outperforms the state-of-the-art but even outperforms the full fine-tuning approach.
  arXiv  Detail & Related papers  (2023-05-15T06:40:56Z)
• Continuous Input Embedding Size Search For Recommender Systems [60.89189829112067]
  Continuous input embedding size search (CIESS) is a novel RL-based method that operates on a continuous search space with arbitrary embedding sizes to choose from.<n> CIESS is also model-agnostic and hence generalizable to a variety of latent factor RSs.<n> experiments on two real-world datasets have shown state-of-the-art performance of CIESS under different memory budgets.
  arXiv  Detail & Related papers  (2023-04-07T06:46:37Z)
• M2TRec: Metadata-aware Multi-task Transformer for Large-scale and Cold-start free Session-based Recommendations [9.327321259021236]
  Session-based recommender systems (SBRSs) have shown superior performance over conventional methods. We propose M2TRec, a Metadata-aware Multi-task Transformer model for session-based recommendations.
  arXiv  Detail & Related papers  (2022-09-23T19:34:29Z)
• Learnable Embedding Sizes for Recommender Systems [34.98757041815557]
  We propose PEP (short for Plug-in Embedding Pruning) to reduce the size of the embedding table while avoiding the drop of recommendation accuracy. PEP achieves strong recommendation performance while reducing 97-99% parameters. PEP only brings an additional 20-30% time cost compared with base models.
  arXiv  Detail & Related papers  (2021-01-19T11:50:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.