mlsanityData to 7 Oct 2026

Sensory-Aware Sequential Recommendation via Review-Distilled Representations

arXiv 2603.02709 ↗First posted 3 Mar 2026cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Sequential recommenders learn behavioral patterns from item identifiers, while the experiential properties that users describe in reviews, such as how products look, feel, smell, taste, or sound, rarely enter item representations in a controlled, auditable form. We present ASER (Attribute-based Sensory-Enhanced Representation), an offline pipeline that fine-tunes a large language model to extract evidence-grounded sensory attribute-value records, such as color: matte black or scent: vanilla, from review text and distills them into a compact student encoder that produces a frozen five-facet sensory bank for each item catalog. At recommendation time the pretrained backbone stays frozen: a lightweight relational metric between the user history and each candidate is learned over the bank, and its correction is applied within a validation-selected magnitude bound. Across five Amazon domains and four backbones, trained within a common experimental pipeline and evaluated by full-catalog leave-one-out ranking without sampled negatives, this integration improves HR@10 and NDCG@10 in all 20 domain-backbone pairs, with average relative gains of 6.1% and 6.4%. A matched non-sensory control channel, built with the same seed model, schema, and pipeline, separates the sources of the gain: the hit-rate improvement follows from structured, evidence-grounded extraction as such, whereas the sensory vocabulary yields a ranking-quality advantage in eight of nine matched comparisons. An audit of the Beauty evaluation catalog finds that 94.8% of retained records are supported by their cited evidence spans, so the extracted signal remains inspectable against its source text.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 41 of its 76 references that are in the map. First posted on arXiv 3 Mar 2026.

Builds on

41 of 76 references.

References · 41

  1. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  2. 2025Qwen3 Technical Reportcited by 8,195
  3. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  4. 2021DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharingcited by 1,219
  5. 2018Self-Attentive Sequential Recommendationcited by 1,169
  6. 2019BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformercited by 952
  7. 2015Image-based Recommendations on Styles and Substitutescited by 702
  8. 2015Session-based Recommendations with Recurrent Neural Networkscited by 685
  9. 2020S^3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximizationcited by 311
  10. 2022Unified Structure Generation for Universal Information Extractioncited by 220
  11. 2023Recommendation as Instruction Following: A Large Language Model Empowered Recommendation Approachcited by 168
  12. 2023Text Is All You Need: Learning Language Representations for Sequential Recommendationcited by 149
29 more
  1. 2023Where to Go Next for Recommender Systems? ID- vs. Modality-based Recommender Models Revisitedcited by 124
  2. 2022Towards Universal Sequence Representation Learning for Recommender Systemscited by 120
  3. 2017Joint Deep Modeling of Users and Items Using Reviews for Recommendationcited by 84
  4. 2021Personalized Transformer for Explainable Recommendationcited by 58
  5. 2024IDGenRec: LLM-RecSys Alignment with Textual ID Learningcited by 50
  6. 2021Non-invasive Self-attention for Side Information Fusion in Sequential Recommendationcited by 39
  7. 2024XRec: Large Language Models for Explainable Recommendationcited by 34
  8. 2022Decoupled Side Information Fusion for Sequential Recommendationcited by 30
  9. 2023An Attentive Inductive Bias for Sequential Recommendation beyond the Self-Attentioncited by 28
  10. 2023Schema-Driven Information Extraction from Heterogeneous Tablescited by 24
  11. 2024ADELIE: Aligning Large Language Models on Information Extractioncited by 24
  12. 2022A Systematic Review and Replicability Study of BERT4Rec for Sequential Recommendationcited by 22
  13. 2024RDRec: Rationale Distillation for LLM-based Recommendationcited by 21
  14. 2023Explainable Recommendation with Personalized Review Retrieval and Aspect Learningcited by 19
  15. 2025LLM2Rec: Large Language Models Are Powerful Embedding Models for Sequential Recommendationcited by 19
  16. 2024SLMRec: Distilling Large Language Models into Small for Sequential Recommendationcited by 18
  17. 2024Sequential Recommendation with Latent Relations based on Large Language Modelcited by 17
  18. 2024Distillation Matters: Empowering Sequential Recommenders to Match the Performance of Large Language Modelcited by 13
  19. 2024Improving Sequential Recommendations with LLMscited by 10
  20. 2024SSD4Rec: A Structured State Space Duality Model for Efficient Sequential Recommendationcited by 9
  21. 2024TiM4Rec: An Efficient Sequential Recommendation Model Based on Time-Aware Structured State Space Duality Modelcited by 9
  22. 2024FineRec:Exploring Fine-grained Sequential Recommendationcited by 8
  23. 2024Review-based Recommender Systems: A Survey of Approaches, Challenges and Future Perspectivescited by 8
  24. 2025Multi-Grained Patch Training for Efficient LLM-based Recommendationcited by 5
  25. 2025DIFF: Dual Side-Information Filtering and Fusion for Sequential Recommendationcited by 4
  26. 2025RecGPT: A Foundation Model for Sequential Recommendationcited by 4
  27. 2024MAPLE: Enhancing Review Generation with Multi-Aspect Prompt LEarning in Explainable Recommendationcited by 3
  28. 2024ASTE Transformer Modelling Dependencies in Aspect-Sentiment Triplet Extractioncited by 2
  29. 2024MARS: Matching Attribute-aware Representations for Text-based Sequential Recommendationcited by 1

Built on by

Not cited yet.