Unsupervised Graph Embeddings for Session-based Recommendation with Item Features (arXiv 2502.13763)

Overview

This forum post on zhichai.net catalogs an arXiv paper from February 2025:

• Title: Unsupervised Graph Embeddings for Session-based Recommendation with Item Features
• Authors: Andreas Peintner, Marta Moscati, Emilia Parada-Cabaleiro, Markus Schedl, Eva Zangerle
• Link: https://arxiv.org/abs/2502.13763
• Category: Sequential Recommendation

The paper studies session-based recommendation (SBRec), where systems must predict a user's next action from an anonymous, short session without long-term profile data. The proposed approach leverages unsupervised graph embeddings enriched with item features, aiming to reduce dependence on supervised training signals.

Key points from the post

• The post positions the paper within the broader landscape of large-scale search, recommendation, and personalization systems, where efficiency, scalability, and user-intent understanding remain core challenges.
• It provides a component-wise breakdown of modern recommender/search pipelines: representation learning, retrievers, rerankers, planners, generators, and feedback mechanisms.
• Typical evaluation setups discussed include datasets (e.g., MS MARCO, BEIR, Natural Questions, public recommendation datasets) and metrics such as nDCG@10, MRR, Recall@k, and Hit@k, with baselines like BM25 and dense retrieval.
• The post emphasizes open problems: evaluation trustworthiness, latency and cost, hallucination and safety, and cross-lingual / multimodal extension.

Engineering checklist (from the post)

| Area | Question | Suggestion | |------|----------|------------| | Data | PII in training/index? Versioning? | Partitioned indexes, anonymization, rollback-capable embeddings | | Latency | p99 budget? Number of retrieval steps? | Cascade with early stopping, cache popular queries, async reranking | | Quality | Do offline gains translate online? | Interleaving experiments, human audits, citation checks | | Safety | Does open retrieval introduce poisoning/bias? | Source allowlists, adversarial detection, output filtering | | Cost | Token and GPU cost per query? | Small-model routing, distillation, hybrid sparse+dense retrieval |

Related entries cited in the post

• BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations (DOI: 10.1145/3357384.3357895)
• EAGER: Two-Stream Generative Recommender with Behavior-Semantic Collaboration (DOI: 10.1145/3637528.3671775)
• Efficient On-Device Session-Based Recommendation (DOI: 10.1145/3580364)
• How to Index Item IDs for Recommendation Foundation Models, P5, SIGIR (DOI: 10.1145/3624918.3625339)
• LLMCDSR: Enhancing Cross-Domain Sequential Recommendation with Large Language Models (DOI: 10.1145/3715099)
• Mamba4Rec: Towards Efficient Sequential Recommendation with Selective State Space Models (arXiv: 2403.03900)

Note

The post itself is a templated summary based on the paper's abstract and public metadata; specific quantitative results should be verified against the original PDF.