RecTemp 2026
Temporal Reasoning in Recommender Systems
Why temporal reasoning?
RecTemp focuses on temporal reasoning in recommender systems, covering changing preferences, short-term and long-term behavior, temporal context integration, and the emerging role of temporal information in LLM-based recommendation pipelines.
Modeling how users evolve over time is essential for recommendation quality, personalization, and adaptability. RecTemp brings together researchers and practitioners interested in time-aware learning, sequential modeling, and emerging recommendation pipelines that reason over temporal behavior.
The workshop is planned as a half-day, in-person event with invited talks, paper presentations, and interactive discussion. It is designed to encourage focused exchange across academia and industry.
Topics of interest
- Sequential and session-based recommendation
- Time-aware user preference learning
- Evolving user preferences and cold-start
- Temporal context integration
- Cross-domain temporal patterns
- LLM-based temporal recommender systems
- Periodic and cyclic behavioral analysis
- Temporal side information and augmentation
TBA
TBA
Call for papers
RecTemp 2026 invites contributions on temporal reasoning in recommender systems, with particular interest in work that models evolving preferences, sequential interactions, contextual dynamics, and the growing role of LLMs in recommendation.
The workshop is dedicated to the exploration and advancement of temporal dynamics in recommender systems. As user preferences, intents, and contexts evolve over time, recommendation models must capture both short-term and long-term behavior to remain accurate, adaptive, and personalized across domains such as e-commerce, media, mobility, travel, and finance.
In this year we especially welcome work that examines how temporal reasoning interacts with large language models (LLMs) and foundation models. Recent LLM-based recommender pipelines increasingly rely on sequential interaction histories, evolving user preferences, and time-dependent contextual signals. RecTemp 2026 therefore aims to provide a focused venue for discussing not only time-aware recommendation methods, but also emerging generative and LLM-based paradigms that reason over user behavior across sessions and longer horizons.
Participants are invited to present novel methods, empirical studies, case studies, theoretical perspectives, and early-stage ideas that help advance temporal reasoning in recommender systems.
Topics of interest
- Case studies highlighting the critical role of temporal factors
- Temporal reasoning in LLM-based recommender systems
- Methods for integrating temporal data into recommendation algorithms
- Sequential, session-based, and time-aware recommendation models
- Solutions to cold-start using temporal data insights
- Cross-domain temporal patterns and temporal side information
- Personalization and group recommendation with temporal analysis
- Use of catalogs, interaction logs, and other diverse temporal data sources
Submission types
- Long papers: 6 to 8 pages plus additional pages for references
- Short papers: 4 pages plus additional pages for references
- Position and Demo papers: 2 pages plus additional pages for references
Guidelines
- Use the ACM Standard SIGCONF templates in two-column conference format
- Submissions are single-blind, so author names should be included
- Papers exceeding page limits or formatting guidelines will be returned without review
- Demos should provide links to the systems presented
- Previously published work should not be submitted unless it includes a significant addition
- An international panel of experts will review all submissions
Important dates
- Paper submission deadlineJuly 20, 2026
- Author notificationAugust 14, 2026
- Camera-ready versionAugust 28, 2026
Organizing committee
The workshop is organized by researchers from academia and industry with long-standing involvement in recommender systems and RecTemp.
Program committee
TBA
Past workshops
RecTemp has brought together researchers interested in temporal dynamics in recommender systems across multiple editions.