Rethinking Recommendation Paradigms: From Pipelines to Agentic Recommender Systems
Source: arXiv:2603.26100 · Published 2026-03-27 · 7 authors (Jinxin Hu, Hao Deng, Lingyu Mu, Hao Zhang, Shizhun Wang, Yu Zhang, et al.)
TL;DR
The paper proposes an Agentic Recommender System (AgenticRS) that replaces the static multi-stage pipeline (recall → ranking → re-ranking) with self-evolving agents, turning recommender systems from manually tuned stacks into systems capable of autonomous optimization and architectural evolution.
Background and Motivation
Large-scale industrial recommenders typically use a fixed multi-stage pipeline and have progressed from collaborative filtering to deep and large pre-trained models. However, both multi-stage and so-called One Model designs remain essentially static:
- Models are black boxes.
- System improvement relies on manual hypotheses and engineering.
- This is hard to scale under heterogeneous data and multi-objective business constraints.
- Unified perspective: reorganizes key recommender modules as agents within a comparable framework.
- Module promotion criteria: a module becomes an agent only when it (1) forms a functionally closed loop, (2) can be independently evaluated, and (3) possesses an evolvable decision space.
- Two self-evolution mechanisms for model agents: 1. Reinforcement-learning-style optimization in well-defined action spaces; 2. LLM-based generation and selection of new architectures and training schemes in open-ended design spaces.
- Individual vs. compositional evolution: distinguishes evolution of a single agent from evolution over how multiple agents are selected and connected.
- Layered inner/outer reward design that couples local module optimization with global business objectives.
- Open problems: evaluation credibility, latency and cost, hallucination and safety, cross-lingual and multimodal scaling.
- Datasets: MS MARCO, BEIR, Natural Questions, domain-specific corpora, public recommendation sets.
- Metrics: nDCG@10, MRR, Recall@k, Hit@k, human preference, task success rate, latency, token cost.
- Baselines: BM25, dense retrieval, cross-encoder re-ranking, retrieval-free LLMs, commercial search APIs.
- A Systematic Framework for Enterprise Knowledge Retrieval
- Retrieval Augmented Generation and Understanding in Vision: A Survey
- Synergizing RAG and Reasoning: A Systematic Review
- AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforcement Learning
- AgentX: Towards Agent-Driven Self-Iteration of Industrial Recommenders
- Agentic Information Retrieval
The paper targets this gap by redefining the responsibility boundaries among retrieval, ranking, generation, and tool calling in the LLM era, where users expect natural-language interaction, multi-hop reasoning, and up-to-date knowledge.
Core Contributions
Method / System Design
The general workflow follows four steps: problem formalization → model/system design → training or construction → inference pipeline.
1. Input & representation: encode queries, documents, and user context into dense/sparse representations or structured prompts. 2. Core modules: retrievers, re-rankers, planners, memory modules, and tool interfaces, chained or composed per task. 3. Learning strategies: supervised fine-tuning, contrastive learning, distillation, reinforcement learning (including process rewards), and bootstrapped data synthesis. 4. Inference strategies: single-pass retrieval, iterative retrieval, parallel sub-queries, early stopping, and budget control.
Experiments and Evaluation (typical setup for this line of work)
Note: quantitative results should be verified against the original PDF; this report is based on the abstract and public metadata.
Key Insights for Search / Rec / Personalization
1. Architecture: cascaded retrieval + re-ranking + generation remains mainstream, but the agentic paradigm makes *retrieval count and policy itself learnable*. 2. Data: high-quality instruction data and click/session logs are both critical; synthetic data must guard against leakage and distribution shift. 3. Evaluation: the gap between offline metrics and online satisfaction is widening; LLM-as-judge needs cross-validation with human evaluation. 4. Product: latency, cost, explainability, and safety are hard constraints for industrial deployment — academic benchmarks alone are insufficient.
Limitations and Future Work
Potential limitations include experiment scale bound by GPU budgets, benchmarks mismatching real user distributions, English-centric data limiting cross-lingual generalization, and safety risks of open-web agent systems. Future directions include more efficient test-time compute allocation, deeper integration with knowledge graphs and structured databases, and causal/fairness constraints for recommendation.
Related Work
Original Abstract
> Large-scale industrial recommenders typically use a fixed multi-stage pipeline (recall, ranking, re-ranking) and have progressed from collaborative filtering to deep and large pre-trained models. However, both multi-stage and so-called One Model designs remain essentially static: models are black boxes, and system improvement relies on manual hypotheses and engineering, which is hard to scale under heterogeneous data and multi-objective business constraints. We propose an Agentic Recommender System (AgenticRS) that reorganizes key modules as agents. Modules are promoted to agents only when they form a functionally closed loop, can be independently evaluated, and possess an evolvable decision space. For model agents, we outline two self-evolution mechanisms: reinforcement learning style optimization in well-defined action spaces, and large language model based generation and selection of new architectures and training schemes in open-ended design spaces. We further distinguish individual evolution of single agents from compositional evolution over how multiple agents are selected and connected, and use a layered inner and outer reward design to couple local optimization with global objectives. This provides a concise blueprint for turning static pipelines into self-evolving agentic recommender systems.