Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ProductResearch: Training E-Commerce Deep Research Agents via Multi-Agent Synthetic Trajectory Distillation

Jiangyuan Wang, Kejun Xiao, Huaipeng Zhao, Tao Luo, Xiaoyi Zeng

TL;DR

ProductResearch is proposed, a multi-agent framework that synthesizes high-fidelity, long-horizon tool-use trajectories for training robust e-commerce shopping agents and establishes multi-agent synthetic trajectory training as an effective and scalable paradigm for enhancing LLM-based shopping assistance.

Abstract

Large Language Model (LLM)-based agents show promise for e-commerce conversational shopping, yet existing implementations lack the interaction depth and contextual breadth required for complex product research. Meanwhile, the Deep Research paradigm, despite advancing information synthesis in web search, suffers from domain gaps when transferred to e-commerce. We propose ProductResearch, a multi-agent framework that synthesizes high-fidelity, long-horizon tool-use trajectories for training robust e-commerce shopping agents. The framework employs a User Agent to infer nuanced shopping intents from behavioral histories, and a Supervisor Agent that orchestrates iterative collaboration with a Research Agent to generate synthetic trajectories culminating in comprehensive, insightful product research reports. These trajectories are rigorously filtered and distilled through a reflective internalization process that consolidates multi-agent supervisory interactions into coherent single-role training examples, enabling effective fine-tuning of LLM agents for complex shopping inquiries. Extensive experiments show that a compact MoE model fine-tuned on our synthetic data achieves substantial improvements over its base model in response comprehensiveness, research depth, and user-perceived utility, approaching the performance of frontier proprietary deep research systems and establishing multi-agent synthetic trajectory training as an effective and scalable paradigm for enhancing LLM-based shopping assistance.

ProductResearch: Training E-Commerce Deep Research Agents via Multi-Agent Synthetic Trajectory Distillation

TL;DR

ProductResearch is proposed, a multi-agent framework that synthesizes high-fidelity, long-horizon tool-use trajectories for training robust e-commerce shopping agents and establishes multi-agent synthetic trajectory training as an effective and scalable paradigm for enhancing LLM-based shopping assistance.

Abstract

Large Language Model (LLM)-based agents show promise for e-commerce conversational shopping, yet existing implementations lack the interaction depth and contextual breadth required for complex product research. Meanwhile, the Deep Research paradigm, despite advancing information synthesis in web search, suffers from domain gaps when transferred to e-commerce. We propose ProductResearch, a multi-agent framework that synthesizes high-fidelity, long-horizon tool-use trajectories for training robust e-commerce shopping agents. The framework employs a User Agent to infer nuanced shopping intents from behavioral histories, and a Supervisor Agent that orchestrates iterative collaboration with a Research Agent to generate synthetic trajectories culminating in comprehensive, insightful product research reports. These trajectories are rigorously filtered and distilled through a reflective internalization process that consolidates multi-agent supervisory interactions into coherent single-role training examples, enabling effective fine-tuning of LLM agents for complex shopping inquiries. Extensive experiments show that a compact MoE model fine-tuned on our synthetic data achieves substantial improvements over its base model in response comprehensiveness, research depth, and user-perceived utility, approaching the performance of frontier proprietary deep research systems and establishing multi-agent synthetic trajectory training as an effective and scalable paradigm for enhancing LLM-based shopping assistance.
Paper Structure (34 sections, 4 equations, 3 figures, 1 table)

This paper contains 34 sections, 4 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Overview of the Multi-Agent Framework
  4. Phase 1: User Profiling and Query Formulation
  5. Phase 2: State-Aware Supervised Iterative Research Execution
  6. State-Specific Verification
  7. Iterative Feedback Loop
  8. Phase 3: Trajectory Refinement and Distillation
  9. Experiments
  10. Experimental Settings
  11. Evaluation of Product Investigation Report
  12. Query-Specific Evaluation Rubrics.
  13. Reference Report Construction.
  14. Main Results
  15. Effect of Context Length
  16. ...and 19 more sections

Figures (3)

  • Figure 1: Overview of the ProductResearch framework. The User Agent infers a persona, research query, and RACE evaluation rubric from behavioral histories. The Research Agent executes Plan→Toolcall→Report reasoning under step-level supervision from the Supervisor Agent, which provides state-specific verification and corrective feedback. Approved trajectories are filtered by length and distilled via reflective internalization into single-role sequences for supervised fine-tuning.
  • Figure 2: Effect of training context length on RACE scores.
  • Figure 3: RACE scores of intermediate reports generated during the iterative synthesis process.