OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System

2025-09-22

AI Summary

Existing industrial search and recommendation systems often naively adopt Transformer architectures without effectively leveraging large language models’ (LLMs) contextual engineering and multi-step reasoning capabilities. Method: This paper pioneers the deep integration of the LLM paradigm into an industrial cascaded ranking system, proposing three core innovations: (1) structured contextual engineering that fuses user preference and scenario-specific signals; (2) a block-wise implicit reasoning mechanism enabling chunked, multi-step semantic inference; and (3) progressive multi-task training grounded in user feedback chains. The approach is built entirely upon a pure Transformer architecture to achieve synergistic capability enhancement between retrieval and ranking modules. Contribution/Results: Deployed in Shopee’s primary search engine, the method yields a >2% lift in per-user GMV and a 2.90% increase in advertising revenue, with sustained, statistically significant improvements across key business metrics.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Learning Preferences or Rankings
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Despite the growing interest in replicating the scaled success of large language models (LLMs) in industrial search and recommender systems, most existing industrial efforts remain limited to transplanting Transformer architectures, which bring only incremental improvements over strong Deep Learning Recommendation Models (DLRMs). From a first principle perspective, the breakthroughs of LLMs stem not only from their architectures but also from two complementary mechanisms: context engineering, which enriches raw input queries with contextual cues to better elicit model capabilities, and multi-step reasoning, which iteratively refines model outputs through intermediate reasoning paths. However, these two mechanisms and their potential to unlock substantial improvements remain largely underexplored in industrial ranking systems. In this paper, we propose OnePiece, a unified framework that seamlessly integrates LLM-style context engineering and reasoning into both retrieval and ranking models of industrial cascaded pipelines. OnePiece is built on a pure Transformer backbone and further introduces three key innovations: (1) structured context engineering, which augments interaction history with preference and scenario signals and unifies them into a structured tokenized input sequence for both retrieval and ranking; (2) block-wise latent reasoning, which equips the model with multi-step refinement of representations and scales reasoning bandwidth via block size; (3) progressive multi-task training, which leverages user feedback chains to effectively supervise reasoning steps during training. OnePiece has been deployed in the main personalized search scenario of Shopee and achieves consistent online gains across different key business metrics, including over $+2%$ GMV/UU and a $+2.90%$ increase in advertising revenue.

Problem

• Integrating LLM-style context engineering into industrial ranking systems
• Enabling multi-step reasoning capabilities for cascade ranking pipelines
• Bridging the gap between LLM breakthroughs and industrial recommendation systems

Innovation

• Structured context engineering augments interaction history
• Block-wise latent reasoning enables multi-step refinement
• Progressive multi-task training supervises reasoning steps