Pailitao-VL: Unified Embedding and Reranker for Real-Time Multi-Modal Industrial Search

TL;DR

Pailitao-VL introduces a unified multi-modal retrieval system that enhances industrial search accuracy and efficiency by shifting to absolute ID recognition and a compare-and-calibrate reranking approach, validated through extensive testing.

Contribution

The paper presents two key innovations: a shift to absolute ID recognition for embeddings and a compare-and-calibrate listwise reranker, improving precision and speed in large-scale industrial search.

Findings

Achieves state-of-the-art retrieval performance.

Demonstrates robustness against environmental noise.

Delivers significant business impact in online tests.

Abstract

In this work, we presented Pailitao-VL, a comprehensive multi-modal retrieval system engineered for high-precision, real-time industrial search. We here address three critical challenges in the current SOTA solution: insufficient retrieval granularity, vulnerability to environmental noise, and prohibitive efficiency-performance gap. Our primary contribution lies in two fundamental paradigm shifts. First, we transitioned the embedding paradigm from traditional contrastive learning to an absolute ID-recognition task. Through anchoring instances to a globally consistent latent space defined by billions of semantic prototypes, we successfully overcome the stochasticity and granularity bottlenecks inherent in existing embedding solutions. Second, we evolved the generative reranker from isolated pointwise evaluation to the compare-and-calibrate listwise policy. By synergizing chunk-based comparative reasoning with calibrated absolute relevance scoring, the system achieves nuanced discriminative resolution while circumventing the prohibitive latency typically associated with conventional reranking methods. Extensive offline benchmarks and online A/B tests on Alibaba e-commerce platform confirm that Pailitao-VL achieves state-of-the-art performance and delivers substantial business impact. This work demonstrates a robust and scalable path for deploying advanced MLLM-based retrieval architectures in demanding, large-scale production environments.