OneSug: A Unified End-to-End Generative Framework for E-commerce Query Suggestion (AAAI 2026)

Forum topic · 小凯 · 2026-07-05

Summary

OneSug is a research paper accepted to AAAI 2026 that proposes a unified, end-to-end generative framework for e-commerce query suggestion. Published in the ACM Digital Library (doi:10.1145/3773966.3777963), the work addresses the limitations of traditional pipeline-based search assistance systems, which handle candidate retrieval, ranking, and generation as separate stages. By framing query suggestion as a single generative task, OneSug aims to improve user intent understanding, multi-turn interaction quality, and system efficiency in large-scale e-commerce search and recommendation settings. This forum post indexes the paper alongside related work on query auto-complete, clarification questions in open-domain QA, and enterprise conversational systems, situating it within the broader evolution from cascaded retrieval-ranking pipelines toward generative and agentic search architectures. The post also discusses practical engineering considerations for deploying such systems, including latency budgets, data quality, safety, and evaluation methodology. Readers should consult the original PDF for exact experimental results and quantitative benchmarks, as this entry summarizes publicly available metadata.

OneSug: The Unified End-to-End Generative Framework for E-commerce Query Suggestion (AAAI 2026)

This entry indexes an AAAI 2026 paper on OneSug, a unified end-to-end generative framework for e-commerce query suggestion, filed under Search Assistance and Recommender Engines.

Paper Information

| Field | Content | |-------|---------| | Title | OneSug: The Unified End-to-End Generative Framework for E-commerce Query Suggestion | | Venue | AAAI 2026 | | Source | https://dl.acm.org/doi/abs/10.1145/3773966.3777963 | | Type | Academic paper |

Context and Motivation

Large-scale search and recommendation systems have traditionally relied on pipeline-based designs that separate retrieval, ranking, and generation. In the LLM era, such fragmentation makes it harder to support natural-language interaction, multi-hop reasoning, and real-time intent understanding. OneSug addresses query suggestion — the task of predicting and generating follow-up queries as users type or interact — as a single generative problem rather than a cascade of independent modules.

Positioning Within the Field

The paper sits at the intersection of information retrieval and large-scale e-commerce search, contributing to the shift from classic cascaded stacks toward generative and agentic search architectures:

Neural IR evolution: from BM25 to dense retrieval, cross-encoder reranking, late interaction, and now generative retrieval with LLMs.
Generative recommendation (Gen-Rec): LLMs bring semantic priors and cold-start capability, but online inference cost and hallucination risk require careful system design.
Agentic search: retrieval becomes an iterative, plannable process rather than a one-shot step, shifting evaluation toward task success and process-level metrics.

Engineering Considerations for Deployment

| Check | Question | Suggestion | |-------|----------|------------| | Latency | What is the p99 budget? How many retrieval steps? | Cascading + early stopping, caching popular queries, async reranking | | Quality | Do offline gains translate to online CTR/satisfaction? | Interleaving experiments, human audit, citation verification | | Safety | Does open retrieval introduce poisoning or bias? | Source whitelisting, adversarial detection, output filtering | | Cost | Per-query token and GPU footprint? | Small-model routing, distillation, hybrid sparse+dense retrieval |

Related Entries

Asking Clarification Questions to Handle Ambiguity in Open-Domain QA
Asking Clarifying Questions in Open-Domain Information-Seeking Conversations
DiAL: Diversity-Aware Listwise Ranking for Query Auto-Complete (EMNLP)
Evaluating Auto-Complete Ranking for Diversity and Relevance (ECIR)
Enhancing Discoverability in Enterprise Conversational Systems
Evaluation and Continual Improvement for an Enterprise AI Assistant

Notes

This entry is based on the paper's public metadata and abstract. Exact experimental results, datasets, and baseline comparisons should be verified against the original PDF via the ACM DL link.