Qian Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Qian Li has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Jie Jiang
Tencent Inc., Shenzhen, China
3 shared publications - Changping Wang
Tencent Inc., Beijing, China
2 shared publications - Huan Yu
Tencent Inc., Beijing, China
2 shared publications - Jun Zhang
Tencent Inc., Beijing, China
2 shared publications - Xun Liu
Tencent, Shenzhen, China
2 shared publications - Anan Liu
Tencent, Shenzhen, China
1 shared publication
Latest Publications
- Open Access
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 1641-1653
UniTraj: Cross-Domain Long-Sequence Modeling for Commercial Recommendation
Long-sequence modeling is increasingly important in recommender systems for capturing users’ evolving and long-term interests. In advertising, however, user interaction histories are often highly sparse due to limited exposure opportunities, making ad-only behavior sequences insufficient for effective long-sequence recommendation. To address this limitation, we propose UniTraj, a practical framework that extends sequence construction beyond the advertising domain by incorporating behaviors from content-consumption scenarios, forming unified commercial trajectories across domains and scenarios. Such unified trajectories provide richer behavioral context, but also introduce substantial heterogeneity in feature taxonomy, behavioral semantics, and optimization targets. In particular, they raise three key challenges: interference among fields from different domains and scenarios, target-specific conflicts in temporal and semantic patterns, and complex high-order dependencies across heterogeneous behavioral signals. To tackle these challenges, UniTraj adopts a two-stage design. In the first stage, it combines hierarchical hard search with a decoupled embedding-based soft search module to retrieve relevant behaviors under complex feature hierarchies while reducing conflicts between retrieval and representation learning. In the second stage, it introduces several decoupled sequence modeling components, including Decoupled Side Information Temporal Interest Networks for mitigating cross-field interference, target-decoupled positional encoding and target-decoupled SASRec for capturing target-aware temporal dynamics, and Deep TIN for modeling high-order behavioral correlations. We deploy UniTraj in a large-scale online advertising system and observe consistent improvements in online business metrics across multiple commercial scenarios. The results demonstrate the effectiveness of unified cross-domain behavior modeling for long-sequence recommendation in sparse advertising environments.
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 1641-1653
Open AccessUniTraj: Cross-Domain Long-Sequence Modeling for Commercial Recommendation
Long-sequence modeling is increasingly important in recommender systems for capturing users’ evolving and long-term interests. In advertising, however, user interaction histories are often highly sparse due to limited exposure opportunities, making ad-only behavior sequences insufficient for effective long-sequence recommendation. To address this limitation, we propose UniTraj, a practical framework that extends sequence construction beyond the advertising domain by incorporating behaviors from content-consumption scenarios, forming unified commercial trajectories across domains and scenarios. Such unified trajectories provide richer behavioral context, but also introduce substantial heterogeneity in feature taxonomy, behavioral semantics, and optimization targets. In particular, they raise three key challenges: interference among fields from different domains and scenarios, target-specific conflicts in temporal and semantic patterns, and complex high-order dependencies across heterogeneous behavioral signals. To tackle these challenges, UniTraj adopts a two-stage design. In the first stage, it combines hierarchical hard search with a decoupled embedding-based soft search module to retrieve relevant behaviors under complex feature hierarchies while reducing conflicts between retrieval and representation learning. In the second stage, it introduces several decoupled sequence modeling components, including Decoupled Side Information Temporal Interest Networks for mitigating cross-field interference, target-decoupled positional encoding and target-decoupled SASRec for capturing target-aware temporal dynamics, and Deep TIN for modeling high-order behavioral correlations. We deploy UniTraj in a large-scale online advertising system and observe consistent improvements in online business metrics across multiple commercial scenarios. The results demonstrate the effectiveness of unified cross-domain behavior modeling for long-sequence recommendation in sparse advertising environments.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7465-7476
S-GRec: Personalized Semantic-Aware Generative Recommendation with Asymmetric Advantage
Generative recommendation models sequence generation to produce items end-to-end, but training from behavioral logs often provides weak supervision on underlying user intent. Although Large Language Models (LLMs) offer rich semantic priors that could supply such supervision, direct adoption in industrial recommendation is hindered by two obstacles: semantic signals can conflict with platform business objectives, and LLM inference is prohibitively expensive at scale. This paper presents S-GRec, a semantic-aware framework that decouples an online lightweight generator from an offline LLM-based semantic judge for train-time supervision. S-GRec introduces a two-stage Personalized Semantic Judge (PSJ) that produces interpretable aspect evidence and learns user-conditional aggregation from pairwise feedback, yielding stable semantic rewards. To prevent semantic supervision from deviating from business goals, Asymmetric Advantage Policy Optimization (A2PO) anchors optimization on business rewards (e.g., eCPM) and injects semantic advantages only when they are consistent. Extensive experiments on public benchmarks and a large-scale production system validate both effectiveness and scalability, including statistically significant gains in CTR and a 1.19% lift in GMV in online A/B tests, without requiring real-time LLM inference.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7465-7476
Open AccessS-GRec: Personalized Semantic-Aware Generative Recommendation with Asymmetric Advantage
Generative recommendation models sequence generation to produce items end-to-end, but training from behavioral logs often provides weak supervision on underlying user intent. Although Large Language Models (LLMs) offer rich semantic priors that could supply such supervision, direct adoption in industrial recommendation is hindered by two obstacles: semantic signals can conflict with platform business objectives, and LLM inference is prohibitively expensive at scale. This paper presents S-GRec, a semantic-aware framework that decouples an online lightweight generator from an offline LLM-based semantic judge for train-time supervision. S-GRec introduces a two-stage Personalized Semantic Judge (PSJ) that produces interpretable aspect evidence and learns user-conditional aggregation from pairwise feedback, yielding stable semantic rewards. To prevent semantic supervision from deviating from business goals, Asymmetric Advantage Policy Optimization (A2PO) anchors optimization on business rewards (e.g., eCPM) and injects semantic advantages only when they are consistent. Extensive experiments on public benchmarks and a large-scale production system validate both effectiveness and scalability, including statistically significant gains in CTR and a 1.19% lift in GMV in online A/B tests, without requiring real-time LLM inference.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7477-7488
GPR: Towards a Generative Pre-trained One-Model Paradigm for Large-Scale Advertising Recommendation
Online advertising recommendation is the economic engine of the digital ecosystem, yet traditional multi-stage cascading systems suffer from objective misalignment and error propagation. Conversely, emerging generative models often struggle with the strict latency and revenue alignment requirements of industrial applications. To address these issues, we propose GPR (Generative Pre-trained Recommender). Instead of the complex retrieval and pre-ranking cascade, GPR employs a unified generative framework to directly generate candidate sets, which are then refined by a lightweight fine-ranking stage for precise business alignment. Unlike prior works, GPR introduces three key technical innovations to ensure industrial feasibility: (1) Unified Representation: We utilize RQ-KMeans+ to map heterogeneous user behaviors and ad content into a shared semantic ID space; (2) Decoupled Architecture: We design a Heterogeneous Hierarchical Decoder (HHD) to decouple user intent modeling from ad generation, enabling efficient long-horizon inference; (3) Value-Aligned Generation: We propose a Value-Guided Trie-Constrained Beam Search and a hierarchical training strategy (MTP + HEPO) to strictly enforce business constraints and maximize commercial value. GPR has been fully deployed in Tencent Weixin Channels, serving over 500 million daily active users. Extensive online A/B testing demonstrates that GPR delivers substantial gains in GMV and CTCVR over highly optimized cascading baselines, validating the generative paradigm as a superior solution for large-scale advertising systems.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7477-7488
Open AccessGPR: Towards a Generative Pre-trained One-Model Paradigm for Large-Scale Advertising Recommendation
Online advertising recommendation is the economic engine of the digital ecosystem, yet traditional multi-stage cascading systems suffer from objective misalignment and error propagation. Conversely, emerging generative models often struggle with the strict latency and revenue alignment requirements of industrial applications. To address these issues, we propose GPR (Generative Pre-trained Recommender). Instead of the complex retrieval and pre-ranking cascade, GPR employs a unified generative framework to directly generate candidate sets, which are then refined by a lightweight fine-ranking stage for precise business alignment. Unlike prior works, GPR introduces three key technical innovations to ensure industrial feasibility: (1) Unified Representation: We utilize RQ-KMeans+ to map heterogeneous user behaviors and ad content into a shared semantic ID space; (2) Decoupled Architecture: We design a Heterogeneous Hierarchical Decoder (HHD) to decouple user intent modeling from ad generation, enabling efficient long-horizon inference; (3) Value-Aligned Generation: We propose a Value-Guided Trie-Constrained Beam Search and a hierarchical training strategy (MTP + HEPO) to strictly enforce business constraints and maximize commercial value. GPR has been fully deployed in Tencent Weixin Channels, serving over 500 million daily active users. Extensive online A/B testing demonstrates that GPR delivers substantial gains in GMV and CTCVR over highly optimized cascading baselines, validating the generative paradigm as a superior solution for large-scale advertising systems.