Author name cluster

Zhaocheng Du

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

1 author row

AAAI Conference 2026 Conference Paper

Length-Adaptive Interest Network for Balancing Long and Short Sequence Modeling in CTR Prediction

Zhicheng Zhang
Zhaocheng Du
Jieming Zhu
Jiwei Tang
Fengyuan Lu
Wang Jiaheng
Song-Li Wu
Qianhui Zhu

User behavior sequences in modern recommendation systems exhibit significant length heterogeneity, ranging from sparse short-term interactions to rich long-term histories. While longer sequences provide more context, we observe that increasing the maximum input sequence length in existing CTR models paradoxically degrades performance for short-sequence users due to attention polarization and length imbalance in training data. To address this, we propose LAIN (Length-Adaptive Interest Network), a plug-and-play framework that explicitly incorporates sequence length as a conditioning signal to balance long- and short-sequence modeling. LAIN consists of three lightweight components: a Spectral Length Encoder that maps length into continuous representations, Length-Conditioned Prompting that injects global contextual cues into both long- and short-term behavior branches, and Length-Modulated Attention that adaptively adjusts attention sharpness based on sequence length. Extensive experiments on three real-world benchmarks across five strong CTR backbones show that LAIN consistently improves overall performance, achieving up to 1.15% AUC gain and 2.25% log loss reduction. Notably, our method significantly improves accuracy for short-sequence users without sacrificing long-sequence effectiveness. Our work offers a general, efficient, and deployable solution to mitigate length-induced bias in sequential recommendation.

PDF Details DOI

AAAI Conference 2026 Conference Paper

REACTION: Parameter-Efficient Learning for Recommendation

Song-Li Wu
Zhaocheng Du
Qinglin Jia
Zhenhua Dong

While deep learning (DL) has demonstrated significant success in recommender systems, it suffers from high computational complexity and poor scalability. In this work, we demonstrate, from an information-theoretic perspective, the redundancy of existing DL-based recommender models in two aspects: (1) Feature Redundancy. We show that many features are highly mutually correlated, noisy, or weakly predictive of user-item interaction labels. (2) Structural Redundancy. We further show that a large proportion of parameters in the dense layers contribute minimally to overall performance, indicating significant redundancy within the model architecture. To address these challenges, we propose REACTION (paRameter-Efficient LeArning for recommendaTION), an information-theoretic framework designed to reduce model complexity without sacrificing performance. REACTION consists of two core components: Adaptive Feature Extraction (AFE) leverages mutual information to project high-dimensional sparse features into a compact, informative subspace. This adaptively filters noisy or weak features, reduces embedding parameters, and preserves implicit feature interactions without explicit high-order computation. Dynamic Tower Fusion (DTF) bridges the representational gap between dual-tower expressiveness and single-tower efficiency. It facilitates rich cross-tower interactions during training, then merges the towers into a unified, low-latency single tower for inference. Extensive experiments on four large-scale benchmarks demonstrate that REACTION not only outperforms existing methods in accuracy but also achieves a drastic reduction in both model parameters and inference costs, thus establishing a new paradigm for efficient and scalable recommendation systems.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Personalized Visual Content Generation in Conversational Systems

Xianquan Wang
Zhaocheng Du
Huibo Xu
Shukang Yin
Yupeng Han
Jieming Zhu
Kai Zhang
Qi Liu

With the rapid progress of large language models (LLMs) and diffusion models, there has been growing interest in personalized content generation. However, current conversational systems often present the same recommended content to all users, falling into the dilemma of "one-size-fits-all. " To break this limitation and boost user engagement, in this paper, we introduce PCG ( P ersonalized Visual C ontent G eneration), a unified framework for personalizing item images within conversational systems. We tackle two key bottlenecks: the depth of personalization and the fidelity of generated images. Specifically, an LLM-powered Inclinations Analyzer is adopted to capture user likes and dislikes from context to construct personalized prompts. Moreover, we design a dual-stage LoRA mechanism—Global LoRA for understanding task-specific visual style, and Local LoRA for capturing preferred visual elements from conversation history. During training, we introduce the visual content condition method to ensure LoRA learns both historical visual context and maintains fidelity to the original item images. Extensive experiments on benchmark conversational datasets—including objective metrics and GPT-based evaluations—demonstrate that our framework outperforms strong baselines, which highlight its potential to redefine personalization in visual content generation for conversational scenarios like e-commerce and real-world recommendation.

PDF Details

NeurIPS Conference 2025 Conference Paper

Process vs. Outcome Reward: Which is Better for Agentic RAG Reinforcement Learning

Wenlin Zhang
Xiangyang Li
Kuicai Dong
Yichao Wang
Pengyue Jia
Xiaopeng Li
Yingyi Zhang
Derong Xu

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by integrating external knowledge, yet traditional RAG systems struggle with static workflows and limited adaptability for complex, multistep reasoning tasks. Agentic RAG systems, such as DeepResearch, address these issues through dynamic retrieval, iterative context refinement, and adaptive workflows. However, recent methods like Search-R1, which rely on outcome-based reinforcement learning, face challenges such as low exploration efficiency, gradient conflict, and sparse reward signals. To tackle these limitations, we introduce ReasonRAG, a novel method that leverages RAG-ProGUIDE—a high-quality dataset providing fine-grained, process-level rewards for query generation, evidence extraction, and answer generation. By employing process-supervised reinforcement learning, ReasonRAG enhances LLMs’ autonomous capabilities in search, query generation, evidence extraction, and answer synthesis. Experimental results show that ReasonRAG, utilizing RAG-ProGUIDE, outperforms existing approaches like Search-R1 and traditional RAG systems, achieving superior performance on five benchmark datasets with only 5k training instances—significantly fewer than the 90k required by Search-R1. Our code is available at https: //github. com/Applied-Machine-Learning-Lab/ReasonRAG.

PDF Details