ZeroSearch: Incentivize the Search Capability of LLMs without Searching

Hao Sun, Z. Qiao, J. Guo, X. Fan, Y. Hou, Y. Jiang, P. Xie, F. Huang, Y. Zhang

Under Review

Homepage | Paper | Code | Models 900+ stars!

We propose a novel reinforcement learning framework that incentivizes the search capability of LLMs without interacting with real search engines.

HRM: Enhancing Agentic Retrieval Augmented Generation With Hierarchical Reward Modeling

Hao Sun, Z. Qiao, B. Wang, G. Chen, Y. Hou, Y. Jiang, P. Xie, F. Huang, Y. Zhang

Under Review

Paper

We propose a novel Agentic RAG framework that independently optimizes planning reasoning and search grounding with dual value models.

Enhancing Retrieval-Augmented Generation via Evidence Tree Search

Hao Sun, H Cai, Y. Li, X. Fan, X. Wei, S. Wang, Y. Zhang, D. Yin

The Annual Meeting of the Association for Computational Linguistics (ACL), 2025

Paper

We propose ETS, a tree-based framework that enhances LLMs’ evidence-grounding through a tree expansion process.

Unveil: Unified Visual-Textual Integration and Distillation for Multi-modal Document Retrieval

Hao Sun, Y. Hou, J. Guo, B. Wang, C. Yang, J. Ni, Y. Zhang

The Annual Meeting of the Association for Computational Linguistics (ACL), 2025

Paper

Transfer the semantic understanding capabilities from the visual-textual embedding model to a purely visual model.

History Semantic Graph Enhanced Conversational KBQA with Temporal Information Modeling

Hao Sun, Y.g Li, L. Deng, B. Li, B. Hui, B. Li, Y. Lan, Y. Zhang, Y. Li

The Annual Meeting of the Association for Computational Linguistics (ACL), 2023

Paper

Modeling history conversation information with History Semantic Graph.

Towards Verifiable Text Generation with Evolving Memory and Self-Reflection

Hao Sun, H. Cai, B. Wang, Y. Hou, X. Wei, S. Wang, Y. Zhang, D. Yin

The Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024

Paper

Improving citation generation with a two-tier verifier and active retrieval mechanism.

Retrieved In-Context Principles from Previous Mistakes

Hao Sun, Y. Jiang, B. Wang, Y. Hou, Y. Zhang, P. Xie, F. Huang

The Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024

Paper

Enabling LLMs to learn from mistakes by providing question-level and task-level principles.

AdaSwitch: Adaptive Switching between Small and Large Agents for Effective Cloud-Local Collaborative Learning

Hao Sun, J. Wu, H. Cai, X. Wei, Y. Feng, B. Wang, S. Wang, Y. Zhang, D. Yin

The Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024

Paper

Enabling adaptive switching between local agent and cloud agent through collaborative learning.

Allies: Prompting Large Language Model with Beam Search

Hao Sun, X. Liu, Y. Gong, Y. Zhang, N. Duan

The Conference on Empirical Methods in Natural Language Processing (EMNLP), 2023, Findings

Paper | Code

Improving the knowledge scope and robustness of LLMs with Beam Search.

LEAD: Liberal Feature-based Distillation for Dense Retrieval

Hao Sun, X. Liu, Y. Gong, A. Dong, J. Lu, Y. Zhang, L. Yang, R. Majumder, N. Duan

The ACM International Conference on Web Search and Data Mining (WSDM), 2024, Oral

Paper | Code

Distilling the intermediate features from teacher to student without the constraints on model architecture or tokenizers.

PeriodicMove: Shift-aware Human Mobility Recovery with Graph Neural Network

Hao Sun, C. Y.g, L. Deng, F. Zhou, F. Huang, K. Zheng

The ACM International Conference on Information and Knowledge Management (CIKM), 2021

Paper | Code

Capturing multi-level periodicity and shifting periodicity of human mobility using attention mechanism.

ConLearn: Contextual-knowledge-aware Concept Prerequisite Relation Learning with Graph Neural Network

Hao Sun, Y. Li, Y. Zhang

The SIAM International Conference on Data Mining (SDM), 2022

Paper | Code

Capturing complex transition patterns between concepts through Graph Neural Network.

Personalized Dynamic Knowledge-Aware Recommendation with Hybrid Explanations

Hao Sun, Z. Wu, Y. Cui, L. Deng, Y. Zhao, K. Zheng

The International Conference on Database Systems for Advanced Applications (DASFAA), 2021

Paper

Providing personalized and hybrid explanations for the recommendations.

Cross-model Control: Improving Multiple Large Language Models in One-time Training

J. Wu, Hao Sun, H Cai, L Su, S Wang, D Yin, X Li, M Gao

The Annual Conference on Neural Information Processing Systems (NeurIPS), 2024

Paper | Code

Improves multiple LLMs in one-time training with a portable tiny language model.

A Neural Corpus Indexer for Document Retrieval

Y. Wang, Y. Hou, H. Wang, Z. Miao, S. Wu, Hao Sun, Q. Chen, Y. Xia, C. Chi, G. Zhao, Z. Liu, X. Xie, H. Allen Sun, W. Deng, Q. Zhang, M. Yang

The Annual Conference on Neural Information Processing Systems (NeurIPS), 2024

Paper | Code 🏆 Outstanding Paper Award!

Propose an end-to-end differentiable document retrieval model that can significantly outperform both inverted index and dense retrieval solutions.

Unraveling the Mechanics of Learning-Based Demonstration Selection for In-Context Learning

H. Liu, W. Wang, Hao Sun, C. TIAN, C. Kong, X. Dong, H. Li

The Annual Meeting of the Association for Computational Linguistics (ACL), 2025

Paper

Analyze the working mechanism of learning-based demonstration selection methods.

PA-RAG: RAG Alignment via Multi-Perspective Preference Optimization

J. Wu, H. Cai, L. Yan, Hao Sun, X. Li, S. Wang, D. Yin, M. Gao

The Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL), 2025

Paper | Code

Optimizing the RAG generator to align with RAG requirements comprehensively.

Sequential-Knowledge-Aware Next POI Recommendation: A Meta-Learning Approach

Y. Cui, Hao Sun, Y. Zhao, H. Yin, K. Zheng

The ACM Transactions on Information Systems (TOIS), 2021

Paper

Utilize sequential, spatiotemporal, and social knowledge to recommend the next POI for a location-based social network user.

Robust domain misinformation detection via multi-modal feature alignment

H. Liu, W. Wang, Hao Sun, A. Rocha, H. Li

IEEE Transactions on Information Forensics and Security (TIFS), 2024

Paper

Propose a novel robust domain and cross-modal approach (RDCM) for multi-modal misinformation detection.

S2TUL: A Semi-Supervised Framework for Trajectory-User Linking

L. Deng, Hao Sun, Y. Zhao, S. Liu, K. Zheng

The ACM International Conference on Web Search and Data Mining (WSDM), 2023

Paper

Capture fine-grained intra-trajectory information by passing the trajectories into the sequential neural networks.

Efficient Trajectory Similarity Computation with Contrastive Learning

L. Deng, Y. Zhao, Z. Fu, Hao Sun, S. Liu, K. Zheng

The ACM International Conference on Information and Knowledge Management (CIKM), 2022

Paper

Employ a contrastive learning mechanism to learn the representations of trajectories, which are then used to calculate the dissimilarity between them.

Efficient and Effective Similar Subtrajectory Search: A Spatial-aware Comprehension Approach

L. Deng, Hao Sun, R. Sun, Y. Zhao, H. Su

The ACM Transactions on Intelligent Systems and Technology (TIST), 2022

Paper

Propose a Similar Subtrajectory Search with a Graph Neural Networks framework.

Fusing Local and Global Mobility Patterns for Trajectory Recovery

L. Deng, Y. Zhao, Hao Sun, C. Yang, J. Xie, K. Zheng

The ACM International Conference on Information and Knowledge Management (CIKM), 2022

Paper

Propose a neural attention model based on graph convolutional networks to enhance the accuracy of trajectory recovery.

Boosting Disfluency Detection with Large Language Model as Disfluency Generator

Z. Cheng, J. Guo, Hao Sun, Y. Zhang

The IEEE International Conference on Multimedia & Expo (ICME), 2024

Paper

Propose a framework that addresses data sparsity issues by generating disfluent data using LLM as augmentation data.

SimCNS: Simple Curriculum Negative Sampling for Multi-Source Dense Retrieval

Hao Sun, X. Liu, Y. Gong, A. Dong, G. Shi, Y. Zhang, L. Yang, N. Duan

Under Review

Paper

Improving multi-source dense retrieval with Curriculum Negative Sampling.