Maosong Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Maosong Sun has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Chi Chen
Tsinghua University, Beijing, China
2 shared publications - Chunyi Peng
Northeastern University, Shenyang, China
2 shared publications - Yukun Yan
Tsinghua University, Beijing, China
2 shared publications - Zhenghao Liu
Northeastern University, Shenyang, China
2 shared publications - Zhiyuan Liu
Tsinghua University
2 shared publications - Ge Yu
Northeastern University, Shenyang, China
1 shared publication
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 9294-9303
Mitigating Visual Hallucinations via Evidence-Guided Multi-Image Reasoning in Visual Retrieval-Augmented Generation
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 9294-9303
Open AccessMitigating Visual Hallucinations via Evidence-Guided Multi-Image Reasoning in Visual Retrieval-Augmented Generation
- 4 October 2026
- Published by Springer Nature
- p. 209-227
NaVLM-PVC: Progressive Visual Compression for Efficient Native-Resolution Encoding in MLLMs
- 4 October 2026
- Published by Springer Nature
- p. 209-227
NaVLM-PVC: Progressive Visual Compression for Efficient Native-Resolution Encoding in MLLMs
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 449-460
Test-Time Deep Thinking to Explore Implicit Rules
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 449-460
Open AccessTest-Time Deep Thinking to Explore Implicit Rules
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1440-1450
Mixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation
Multimodal Retrieval-Augmented Generation (MRAG) has shown promise in mitigating hallucinations in Multimodal Large Language Models (MLLMs) by incorporating external knowledge. However, existing methods typically adhere to rigid retrieval paradigms by mimicking fixed retrieval trajectories and thus fail to fully exploit the knowledge of different retrieval experts through dynamic interaction based on the model's knowledge needs or evolving reasoning states. To overcome this limitation, we introduce Mixture-of-Retrieval Experts (MoRE), a novel framework that enables MLLMs to collaboratively interact with diverse retrieval experts for more effective knowledge exploitation. Specifically, MoRE learns to dynamically determine which expert to engage with, conditioned on the evolving reasoning state. To effectively train this capability, we propose Stepwise Group Relative Policy Optimization (Step-GRPO), which goes beyond sparse outcome-based supervision by encouraging MLLMs to interact with multiple retrieval experts and synthesize fine-grained rewards, thereby teaching the MLLM to fully coordinate all experts when answering a given query. Experimental results on diverse open-domain QA benchmarks demonstrate the effectiveness of MoRE, achieving average performance gains of over 7% compared to competitive baselines. Notably, MoRE exhibits strong adaptability by dynamically coordinating heterogeneous experts to precisely locate relevant information, validating its capability for robust, reasoning-driven expert collaboration. All codes and data are released on https://github.com/OpenBMB/MoRE.
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1440-1450
Open AccessMixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation
Multimodal Retrieval-Augmented Generation (MRAG) has shown promise in mitigating hallucinations in Multimodal Large Language Models (MLLMs) by incorporating external knowledge. However, existing methods typically adhere to rigid retrieval paradigms by mimicking fixed retrieval trajectories and thus fail to fully exploit the knowledge of different retrieval experts through dynamic interaction based on the model's knowledge needs or evolving reasoning states. To overcome this limitation, we introduce Mixture-of-Retrieval Experts (MoRE), a novel framework that enables MLLMs to collaboratively interact with diverse retrieval experts for more effective knowledge exploitation. Specifically, MoRE learns to dynamically determine which expert to engage with, conditioned on the evolving reasoning state. To effectively train this capability, we propose Stepwise Group Relative Policy Optimization (Step-GRPO), which goes beyond sparse outcome-based supervision by encouraging MLLMs to interact with multiple retrieval experts and synthesize fine-grained rewards, thereby teaching the MLLM to fully coordinate all experts when answering a given query. Experimental results on diverse open-domain QA benchmarks demonstrate the effectiveness of MoRE, achieving average performance gains of over 7% compared to competitive baselines. Notably, MoRE exhibits strong adaptability by dynamically coordinating heterogeneous experts to precisely locate relevant information, validating its capability for robust, reasoning-driven expert collaboration. All codes and data are released on https://github.com/OpenBMB/MoRE.