Feng Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Feng Sun has a total of 83 co-authors. Here are the top co-authors by number of shared publications:
- Dandan Su
Department of Mechanical Design and Manufacture and Automation, Hebei University, Baoding 071000, PR China
4 shared publications - Qi Zhang
Microsoft Corporation
3 shared publications - Weiwei Deng
Department of Mechanics and Aerospace Engineering, Southern University of Science and Technology (SUSTech) 2 , Shenzhen 518055
3 shared publications Yujie Yin
3 shared publications- Haizhen Huang
Microsoft Corporation
2 shared publications - Minghui Song
Harbin Institute of Technology
2 shared publications
Latest Publications
- 20 October 2026
- Published by Royal Society of Chemistry (RSC) in Energy & Environmental Science
- Vol. 19 (20) , 7183-7195
Bio-inspired self-assembly of ordered interface topology for robust composite solid electrolytes
A bio-inspired self-assembly strategy enables composite solid-state electrolytes with fast ion transport, stable interfaces, strong dendrite suppression, high conductivity, mechanical robustness, and excellent cycling stability.
- 20 October 2026
- Published by Royal Society of Chemistry (RSC) in Energy & Environmental Science
- Vol. 19 (20) , 7183-7195
Bio-inspired self-assembly of ordered interface topology for robust composite solid electrolytes
A bio-inspired self-assembly strategy enables composite solid-state electrolytes with fast ion transport, stable interfaces, strong dendrite suppression, high conductivity, mechanical robustness, and excellent cycling stability.
- Open Access
AdNanny: One Reasoning LLM for All Offline Ads Recommendation Tasks
Large Language Models (LLMs) have shown strong capabilities in Natural Language Understanding and Generation, but deploying them directly in online advertising systems is often impractical due to strict millisecond-level latency constraints. This has motivated the use of LLMs offline to improve retrieval, ranking, and recommendation models. Existing solutions typically fine-tune separate LLMs for individual tasks such as query-ad relevance labeling, keyword-based query generation, and user profiling. This results in redundant models, high maintenance cost, and limited performance gains despite substantial overlap in domain knowledge and reasoning patterns. We introduce AdNanny, a unified reasoning-centric LLM that serves as a shared backbone for offline advertising tasks. AdNanny is obtained by fine-tuning a public 671B-parameter DeepSeek-R1 checkpoint using a scalable training system that supports hybrid dense-MoE parallelism. We construct reasoning-augmented corpora that pair structured supervision with step-by-step natural language explanations. A multi-task supervised fine-tuning stage with adaptive reweighting enables AdNanny to handle diverse labeling and generation tasks in a consistent reasoning format. This is followed by reinforcement learning using downstream advertising metrics to align model behavior with online retrieval and ranking objectives. AdNanny is deployed in production within Bing Ads, where it significantly reduces manual labeling effort and improves accuracy across multiple offline tasks. By consolidating many task-specific models into a single reasoning-centric foundation model, AdNanny provides a scalable and cost-effective solution for large-scale advertising systems.
Open AccessAdNanny: One Reasoning LLM for All Offline Ads Recommendation Tasks
Large Language Models (LLMs) have shown strong capabilities in Natural Language Understanding and Generation, but deploying them directly in online advertising systems is often impractical due to strict millisecond-level latency constraints. This has motivated the use of LLMs offline to improve retrieval, ranking, and recommendation models. Existing solutions typically fine-tune separate LLMs for individual tasks such as query-ad relevance labeling, keyword-based query generation, and user profiling. This results in redundant models, high maintenance cost, and limited performance gains despite substantial overlap in domain knowledge and reasoning patterns. We introduce AdNanny, a unified reasoning-centric LLM that serves as a shared backbone for offline advertising tasks. AdNanny is obtained by fine-tuning a public 671B-parameter DeepSeek-R1 checkpoint using a scalable training system that supports hybrid dense-MoE parallelism. We construct reasoning-augmented corpora that pair structured supervision with step-by-step natural language explanations. A multi-task supervised fine-tuning stage with adaptive reweighting enables AdNanny to handle diverse labeling and generation tasks in a consistent reasoning format. This is followed by reinforcement learning using downstream advertising metrics to align model behavior with online retrieval and ranking objectives. AdNanny is deployed in production within Bing Ads, where it significantly reduces manual labeling effort and improves accuracy across multiple offline tasks. By consolidating many task-specific models into a single reasoning-centric foundation model, AdNanny provides a scalable and cost-effective solution for large-scale advertising systems.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 12768-12780
MAIN: Mutual Alignment Is Necessary for instruction tuning
Fanyi Yang, Jianfeng Liu, Xin Zhang, Haoyu Liu, Xixin Cao, Yuefeng Zhan, Hao Sun, Weiwei Deng, Feng Sun, Qi Zhang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 12768-12780
Open AccessMAIN: Mutual Alignment Is Necessary for instruction tuning
Fanyi Yang, Jianfeng Liu, Xin Zhang, Haoyu Liu, Xixin Cao, Yuefeng Zhan, Hao Sun, Weiwei Deng, Feng Sun, Qi Zhang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- Open Access
- 29 November 2024
- Published by Qeios Ltd in Qeios Preprints
StreamAdapter: Efficient Test Time Adaptation from Contextual Streams
In-context learning (ICL) allows large language models (LLMs) to adapt to new tasks directly from the given demonstrations without requiring gradient updates. While recent advances have expanded context windows to accommodate more demonstrations, this approach increases inference costs without necessarily improving performance. To mitigate these issues, We propose StreamAdapter, a novel approach that directly updates model parameters from context at test time, eliminating the need for explicit in-context demonstrations. StreamAdapter employs context mapping and weight absorption mechanisms to dynamically transform ICL demonstrations into parameter updates with minimal additional parameters. By reducing reliance on numerous in-context examples, StreamAdapter significantly reduce inference costs and allows for efficient inference with constant time complexity, regardless of demonstration count. Extensive experiments across diverse tasks and model architectures demonstrate that StreamAdapter achieves comparable or superior adaptation capability to ICL while requiring significantly fewer demonstrations. The superior task adaptation and context encoding capabilities of StreamAdapter on both language understanding and generation tasks provides a new perspective for adapting LLMs at test time using context, allowing for more efficient adaptation across scenarios and more cost-effective inference.
- 29 November 2024
- Published by Qeios Ltd in Qeios Preprints
Open AccessStreamAdapter: Efficient Test Time Adaptation from Contextual Streams
In-context learning (ICL) allows large language models (LLMs) to adapt to new tasks directly from the given demonstrations without requiring gradient updates. While recent advances have expanded context windows to accommodate more demonstrations, this approach increases inference costs without necessarily improving performance. To mitigate these issues, We propose StreamAdapter, a novel approach that directly updates model parameters from context at test time, eliminating the need for explicit in-context demonstrations. StreamAdapter employs context mapping and weight absorption mechanisms to dynamically transform ICL demonstrations into parameter updates with minimal additional parameters. By reducing reliance on numerous in-context examples, StreamAdapter significantly reduce inference costs and allows for efficient inference with constant time complexity, regardless of demonstration count. Extensive experiments across diverse tasks and model architectures demonstrate that StreamAdapter achieves comparable or superior adaptation capability to ICL while requiring significantly fewer demonstrations. The superior task adaptation and context encoding capabilities of StreamAdapter on both language understanding and generation tasks provides a new perspective for adapting LLMs at test time using context, allowing for more efficient adaptation across scenarios and more cost-effective inference.
Performance Degradation Mechanism of the Si@N, S‐Doped Carbon Anode in Sulfide‐Based All‐Solid‐State Batteries
Silicon (Si) anode is a promising anode material for all‐solid‐state lithium batteries with ultra‐high theoretical specific capacity and low lithium dendrite risk. However, the inevitable vast volume expansion of Si anode during charge/discharge is recognized as a major limitation preventing its commercial application. Herein, an N, S self‐doped amorphous carbon layer coated on porous micron‐sized Si (p‐mSi@C) is designed to construct an electron/ion conducting network while ensuring structural and interfacial stability. Uneven distribution of von mises stresses during p‐mSi lithiation leads to irregular volume expansion and even fragmentation. Meanwhile, the growth of by‐products at the interface between p‐mSi and electrolyte contact leads to a rapid capacity decay. Compared to p‐mSi anode, p‐mSi@C reduces the risk of fragmentation thanks to the stress‐absorbing effect of amorphous carbon, delivering excellent electrochemical performance (2679.65 mAh g −1 at 0.2 mA cm −2 with initial coulombic efficiency of 84%). More importantly, the chemical failure mechanisms of p‐mSi and p‐mSi@C composite anodes are revealed through structural characterization, chemical analysis, and simulation, which provides the necessary theoretical guidance for practicalization.
Performance Degradation Mechanism of the Si@N, S‐Doped Carbon Anode in Sulfide‐Based All‐Solid‐State Batteries
Silicon (Si) anode is a promising anode material for all‐solid‐state lithium batteries with ultra‐high theoretical specific capacity and low lithium dendrite risk. However, the inevitable vast volume expansion of Si anode during charge/discharge is recognized as a major limitation preventing its commercial application. Herein, an N, S self‐doped amorphous carbon layer coated on porous micron‐sized Si (p‐mSi@C) is designed to construct an electron/ion conducting network while ensuring structural and interfacial stability. Uneven distribution of von mises stresses during p‐mSi lithiation leads to irregular volume expansion and even fragmentation. Meanwhile, the growth of by‐products at the interface between p‐mSi and electrolyte contact leads to a rapid capacity decay. Compared to p‐mSi anode, p‐mSi@C reduces the risk of fragmentation thanks to the stress‐absorbing effect of amorphous carbon, delivering excellent electrochemical performance (2679.65 mAh g −1 at 0.2 mA cm −2 with initial coulombic efficiency of 84%). More importantly, the chemical failure mechanisms of p‐mSi and p‐mSi@C composite anodes are revealed through structural characterization, chemical analysis, and simulation, which provides the necessary theoretical guidance for practicalization.
- 9 September 2024
- Published by Royal Society of Chemistry (RSC) in Journal of Materials Chemistry A
- Vol. 12 (39) , 27011-27021
High performance P-based argyrodite sulfide electrolytes enabled by Sb-based argyrodite doping for all-solid-state lithium metal batteries
An ingenious yet effective strategy was applied to prepare P-based Li-argyrodites with Sb-based Li-argyrodite doping. The incorporation of Sn, Sb and I endows the novel ternary-doped Li 6 PS 5 Cl electrolyte with much enhanced integrated performance.
- 9 September 2024
- Published by Royal Society of Chemistry (RSC) in Journal of Materials Chemistry A
- Vol. 12 (39) , 27011-27021
High performance P-based argyrodite sulfide electrolytes enabled by Sb-based argyrodite doping for all-solid-state lithium metal batteries
An ingenious yet effective strategy was applied to prepare P-based Li-argyrodites with Sb-based Li-argyrodite doping. The incorporation of Sn, Sb and I endows the novel ternary-doped Li 6 PS 5 Cl electrolyte with much enhanced integrated performance.
- Open Access
Auto Search Indexer for End-to-End Document Retrieval
Generative retrieval, which is a new advanced paradigm for document retrieval, has recently attracted research interests, since it encodes all documents into the model and directly generates the retrieved documents. However, its power is still underutilized since it heavily relies on the "preprocessed" document identifiers (docids), thus limiting its retrieval performance and ability to retrieve new documents. In this paper, we propose a novel fully end-to-end retrieval paradigm. It can not only end-to-end learn the best docids for existing and new documents automatically via a semantic indexing module, but also perform end-to-end document retrieval via an encoder-decoder-based generative model, namely Auto Search Indexer (ASI). Besides, we design a reparameterization mechanism to combine the above two modules into a joint optimization framework. Extensive experimental results demonstrate the superiority of our model over advanced baselines on both public and industrial datasets and also verify the ability to deal with new documents.
Open AccessAuto Search Indexer for End-to-End Document Retrieval
Generative retrieval, which is a new advanced paradigm for document retrieval, has recently attracted research interests, since it encodes all documents into the model and directly generates the retrieved documents. However, its power is still underutilized since it heavily relies on the "preprocessed" document identifiers (docids), thus limiting its retrieval performance and ability to retrieve new documents. In this paper, we propose a novel fully end-to-end retrieval paradigm. It can not only end-to-end learn the best docids for existing and new documents automatically via a semantic indexing module, but also perform end-to-end document retrieval via an encoder-decoder-based generative model, namely Auto Search Indexer (ASI). Besides, we design a reparameterization mechanism to combine the above two modules into a joint optimization framework. Extensive experimental results demonstrate the superiority of our model over advanced baselines on both public and industrial datasets and also verify the ability to deal with new documents.
- 10 September 2023
- Published by Elsevier in Journal of Power Sources
- Vol. 584
A novel 3D fine-mesh flow field design and performance analysis for proton exchange membrane fuel cells
- 10 September 2023
- Published by Elsevier in Journal of Power Sources
- Vol. 584
A novel 3D fine-mesh flow field design and performance analysis for proton exchange membrane fuel cells
- 14 August 2023
- Published by Taylor & Francis in International Journal of Green Energy
- Vol. 21 (6) , 1253-1266
Numerical simulation of holomorphic microscopic metal foam as cathode flow field in proton exchange membrane fuel cells
- 14 August 2023
- Published by Taylor & Francis in International Journal of Green Energy
- Vol. 21 (6) , 1253-1266
Numerical simulation of holomorphic microscopic metal foam as cathode flow field in proton exchange membrane fuel cells