* denotes equal contribution

2026

Explainable Forensics of Manipulated Segments in Untrimmed Long Videos
Explainable Forensics of Manipulated Segments in Untrimmed Long Videos

Y. Feng*, J. Li*, Q. Lu*, W. Ji, J. Zhang, F. Shen, X. Li, Y. Jia, Q. Chen, L. Wang, W. Li, J. Qin

International Conference on Machine Learning (ICML) 2026 CCF-A

We propose a framework for explainable forensics of AI-manipulated segments in untrimmed long videos, providing temporal localization and natural language explanations.

[Paper]

Explainable Forensics of Manipulated Segments in Untrimmed Long Videos

Y. Feng*, J. Li*, Q. Lu*, W. Ji, J. Zhang, F. Shen, X. Li, Y. Jia, Q. Chen, L. Wang, W. Li, J. Qin

International Conference on Machine Learning (ICML) 2026 CCF-A

We propose a framework for explainable forensics of AI-manipulated segments in untrimmed long videos, providing temporal localization and natural language explanations.

[Paper]

SAM3-I: Segment Anything with Instructions
SAM3-I: Segment Anything with Instructions

J. Li*, Y. Feng*, Y. Guo*, J. Huang*, Y. Piao, Q. Bi, M. Zhang, X. Zhao, Q. Chen, S. Zou, W. Ji, H. Lu, L. Cheng

Annual Meeting of the Association for Computational Linguistics (ACL) 2026 CCF-A

We present SAM3-I, an instruction-guided segmentation model that extends SAM to handle diverse segmentation tasks through natural language instructions.

[Paper]

SAM3-I: Segment Anything with Instructions

J. Li*, Y. Feng*, Y. Guo*, J. Huang*, Y. Piao, Q. Bi, M. Zhang, X. Zhao, Q. Chen, S. Zou, W. Ji, H. Lu, L. Cheng

Annual Meeting of the Association for Computational Linguistics (ACL) 2026 CCF-A

We present SAM3-I, an instruction-guided segmentation model that extends SAM to handle diverse segmentation tasks through natural language instructions.

[Paper]

2025

MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection
MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection

S. Yang*, Y. Feng*, Y. Liu, J. Zhang, J. Qin

Annual Conference on Neural Information Processing Systems (NeurIPS) 2025 CCF-A

We propose MoniTor, which leverages LLMs with instruction tuning for online video anomaly detection in streaming scenarios.

[Paper]

MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection

S. Yang*, Y. Feng*, Y. Liu, J. Zhang, J. Qin

Annual Conference on Neural Information Processing Systems (NeurIPS) 2025 CCF-A

We propose MoniTor, which leverages LLMs with instruction tuning for online video anomaly detection in streaming scenarios.

[Paper]

MUVR: A Multi-Modal Untrimmed Video Retrieval Benchmark with Multi-Level Visual Correspondence
MUVR: A Multi-Modal Untrimmed Video Retrieval Benchmark with Multi-Level Visual Correspondence

Y. Feng, J. Hu, Q. Lu, J. Niu, L. Tan, S. Yuan, Z. Yan, Y. Jia, Q. He, S. Ge, E.Q. Chen, W. Li, L. Wang, J. Qin

Annual Conference on Neural Information Processing Systems (NeurIPS) 2025 CCF-A

We introduce MUVR, a large-scale benchmark for multi-modal untrimmed video retrieval with multi-level visual correspondence annotations.

[Paper]

MUVR: A Multi-Modal Untrimmed Video Retrieval Benchmark with Multi-Level Visual Correspondence

Y. Feng, J. Hu, Q. Lu, J. Niu, L. Tan, S. Yuan, Z. Yan, Y. Jia, Q. He, S. Ge, E.Q. Chen, W. Li, L. Wang, J. Qin

Annual Conference on Neural Information Processing Systems (NeurIPS) 2025 CCF-A

We introduce MUVR, a large-scale benchmark for multi-modal untrimmed video retrieval with multi-level visual correspondence annotations.

[Paper]

Training-free Dense Video Captioning with Large-scale Pretrained Models
Training-free Dense Video Captioning with Large-scale Pretrained Models

Y. Feng*, Z. Yan*, Y. Jia, Q. Chen, J. Qin

Machine Intelligence Research (MIR) 2025 JCR Q1 (IF 10.0)

We propose FreeDVC, a training-free framework for dense video captioning that leverages large-scale pretrained models.

[Paper]

Training-free Dense Video Captioning with Large-scale Pretrained Models

Y. Feng*, Z. Yan*, Y. Jia, Q. Chen, J. Qin

Machine Intelligence Research (MIR) 2025 JCR Q1 (IF 10.0)

We propose FreeDVC, a training-free framework for dense video captioning that leverages large-scale pretrained models.

[Paper]

2024

Doubly Contrastive Learning for Source-Free Domain Adaptive Person Search
Doubly Contrastive Learning for Source-Free Domain Adaptive Person Search

Y. Jia, R. Quan, Y. Feng, H. Chen, J. Qin

AAAI Conference on Artificial Intelligence (AAAI) 2025 CCF-A

We propose a doubly contrastive learning framework for source-free domain adaptive person search.

[Paper]

Doubly Contrastive Learning for Source-Free Domain Adaptive Person Search

Y. Jia, R. Quan, Y. Feng, H. Chen, J. Qin

AAAI Conference on Artificial Intelligence (AAAI) 2025 CCF-A

We propose a doubly contrastive learning framework for source-free domain adaptive person search.

[Paper]

2023

RefineTAD: Learning Proposal-free Refinement for Temporal Action Detection
RefineTAD: Learning Proposal-free Refinement for Temporal Action Detection

Y. Feng, Z. Zhang, R. Quan, L. Wang, J. Qin

ACM International Conference on Multimedia (ACM MM) 2023 CCF-A Honorable Mention Award (Top 1‰)

We propose RefineTAD, a proposal-free framework for temporal action detection that achieves state-of-the-art performance with a novel refinement mechanism.

[Paper]

RefineTAD: Learning Proposal-free Refinement for Temporal Action Detection

Y. Feng, Z. Zhang, R. Quan, L. Wang, J. Qin

ACM International Conference on Multimedia (ACM MM) 2023 CCF-A Honorable Mention Award (Top 1‰)

We propose RefineTAD, a proposal-free framework for temporal action detection that achieves state-of-the-art performance with a novel refinement mechanism.

[Paper]