Author name cluster

Luyuan Chen

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.

6 papers

1 author row

AAAI Conference 2025 Conference Paper

MHBench: Demystifying Motion Hallucination in VideoLLMs

Ming Kong
Xianzhou Zeng
Luyuan Chen
Yadong Li
Bo Yan
Qiang Zhu

Similar to Language or Image LLMs, VideoLLMs are also plagued by hallucination issues. Hallucinations in videos not only manifest in the spatial dimension regarding the perception of the existence of visual objects (static) but also the temporal dimension influencing the perception of actions and events (dynamic). This paper introduces the concept of Motion Hallucination for the first time, exploring the hallucination phenomena caused by insufficient motion perception capabilities in VideoLMMs, as well as how to detect, evaluate, and mitigate the hallucination. To this end, we propose the first benchmark for assessing motion hallucination MHBench, which consists of 1,200 videos of 20 different action categories. By constructing a collection of adversarial triplet types of videos (original/antonym/incomplete), we achieve a comprehensive evaluation of motion hallucination. Furthermore, we present a Motion Contrastive Decoding (MotionCD) method, which employs bidirectional motion elimination between the original video and its reverse playback to construct an amateur model that removes the influence of motion while preserving visual information, thereby effectively suppressing motion hallucination. Extensive experiments on MHBench reveal that current state-of-the-art VideoLLMs significantly suffer from motion hallucination, while the introduction of MotionCD can effectively mitigate this issue, achieving up to a 15.1% performance improvement. We hope this work will guide future efforts in avoiding and mitigating hallucinations in VideoLLMs.

PDF Details DOI

AAAI Conference 2025 Conference Paper

MoLE:Decoding by Mixture of Layer Experts Alleviates Hallucination in Large Vision-Language Models

Tian Liang
Yuetian Du
Jing Huang
Ming Kong
Luyuan Chen
Yadong Li
Siye Chen
Qiang Zhu

Recent advancements in Large Vision-Language Models (LVLMs) highlight their ability to integrate and process multi-modal information. However, hallucinations—where generated content is inconsistent with input vision and instructions—remain a challenge. In this paper, we analyze LVLMs' layer-wise decoding and identify that hallucinations can arise during the reasoning and factual information injection process. Additionally, as the number of generated tokens increases, the forgetting of the original prompt may also lead to hallucinations.To address this, we propose a training-free decoding method called Mixture of Layer Experts (MoLE). MoLE leverages a heuristic gating mechanism to dynamically select multiple layers of LVLMs as expert layers: the Final Expert, the Second Opinion expert, and the Prompt Retention Expert. By the cooperation of each expert, MoLE enhances the robustness and faithfulness of the generation process. Our extensive experiments demonstrate that MoLE significantly reduces hallucinations, outperforming the current state-of-the-art decoding techniques across three mainstream LVLMs and two established hallucination benchmarks. Moreover, our method reveals the potential of LVLMs to independently produce more reliable and accurate outputs.

PDF Details DOI

EAAI Journal 2023 Journal Article

Permutation Jensen–Shannon divergence for Random Permutation Set

Luyuan Chen
Yong Deng
Kang Hao Cheong

Random Permutation Set (RPS) considers the permutation of elements for a certain set, which is an efficient tool for dealing with uncertainty with ordered information. An important feature of RPS theory is that in the fusion rule of RPS sources, the fusion order has a great impact on fusion results. However, how to determine the fusion order has not yet been discussed. To address this problem, this paper first proposes Permutation Jensen–Shannon (PJS) divergence for measuring the distance between two RPSs. Based on PJS divergence, a new Reliability Assessment algorithm, named RAPJS, is then presented for determining the fusion order of RPSs. The proposed PJS divergence satisfies the properties of non–degeneracy, boundary, and symmetry, and has desirable compatibility with Belief Jensen–Shannon divergence and Jensen–Shannon divergence under certain conditions. The presented RAPJS makes use of the divergence information to calculate the reliability degree of RPS sources, the RPS with a higher reliability is fused first. Experiment results in threat assessment reveal that the presented RAPJS algorithm can determine the fusion order reasonably and effectively. The assessment results using the proposed RAPJS algorithm has the highest target recognition rate compared to other results under different fusion orders.

Details DOI

EAAI Journal 2021 Journal Article

Probability transformation of mass function: A weighted network method based on the ordered visibility graph

Luyuan Chen
Yong Deng
Kang Hao Cheong

Transform of basic probability assignment to probability distribution is an important aspect of decision making process. To address this issue, a weighted network method based on the ordered visibility graph is proposed in this paper, named OVGWP. In this proposed method, the information volume of focal elements is calculated by belief entropy. The entropy value is used to determine the rank of each proposition. After generating the rank, a weighted network corresponding to the given basic probability assignment can be constructed. The global ratio for proportional belief transformation is determined by the degree of nodes and its weighted edges in the network. Compared with existing ordered visibility graph probability, we have considered not only the belief value itself, but also the cardinality of basic probability assignment. Hence the proposed OVGWP considers a much more comprehensive information for transformation. Experimental results reveal that OVGWP produces an effective and reasonable transformation performance compared with existing methods. If the basic probability assignment is given as m ( Θ ) = 1, the proposed OVGWP has the same result with pignistic probability transformation. The proposed OVGWP satisfies the consistency of the upper and lower boundaries.

Details DOI

EAAI Journal 2019 Journal Article

A novel evidential FMEA method by integrating fuzzy belief structure and grey relational projection method

Zhen Li
Luyuan Chen

Failure mode and effects analysis (FMEA) has been applied extensively in reliability engineering domain. Risk priority number (RPN), which is the product of occurrence (O), severity (S) and detection ( D ) of a failure is the most important measure used in FMEA for prioritizing risk. In this paper, a novel evidential FMEA integrating fuzzy belief structure and grey relational projection method(GRPM) is presented to avoid the use of traditional RPN, for it has been criticized with some weaknesses. First, a new distribution form of assessments applying fuzzy belief structure to represent the experts’ opinions in a more flexible and reasonable manner is constructed. Second, a new method to transform the experts’ fuzzy opinions into basic probability assignments (BPAs) is proposed. Third, to address the problems existing in conflicting evidence combination, the differences of experts’ authorities are taken into consideration, which is represented by the discounting coefficient. Besides, GPRM is adopted to address a more comprehensive assessments by fusing the 3 diverse criterion with weights respectively, which can overcome the limitations in traditional FMEA. An example is illustrated to show the practical application of the proposed FMEA methodology in engineering area.

Details DOI

EAAI Journal 2018 Journal Article

A new failure mode and effects analysis model using Dempster–Shafer evidence theory and grey relational projection method

Luyuan Chen
Yong Deng

Failure mode and effects analysis (FMEA) is an important analytical tool in reliability engineering to identify the critical potential failure modes. In this paper, a new FMEA model using Dempster–Shafer evidence theory (DSET) and grey relational projection method (GRPM) is proposed, which mainly manages two critical issues of FMEA: the presentation and handling of various types of uncertainty and the ranking of risk priorities of failure modes. DSET has a good advantage to express and model the assessment results of risk factors. GRPM is used to determine the risk priority order of the identified failure modes, where the double reference points (the positive/negative ideal alternative) are applied. Two illustrative cases are provided to demonstrate the effectiveness and practicality of the proposed method.

Details DOI