Author name cluster

Yuan Meng

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.

9 papers

2 author rows

EAAI Journal 2026 Journal Article

An improved DeepLab version three plus algorithm for segmenting heartwood and sapwood on wooden flooring surface

Yuan Meng
Rongrong Li

Wood is physiologically divided into two distinct regions, sapwood (lighter outer wood) and heartwood (darker inner core). Accurate sapwood identification is crucial for grading wooden flooring. Current sapwood selection relies on manual inspection, which is inefficient and error-prone, driving demand for automated, real-time machine vision solutions on production lines. However, most existing segmentation methods struggled to achieve high segmentation accuracy, particularly under limited computational resources. To address this, a lightweight DeepLab version three plus (DeepLabV3+) segmentation model is proposed for precise sapwood and heartwood detection under limited computational resources. A dedicated wood flooring dataset was established, and an Open Computer Vision (OpenCV)-based preprocessing method was applied to accelerate annotation and enhance label consistency. The Mobile Network version two (MobileNetV2) backbone replaces Xception to reduce model complexity, while an Efficient Multi-Scale Attention (EMA) module is integrated into the Atrous Spatial Pyramid Pooling (ASPP) block, and a Strip Pooling (SP) module is incorporated into the decoder to strengthen contextual and boundary representation. A hybrid Focal Dice loss further improves optimization stability and segmentation precision. Experimental evaluations show that the proposed model achieves 93. 45 % mean intersection-over-union (mIoU), 95. 90 % mean pixel accuracy (mPA), and 97. 15 % mean precision (mPrecision), outperforming several state-of-the-art segmentation algorithms. With only 6. 09 million parameters and an inference speed of 82. 65 frames per second, the model demonstrates strong potential for intelligent wood grading and real-time manufacturing automation.

Details DOI

IROS Conference 2025 Conference Paper

CIT: Context-Based Biased Batch-Sampling for Almost-Surely Asymptotically Optimal Motion Planning

Liding Zhang
Yankun Wei
Kuanqi Cai
Zhenshan Bing
Yuan Meng
Fan Wu 0015
Sami Haddadin
Alois C. Knoll

This paper introduces Context Informed Trees (CIT*), a sampling-based motion planning algorithm that enhances exploration efficiency by biasing sampling based on uncertainty estimation from local samples and connectivity information obtained during the search process. CIT* is based on Flexible Informed Trees (FIT*) and incorporates three key components: region-based sampling, uncertainty-driven weighting, and connection-greedy prioritization (CGP). It generates regions from sampled states based on local obstacle proximity, assigning weights to these regions using probability uncertainty estimation via kernel density estimation (KDE) classification. To further refine the sampling focus, CGP prioritizes regions that exhibit strong connectivity in previous searches, ensuring that exploration is directed toward unknown and critical areas that have a higher likelihood of contributing to feasible and efficient paths. The sampling process is then guided by a mixture of Gaussian distributions centered on weighted regions, where the weighting biases sampling toward more critical regions, thereby improving search efficiency and accelerating convergence. Benchmark evaluations demonstrate that CIT* improves efficiency by reducing reliance on random sampling, which often leads to slower solution discovery and higher path costs. With biased sampling, CIT* maintains strong performance in solving complex motion planning problems in ${\mathbb{R}^4}$ to ${\mathbb{R}^{16}}$ and has been demonstrated on a real-world manipulation task. A video showcasing our method and experimental results is available at: https://youtu.be/SG2cy9WmjD0.

Details

NeurIPS Conference 2025 Conference Paper

CPRet: A Dataset, Benchmark, and Model for Retrieval in Competitive Programming

Han Deng
Yuan Meng
Shixiang Tang
Wanli Ouyang
Xinzhu Ma

Competitive programming is widely used to evaluate the coding and reasoning abilities of large language models. However, the growing presence of duplicate or highly similar problems raises concerns not only about competition fairness, but also about the validity of competitive programming as a benchmark for model evaluation. We introduce a retrieval-oriented benchmark suite for competitive programming, covering four retrieval tasks—two code-centric (Text-to-Code, Code-to-Code) and two newly proposed problem-centric tasks (Problem-to-Duplicate, Simplified-to-Full)—built from a combination of automatically crawled problem–solution data and manually curated annotations. Our contribution includes both high-quality training data and temporally separated test sets for reliable evaluation. We develop two task-specialized retrievers based on this dataset: CPRetriever-Code, trained with a novel Group-InfoNCE loss for problem–code alignment, and CPRetriever-Prob, fine-tuned for problem-level similarity. Both models achieve strong results and are open-sourced for local use. Finally, we analyze LiveCodeBench and find that high-similarity problems inflate model pass rates and reduce differentiation, underscoring the need for similarity-aware evaluation in future benchmarks.

PDF Details

ECAI Conference 2025 Conference Paper

Dynamic Model Fusion for Multi-Source Test-Time Adaptation

Yuan Xue 0013
Qinting Jiang
Yuan Meng
Xingxuan Zhang
Chen Tang
Jingyan Jiang
Zhi Wang 0001

Deep Neural Networks suffer significant performance degradation when faced with distribution shifts between training and test data. Test-time adaptation (TTA) has emerged as a practical solution that enables models to adapt to the shifted test distribution. Currently, most existing TTA methods are designed around a single model, which incorporate limited information from a singular data distribution. In practice, pre-trained models derived from diverse source domains are readily accessible, each capturing a distinct data distribution and containing complementary information. To exploit this diversity, we propose Model Fusion-based multi-source Test-Time Adaptation (MFTTA), which constructs a target model by fusing the parameters of multiple source models. Drawing inspiration from deep model fusion, we introduce a fine-grained fusion mechanism governed by an off-policy reinforcement learning agent, which dynamically assigns fusion weights based on the current data distribution. Furthermore, we design a correlation-aware model update strategy that prioritizes the source model most relevant to the incoming test data. Extensive experiments on standard out-of-distribution benchmarks demonstrate that our method effectively integrates knowledge from multiple source models, adapts robustly to dynamic distribution shifts, and alleviates the problem of forgetting in long-term adaptation.

Details

AAAI Conference 2025 Conference Paper

JAQ: Joint Efficient Architecture Design and Low-Bit Quantization with Hardware-Software Co-Exploration

Mingzi Wang
Yuan Meng
Chen Tang
Weixiang Zhang
Yijian Qin
Yang Yao
Yingxin Li
Tongtong Feng

The co-design of neural network architectures, quantization precisions, and hardware accelerators offers a promising approach to achieving an optimal balance between performance and efficiency, particularly for model deployment on resource-constrained edge devices. In this work, we propose the JAQ Framework, which jointly optimizes the three critical dimensions. However, effectively automating the design process across the vast search space of those three dimensions poses significant challenges, especially when pursuing extremely low-bit quantization. Specifical, the primary challenges include: (1) Memory overhead in software-side: Low-precision quantization-aware training can lead to significant memory usage due to storing large intermediate features and latent weights for backpropagation, potentially causing memory exhaustion. (2) Search time-consuming in hardware-side: The discrete nature of hardware parameters and the complex interplay between compiler optimizations and individual operators make the accelerator search time-consuming. To address these issues, JAQ mitigates the memory overhead through a channel-wise sparse quantization (CSQ) scheme, selectively applying quantization to the most sensitive components of the model during optimization. Additionally, JAQ designs BatchTile, which employs a hardware generation network to encode all possible tiling modes, thereby speeding up the search for the optimal compiler mapping strategy. Extensive experiments demonstrate the effectiveness of JAQ, achieving approximately 7% higher Top-1 accuracy on ImageNet compared to previous methods and reducing the hardware search time per iteration to 0.15 seconds.

PDF Details DOI

ICML Conference 2025 Conference Paper

Peripheral Memory for LLMs: Integration of Sequential Memory Banks with Adaptive Querying

Songlin Zhai
Yuan Meng
Yongrui Chen 0002
Yiwei Wang
Guilin Qi

Large Language Models (LLMs) have revolutionized various natural language processing tasks with their remarkable capabilities. However, challenges persist in effectively integrating new knowledge into LLMs without compromising their performance, particularly in the Large Language Models (LLMs) have revolutionized various natural language processing tasks with their remarkable capabilities. However, a challenge persists in effectively processing new information, particularly in the area of long-term knowledge updates without compromising model performance. To address this challenge, this paper introduces a novel memory augmentation framework that conceptualizes memory as a peripheral component (akin to physical RAM), with the LLM serving as the information processor (analogous to a CPU). Drawing inspiration from RAM architecture, we design memory as a sequence of memory banks, each modeled using Kolmogorov-Arnold Network (KAN) to ensure smooth state transitions. Memory read and write operations are dynamically controlled by query signals derived from the LLMs’ internal states, closely mimicking the interaction between a CPU and RAM. Furthermore, a dedicated memory bank is used to generate a mask value that indicates the relevance of the retrieved data, inspired by the sign bit in binary coding schemes. The retrieved memory feature is then integrated as a prefix to enhance the model prediction. Extensive experiments on knowledge-based model editing validate the effectiveness and efficiency of our peripheral memory.

Details

ECAI Conference 2025 Conference Paper

Safe APG: Accelerated Policy Gradient Algorithm for Secure Policy Updating in Reinforcement Learning

Jianan Lin
Yao Chen 0003
Zhengyang Ji
Yuan Meng
Bo Hou
Shaolin Tan

Inverse reinforcement learning (IRL) aims to infer the reward function from expert demonstrations. However, as IRL techniques are increasingly applied in high-stakes domains such as autonomous driving and military decision-making, reward function leakage has emerged as a critical risk, potentially leading to severe security threats and unintended consequences. To address this challenge, we propose Safe Accelerated Policy Gradient (Safe APG), a method designed to enhance learning security of the demonstrating agent by preventing observers from inferring its reward function. The core idea behind Safe APG is to incorporate a delicately constructed and theoretically guaranteed structural noise into Nesterov’s Accelerated Gradient (NAG) for policy updating, with the goal of concealing critical gradient information from the learning agent as well as keeping the geometric convergence property of NAG. The results from numerical experiments and simulations in reinforcement learning environments demonstrate that the proposed method not only significantly mitigates reward function leakage, but also achieves superior convergence rates even under the perturbation of the introduced structural noise.

Details

IROS Conference 2023 Conference Paper

Contact-Aware Shaping and Maintenance of Deformable Linear Objects With Fixtures

Kejia Chen 0005
Zhenshan Bing
Fan Wu 0015
Yuan Meng
André Kraft
Sami Haddadin
Alois C. Knoll

Studying the manipulation of deformable linear objects has significant practical applications in industry, including car manufacturing, textile production, and electronics automation. However, deformable linear object manipulation poses a significant challenge in developing planning and control algorithms, due to the precise and continuous control required to effectively manipulate the deformable nature of these objects. In this paper, we propose a new framework to control and maintain the shape of deformable linear objects with two robot manipulators utilizing environmental contacts. The framework is composed of a shape planning algorithm which automatically generates appropriate positions to place fixtures, and an object-centered skill engine which includes task and motion planning to control the motion and force of both robots based on the object status. The status of the deformable linear object is estimated online utilizing visual as well as force information. The framework manages to handle a cable routing task in real-world experiments with two Panda robots and especially achieves contact-aware and flexible clip fixing with challenging fixtures.

Details

AAAI Conference 2023 Conference Paper

DNG: Taxonomy Expansion by Exploring the Intrinsic Directed Structure on Non-gaussian Space

Songlin Zhai
Weiqing Wang
Yuanfang Li
Yuan Meng

Taxonomy expansion is the process of incorporating a large number of additional nodes (i.e., ''queries'') into an existing taxonomy (i.e., ''seed''), with the most important step being the selection of appropriate positions for each query. Enormous efforts have been made by exploring the seed's structure. However, existing approaches are deficient in their mining of structural information in two ways: poor modeling of the hierarchical semantics and failure to capture directionality of the is-a relation. This paper seeks to address these issues by explicitly denoting each node as the combination of inherited feature (i.e., structural part) and incremental feature (i.e., supplementary part). Specifically, the inherited feature originates from ''parent'' nodes and is weighted by an inheritance factor. With this node representation, the hierarchy of semantics in taxonomies (i.e., the inheritance and accumulation of features from ''parent'' to ''child'') could be embodied. Additionally, based on this representation, the directionality of the is-a relation could be easily translated into the irreversible inheritance of features. Inspired by the Darmois-Skitovich Theorem, we implement this irreversibility by a non-Gaussian constraint on the supplementary feature. A log-likelihood learning objective is further utilized to optimize the proposed model (dubbed DNG), whereby the required non-Gaussianity is also theoretically ensured. Extensive experimental results on two real-world datasets verify the superiority of DNG relative to several strong baselines.

PDF Details DOI