Selected work

Representative Papers

Optimal Design of Continuum Robots With Reachability Constraints

Apr 01, 2021IEEE Robotics and Automation Letters

Addressing the structural design challenge of multi-segment continuum robots caused by curvature coupling, this paper proposes a joint topology and sizing optimization method with workspace reachability as a hard constraint. The approach innovatively integrates numerical reachability analysis, torque-aware inverse kinematics modeling, and the Estimation of Distribution Algorithm (EDA), enabling minimization of joint torques under forward/inverse kinematic consistency constraints. Compared to conventional genetic algorithms, EDA improves the composite performance—measuring both robot length and actuation energy consumption—by 4–15% across three representative tasks, significantly enhancing solution quality and convergence efficiency. To the best of our knowledge, this work is the first to synergistically combine reachability analysis, torque-aware kinematics, and EDA for continuum robot structural optimization. It establishes a novel paradigm for autonomous configuration design that simultaneously achieves high workspace reachability and low energy consumption.

9 citationsRead paper

Configuration Design of Mechanical Assemblies using an Estimation of Distribution Algorithm and Constraint Programming

Jun 01, 2019IEEE Congress on Evolutionary Computation

This paper addresses the challenging configuration design optimization problem in mechanical engineering—characterized by discrete decision variables, stringent constraints, and black-box objective evaluations. To tackle these challenges, we propose a novel evolutionary algorithm integrating the Bivariate Marginal Distribution Algorithm (BMDA) with Constraint Programming (CP). Key innovations include an adaptive chi-square test for dependency identification among variables and a Gibbs sampling strategy for probabilistic model-based offspring generation, complemented by a CP-driven feasibility repair operator to ensure constraint satisfaction. Evaluated on a vehicle suspension design case study, the method achieves a 37% faster convergence rate and improves the best-found solution performance by 22% compared to standard genetic algorithms and other estimation-of-distribution algorithms (EDAs), while satisfying 100% of all engineering constraints. These results demonstrate the framework’s efficacy, robustness, and practical applicability in automated, complex mechanical configuration design.

3 citationsRead paper

FLOORA: A Human-Aligned Domain-Specific Language Model for Architectural Design

Sep 28, 2026

This study addresses the limited generative capability of general-purpose large language models in structured engineering tasks such as architectural layout generation. To this end, we propose a lightweight, domain-specific language (DSL) model tailored for architectural design. Methodologically, we design a token-efficient DSL paired with a custom tokenizer, and optimize human alignment through supervised fine-tuning (SFT) combined with reinforcement learning incorporating verifiable rewards. Experimental results demonstrate that this compact model substantially outperforms large frontier models, achieving a 96% win rate under vision-language model evaluation, an 89.3% preference rate in human assessment, and significantly improved geometric validity. Overall, this work establishes an efficient, domain-adapted training paradigm for vertical applications.

0 citationsRead paper

ReFM: Semantic-Aware Refinement Flow Model for Motion Retargeting

Sep 25, 2026

This study addresses the challenge of learning semantics for motion retargeting without high-quality paired data, as well as the bottleneck whereby a single regression objective struggles to jointly satisfy physical and semantic constraints. To this end, we propose a source-mesh-agnostic, energy-guided model. Methodologically, retargeting is reformulated as a progressive refinement pipeline compatible with diverse initialization strategies. An SO(3) normalizer and a contrastive cross-character semantic encoder are introduced to enforce semantic consistency without paired supervision. Furthermore, the framework integrates flow matching, inverse kinematics, and energy function optimization. By effectively preserving semantic intent while ensuring physical plausibility, the proposed approach significantly outperforms conventional direct regression methods, yielding substantial improvements in overall retargeting quality.

0 citationsRead paper

From Interpretability to Control: Insights from Six Years of the TrustNLP Workshop

Aug 11, 2026

This study systematically traces the evolutionary trajectory of Trustworthy Natural Language Processing (TrustNLP) research, focusing on the dynamic development across six key trust dimensions. Through an analysis of 144 papers from six TrustNLP workshops—integrated with multidimensional taxonomies from TrustLLM and DecodingTrust, cross-conference thematic comparisons, and temporal modeling—the work reveals a paradigm shift from post-hoc explainability of static models toward mechanistic understanding and proactive control in generative models. The findings indicate an explosive growth in truthfulness research (37% of papers) and a U-shaped resurgence in explainability. Moreover, the release of high-impact chat models has significantly accelerated trust-related investigations across all dimensions, with TrustNLP’s topical distribution closely mirroring that of mainstream NLP conferences.

0 citationsRead paper

Recent publications

Latest Papers

FLOORA: A Human-Aligned Domain-Specific Language Model for Architectural Design

Sep 28, 2026

This study addresses the limited generative capability of general-purpose large language models in structured engineering tasks such as architectural layout generation. To this end, we propose a lightweight, domain-specific language (DSL) model tailored for architectural design. Methodologically, we design a token-efficient DSL paired with a custom tokenizer, and optimize human alignment through supervised fine-tuning (SFT) combined with reinforcement learning incorporating verifiable rewards. Experimental results demonstrate that this compact model substantially outperforms large frontier models, achieving a 96% win rate under vision-language model evaluation, an 89.3% preference rate in human assessment, and significantly improved geometric validity. Overall, this work establishes an efficient, domain-adapted training paradigm for vertical applications.

0 citationsRead paper

ReFM: Semantic-Aware Refinement Flow Model for Motion Retargeting

Sep 25, 2026

This study addresses the challenge of learning semantics for motion retargeting without high-quality paired data, as well as the bottleneck whereby a single regression objective struggles to jointly satisfy physical and semantic constraints. To this end, we propose a source-mesh-agnostic, energy-guided model. Methodologically, retargeting is reformulated as a progressive refinement pipeline compatible with diverse initialization strategies. An SO(3) normalizer and a contrastive cross-character semantic encoder are introduced to enforce semantic consistency without paired supervision. Furthermore, the framework integrates flow matching, inverse kinematics, and energy function optimization. By effectively preserving semantic intent while ensuring physical plausibility, the proposed approach significantly outperforms conventional direct regression methods, yielding substantial improvements in overall retargeting quality.

0 citationsRead paper

From Interpretability to Control: Insights from Six Years of the TrustNLP Workshop

Aug 11, 2026

This study systematically traces the evolutionary trajectory of Trustworthy Natural Language Processing (TrustNLP) research, focusing on the dynamic development across six key trust dimensions. Through an analysis of 144 papers from six TrustNLP workshops—integrated with multidimensional taxonomies from TrustLLM and DecodingTrust, cross-conference thematic comparisons, and temporal modeling—the work reveals a paradigm shift from post-hoc explainability of static models toward mechanistic understanding and proactive control in generative models. The findings indicate an explosive growth in truthfulness research (37% of papers) and a U-shaped resurgence in explainability. Moreover, the release of high-impact chat models has significantly accelerated trust-related investigations across all dimensions, with TrustNLP’s topical distribution closely mirroring that of mainstream NLP conferences.

0 citationsRead paper

CNS-Edit++: Category-Agnostic 3D Editing with Coupled Neural Shape Representation

Jul 17, 2026

Existing 3D editing methods often rely on category-specific models, limiting their generalizability. This work proposes a category-agnostic framework for versatile 3D shape editing based on Coupled Neural Shapes (CNS), which integrates a global semantic latent code with a 3D neural feature volume to enable diverse editing operations through joint optimization. The approach introduces two key innovations: a KV-cache replacement mechanism and latent feature regularization, which together allow precise control over edited regions while preserving geometric consistency in non-edited areas. Compatible with various 3D generative models, the method achieves state-of-the-art quantitative and qualitative results across multiple benchmarks, significantly enhancing both editing flexibility and shape fidelity.

0 citationsRead paper

EPS3D: End-to-End Feed-Forward 3D Panoptic Segmentation

Jun 07, 2026

This work addresses key limitations in open-vocabulary 3D panoptic segmentation—namely, reliance on preprocessing pipelines, error propagation, and inconsistency between semantic and instance predictions—by introducing the first end-to-end feedforward framework that directly predicts 3D semantic and instance features from multi-view images. The core innovation lies in the bidirectional Ins2Sem and Sem2Ins mutual enhancement modules, which explicitly model consistency between semantic and instance representations. Integrated with multi-view feature fusion and a distillation-based training strategy, the method achieves state-of-the-art performance on benchmarks such as Replica, improving semantic mIoU by 13% over existing approaches. It also enables real-time inference at just one second per scene, offering an optimal balance of accuracy and efficiency for applications like robotic manipulation and 3D editing.

0 citationsRead paper