All # 3D vision # Embodied Perception # Autonomous Driving # Robotics # 3D Reconstruction # 3D Editing # 3D Scene Understanding # Benchmarks # Adversarial Attacks # Others
My Research

My current research focuses on developing robust 3D visual representations that bridge geometric reconstruction, semantic understanding, and downstream decision-making. Traditionally, 3D reconstruction has been viewed primarily as a graphics or geometry problem, while reasoning and navigation have been studied separately in robotics. My work connects these areas by exploring how 3D scene representations can move beyond passive reconstruction to become active, useful computational structures for perception, memory, and reasoning in complex real-world environments. I investigate how explicit geometric representations can enhance spatial perception in autonomous vehicles, how representations like 3D Gaussian Splatting (3DGS) can function as a persistent memory for embodied agents, and how geometry-aware representations can improve semantic understanding.

Representative publications are highlighted.

2026

arXiv
Reconstruction Matters: Learning Geometry-Aligned BEV Representation through 3D Gaussian Splatting

Yiren Lu; Xin Ye; Burhaneddin Yaman; Jingru Luo; Zhexiao Xiong; Liu Ren; Yu Yin.

arXiv preprint 2026

We propose Splat2BEV, a Gaussian Splatting-assisted BEV framework that learns emantically rich and geometrically precise BEV feature representations.

[Paper] [Project]

# 3D vision # 3D Scene Understanding # Autonomous Driving
Reconstruction Matters: Learning Geometry-Aligned BEV Representation through 3D Gaussian Splatting

Yiren Lu; Xin Ye; Burhaneddin Yaman; Jingru Luo; Zhexiao Xiong; Liu Ren; Yu Yin.

We propose Splat2BEV, a Gaussian Splatting-assisted BEV framework that learns emantically rich and geometrically precise BEV feature representations.

[Paper] [Project]

# 3D vision # 3D Scene Understanding # Autonomous Driving
arXiv
arXiv
GSMem: 3D Gaussian Splatting as Persistent Spatial Memory for Zero-Shot Embodied Exploration and Reasoning

Yiren Lu; Yi Du; Disheng Liu; Yunlai Zhou; Chen Wang; Yu Yin.

arXiv preprint 2026

GSMem is a zero-shot embodied exploration and reasoning framework that utilize 3DGS as persistent memory.

[Paper] [Project]

# 3D vision # Embodied Perception # Robotics
GSMem: 3D Gaussian Splatting as Persistent Spatial Memory for Zero-Shot Embodied Exploration and Reasoning

Yiren Lu; Yi Du; Disheng Liu; Yunlai Zhou; Chen Wang; Yu Yin.

GSMem is a zero-shot embodied exploration and reasoning framework that utilize 3DGS as persistent memory.

[Paper] [Project]

# 3D vision # Embodied Perception # Robotics
arXiv
DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filtering
arXiv
DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filtering

Yiran Qiao; Yiren Lu; Yunlai Zhou; Rui Yang; Linlin Hou; Yu Yin; Jing Ma.

arXiv preprint 2026

We introduce a novel frequency-aware defense strategy DefenseSplat for 3DGS.

[Paper]

# 3D vision # Adversarial Attacks
DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filtering

Yiran Qiao; Yiren Lu; Yunlai Zhou; Rui Yang; Linlin Hou; Yu Yin; Jing Ma.

We introduce a novel frequency-aware defense strategy DefenseSplat for 3DGS.

[Paper]

# 3D vision # Adversarial Attacks
arXiv
AdvSplat: Adversarial Attacks on Feed-Forward Gaussian Splatting Models
arXiv
AdvSplat: Adversarial Attacks on Feed-Forward Gaussian Splatting Models

Yiran Qiao; Yiren Lu; Yunlai Zhou; Rui Yang; Linlin Hou; Yu Yin; Jing Ma.

arXiv preprint 2026

We introduce AdvSplat, the first systematic study of adversarial attacks on feed-forward 3DGS.

[Paper]

# 3D vision # Adversarial Attacks
AdvSplat: Adversarial Attacks on Feed-Forward Gaussian Splatting Models

Yiran Qiao; Yiren Lu; Yunlai Zhou; Rui Yang; Linlin Hou; Yu Yin; Jing Ma.

We introduce AdvSplat, the first systematic study of adversarial attacks on feed-forward 3DGS.

[Paper]

# 3D vision # Adversarial Attacks
arXiv
UniDrive-WM: Unified Understanding, Planning and Generation World Model For Autonomous Driving
arXiv
UniDrive-WM: Unified Understanding, Planning and Generation World Model For Autonomous Driving

Zhexiao Xiong; Xin Ye; Burhan Yaman; Sheng Cheng; Yiren Lu; Jingru Luo; Nathan Jacobs; Liu Ren.

arXiv preprint 2026

We propose UniDrive-WM, a unified VLM-based world model that jointly performs scene understanding, trajectory planning, and trajectory-conditioned future image generation within a single architecture.

[Paper] [Project]

# Embodied Perception # Autonomous Driving
UniDrive-WM: Unified Understanding, Planning and Generation World Model For Autonomous Driving

Zhexiao Xiong; Xin Ye; Burhan Yaman; Sheng Cheng; Yiren Lu; Jingru Luo; Nathan Jacobs; Liu Ren.

We propose UniDrive-WM, a unified VLM-based world model that jointly performs scene understanding, trajectory planning, and trajectory-conditioned future image generation within a single architecture.

[Paper] [Project]

# Embodied Perception # Autonomous Driving
arXiv

2025

Segment then Splat: Unified 3D Open-Vocabulary Segmentation via Gaussian Splatting
NeurIPS
Segment then Splat: Unified 3D Open-Vocabulary Segmentation via Gaussian Splatting

Yiren Lu; Yunlai Zhou; Yiran Qiao; Chaoda Song; Tuo Liang; Jing Ma; Yu Yin.

NeurIPS 2025

We propose Segment then Splat, which performs segmentation before reconstruction by dividing Gaussians into object sets upfront, eliminating semantic/geometric ambiguity and accelerating optimization.

[Paper] [Project] [Code]

# 3D vision # 3D scene understanding
Segment then Splat: Unified 3D Open-Vocabulary Segmentation via Gaussian Splatting

Yiren Lu; Yunlai Zhou; Yiran Qiao; Chaoda Song; Tuo Liang; Jing Ma; Yu Yin.

We propose Segment then Splat, which performs segmentation before reconstruction by dividing Gaussians into object sets upfront, eliminating semantic/geometric ambiguity and accelerating optimization.

[Paper] [Project] [Code]

# 3D vision # 3D scene understanding
NeurIPS
Fix False Transparency by Noise Guided Splatting
NeurIPS
Fix False Transparency by Noise Guided Splatting

Aly El Hakie*; Yiren Lu*; Yu Yin; Michael W. Jenkins; Yehe Liu. (* equal contribution)

NeurIPS 2025

We propose Noise Guided Splatting to address false transparency artifacts in 3D Gaussian Splatting by injecting opaque noise Gaussians in object volumes during training.

[Paper] [Project] [Code]

# 3D vision # 3D Reconstruction
Fix False Transparency by Noise Guided Splatting

Aly El Hakie*; Yiren Lu*; Yu Yin; Michael W. Jenkins; Yehe Liu. (* equal contribution)

We propose Noise Guided Splatting to address false transparency artifacts in 3D Gaussian Splatting by injecting opaque noise Gaussians in object volumes during training.

[Paper] [Project] [Code]

# 3D vision # 3D Reconstruction
NeurIPS
BARD-GS: Blur-Aware Reconstruction of Dynamic Scenes via Gaussian Splatting
CVPR
BARD-GS: Blur-Aware Reconstruction of Dynamic Scenes via Gaussian Splatting

Yiren Lu; Yunlai Zhou; Disheng Liu; Tuo Liang; Yu Yin.

CVPR 2025

We introduce BARD-GS, a robust dynamic scene reconstruction method that explicitly decomposes motion blur into camera and object components to handle blurry inputs and imprecise camera poses.

[Paper] [Project] [Code] [Data]

# 3D vision # 3D Reconstruction
BARD-GS: Blur-Aware Reconstruction of Dynamic Scenes via Gaussian Splatting

Yiren Lu; Yunlai Zhou; Disheng Liu; Tuo Liang; Yu Yin.

We introduce BARD-GS, a robust dynamic scene reconstruction method that explicitly decomposes motion blur into camera and object components to handle blurry inputs and imprecise camera poses.

[Paper] [Project] [Code] [Data]

# 3D vision # 3D Reconstruction
CVPR
Spatial Intelligence in Vision-Language Models: A Comprehensive Survey
arXiv
Spatial Intelligence in Vision-Language Models: A Comprehensive Survey

Disheng Liu; Tuo Liang; Zhe Hu; Jierui Peng; Yiren Lu; Yi Xu; Yun Fu; Yu Yin.

arXiv 2025

A survey on spatial reasoning in Vision-Language Models.

[Paper] [Project] [Code]

# Others
Spatial Intelligence in Vision-Language Models: A Comprehensive Survey

Disheng Liu; Tuo Liang; Zhe Hu; Jierui Peng; Yiren Lu; Yi Xu; Yun Fu; Yu Yin.

A survey on spatial reasoning in Vision-Language Models.

[Paper] [Project] [Code]

# Others
arXiv
Counterfactual Visual Explanation via Causally-Guided Adversarial Steering
arXiv
Counterfactual Visual Explanation via Causally-Guided Adversarial Steering

Yiran Qiao; Disheng Liu; Yiren Lu; Yu Yin; Mengnan Du; Jing Ma.

arXiv preprint 2025

We propose a causally-guided adversarial framework for counterfactual visual explanation that avoids spurious perturbations and improves explanation quality.

[Paper]

# Adversarial Attacks
Counterfactual Visual Explanation via Causally-Guided Adversarial Steering

Yiran Qiao; Disheng Liu; Yiren Lu; Yu Yin; Mengnan Du; Jing Ma.

We propose a causally-guided adversarial framework for counterfactual visual explanation that avoids spurious perturbations and improves explanation quality.

[Paper]

# Adversarial Attacks
arXiv
When 'YES' Meets 'BUT': Can Large Models Comprehend Contradictory Humor Through Comparative Reasoning?
arXiv
When 'YES' Meets 'BUT': Can Large Models Comprehend Contradictory Humor Through Comparative Reasoning?

Tuo Liang; Zhe Hu; Jing Li; Hao Zhang; Yiren Lu; Yunlai Zhou; Yiran Qiao; Disheng Liu; Jeirui Peng; Jing Ma; Yu Yin.

arXiv preprint 2025

Extension of Yes-But, Yes-But (V2).

[Paper]

# Benchmarks
When 'YES' Meets 'BUT': Can Large Models Comprehend Contradictory Humor Through Comparative Reasoning?

Tuo Liang; Zhe Hu; Jing Li; Hao Zhang; Yiren Lu; Yunlai Zhou; Yiran Qiao; Disheng Liu; Jeirui Peng; Jing Ma; Yu Yin.

Extension of Yes-But, Yes-But (V2).

[Paper]

# Benchmarks
arXiv
CAUSAL3D: A Comprehensive Benchmark for Causal Learning from Visual Data
arXiv
CAUSAL3D: A Comprehensive Benchmark for Causal Learning from Visual Data

Disheng Liu; Yiran Qiao; Wuche Liu; Yiren Lu; Yunlai Zhou; Tuo Liang; Yu Yin; Jing Ma.

arXiv 2025

We introduce Causal3D, a novel and comprehensive benchmark that integrates structured data (tables) with corresponding visual representations (images) to evaluate causal reasoning.

[Paper]

# Benchmarks
CAUSAL3D: A Comprehensive Benchmark for Causal Learning from Visual Data

Disheng Liu; Yiran Qiao; Wuche Liu; Yiren Lu; Yunlai Zhou; Tuo Liang; Yu Yin; Jing Ma.

We introduce Causal3D, a novel and comprehensive benchmark that integrates structured data (tables) with corresponding visual representations (images) to evaluate causal reasoning.

[Paper]

# Benchmarks
arXiv

2024

View-consistent Object Removal in Radiance Fields
ACM MM
View-consistent Object Removal in Radiance Fields

Yiren Lu; Jing Ma; Yu Yin.

ACM MM 2024

We introduce a novel radiance field editing pipeline that significantly enhances consistency by requiring inpainting of only a single reference image.

[Paper] [Project]

# 3D vision # 3D Editing
View-consistent Object Removal in Radiance Fields

Yiren Lu; Jing Ma; Yu Yin.

We introduce a novel radiance field editing pipeline that significantly enhances consistency by requiring inpainting of only a single reference image.

[Paper] [Project]

# 3D vision # 3D Editing
ACM MM
iSLAM: Imperative SLAM
RA-L
iSLAM: Imperative SLAM

Taimeng Fu; Shaoshu Su; Yiren Lu; Chen Wang.

Robotics and Automation Letters (RA-L) 2024

We proposed a novel self-supervised imperative learning framework, named imperative SLAM (iSLAM), which fosters reciprocal correction between the front-end and back-end and enhances performance without external supervision.

[Paper] [Project] [Code] [Comment]

# 3D vision # Robotics
iSLAM: Imperative SLAM

Taimeng Fu; Shaoshu Su; Yiren Lu; Chen Wang.

We proposed a novel self-supervised imperative learning framework, named imperative SLAM (iSLAM), which fosters reciprocal correction between the front-end and back-end and enhances performance without external supervision.

[Paper] [Project] [Code] [Comment]

# 3D vision # Robotics
RA-L
Cracking the Code of Juxtaposition: Can AI Models Understand the Humorous Contradictions
NeurIPS
Cracking the Code of Juxtaposition: Can AI Models Understand the Humorous Contradictions

Zhe Hu; Tuo Liang; Jing Li; Yiren Lu; Yunlai Zhou; Yiran Qiao; Jing Ma; Yu Yin.

NeurIPS 2024

We introduce the YesBut benchmark, which comprises tasks aimed at assessing AI's capabilities in understanding comics with contradictory narratives.

[Paper] [Project]

# Benchmarks
Cracking the Code of Juxtaposition: Can AI Models Understand the Humorous Contradictions

Zhe Hu; Tuo Liang; Jing Li; Yiren Lu; Yunlai Zhou; Yiran Qiao; Jing Ma; Yu Yin.

We introduce the YesBut benchmark, which comprises tasks aimed at assessing AI's capabilities in understanding comics with contradictory narratives.

[Paper] [Project]

# Benchmarks
NeurIPS

2023

PyPose v0.6: The Imperative Programming Interface for Robotics
IROS
PyPose v0.6: The Imperative Programming Interface for Robotics

Zitong Zhan; Xiangfu Li; Qihang Li; Haonan He; Abhinav Pandey; Haitao Xiao; Yangmengfei Xu; Xiangyu Chen; Kuan Xu; Kun Cao; Zhipeng Zhao; Zihan Wang; Huan Xu; Zihang Fang; Yutian Chen; Wentao Wang; Xu Fang; Yi Du; Tianhao Wu; Xiao Lin; Yuheng Qiu; Fan Yang; Jingnan Shi; Shaoshu Su; Yiren Lu; Taimeng Fu; Karthik Dantu; Jiajun Wu; Lihua Xie; Marco Hutter; Luca Carlone; Sebastian Scherer; Daning Huang; Yaoyu Hu; Junyi Geng; Chen Wang.

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) Workshop 2023

[Paper]

# Robotics
PyPose v0.6: The Imperative Programming Interface for Robotics

Zitong Zhan; Xiangfu Li; Qihang Li; Haonan He; Abhinav Pandey; Haitao Xiao; Yangmengfei Xu; Xiangyu Chen; Kuan Xu; Kun Cao; Zhipeng Zhao; Zihan Wang; Huan Xu; Zihang Fang; Yutian Chen; Wentao Wang; Xu Fang; Yi Du; Tianhao Wu; Xiao Lin; Yuheng Qiu; Fan Yang; Jingnan Shi; Shaoshu Su; Yiren Lu; Taimeng Fu; Karthik Dantu; Jiajun Wu; Lihua Xie; Marco Hutter; Luca Carlone; Sebastian Scherer; Daning Huang; Yaoyu Hu; Junyi Geng; Chen Wang.

[Paper]

# Robotics
IROS

2022

Multical: Spatiotemporal calibration for multiple IMUs, cameras and LiDARs
IROS
Multical: Spatiotemporal calibration for multiple IMUs, cameras and LiDARs

Xiangyang Zhi; Jiawei Hou; Yiren Lu; Laurent Kneip; Soren Schwertfeger.

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2022

This paper presents a general sensor calibration method, named Multical, that makes use of multiple planar calibration targets whose poses will be estimated alongside spatiotemporal calibration.

[Paper] [Code]

# Robotics
Multical: Spatiotemporal calibration for multiple IMUs, cameras and LiDARs

Xiangyang Zhi; Jiawei Hou; Yiren Lu; Laurent Kneip; Soren Schwertfeger.

This paper presents a general sensor calibration method, named Multical, that makes use of multiple planar calibration targets whose poses will be estimated alongside spatiotemporal calibration.

[Paper] [Code]

# Robotics
IROS
End to End Face Reconstruction via Differentiable PnP
ECCV
End to End Face Reconstruction via Differentiable PnP

Yiren Lu; Huawei Wei.

European Conference on Computer Vision 2022

[Paper]

# 3D vision
End to End Face Reconstruction via Differentiable PnP

Yiren Lu; Huawei Wei.

[Paper]

# 3D vision
ECCV

2020

Improving CNN-based planar object detection with geometric prior knowledge
SSRR
Improving CNN-based planar object detection with geometric prior knowledge

Jianxiong Cai; Jiawei Hou; Yiren Lu; Hongyu Chen; Laurent Kneip; Soren Schwertfeger.

IEEE International Symposium on Safety, Security, and Rescue Robotics 2020

[Paper]

# Robotics
Improving CNN-based planar object detection with geometric prior knowledge

Jianxiong Cai; Jiawei Hou; Yiren Lu; Hongyu Chen; Laurent Kneip; Soren Schwertfeger.

[Paper]

# Robotics
SSRR