Skip to content
View s-elim's full-sized avatar
🎯
Focusing
🎯
Focusing

Highlights

  • Pro

Block or report s-elim

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
s-elim/README.md
Md Selim Sarowar — M.Sc. researcher in world models, vision-language-action models and agentic robotics at the Advanced Visual Intelligence Lab, Yeungnam University

Homepage  Google Scholar  LinkedIn  ResearchGate  Curriculum vitae  Email

“I aim to become a rising star in Physical Super Intelligence.”

Open to research collaborations and PhD opportunities · get in touch

About · News · Research · Publications · In Progress · Toolbox · Journey · Honors · Contact

About

I am an M.Sc. (Research) student in Computer Vision and AI at Yeungnam University (YU), South Korea, and a member of the Advanced Visual Intelligence Lab advised by Prof. Sungho Kim. I am a Global Korea Scholarship (GKS) scholar funded by NIIED, with additional support from the RLRC and RISE industry-collaboration projects, and I currently hold a perfect CGPA of 4.5/4.5, ranked first in my program.

My research spans 3D computer vision and spatial intelligence, vision-language-action (VLA) models, latent world models, and agentic robotics for robot manipulation. The goal is embodied agents that perceive, reason, and act in the physical world, integrating visual-spatial representations, predictive world models, multimodal reasoning, and action generation, working toward physical super intelligence.

PROGRAM
M.Sc. Computer Vision & AI
Yeungnam University · 4.5/4.5, ranked first
LAB
Advanced Visual Intelligence Lab
Prof. Sungho Kim · Graduate researcher
FUNDING
Global Korea Scholarship
NIIED, Republic of Korea
PUBLICATIONS
2 × BMVC'26 · 2 × SCIE-Q1
First-author BMVC'26 and IEEE Access
PATENTS
4 granted
Korea · United Kingdom · India
PATH
Bangladesh → India → Taiwan → Korea
Bengali · English · Hindi · Korean

World Models H/V-JEPA VLAs Agentic Robotics Spatial Intelligence 6D Pose Estimation Causal Learning Self Modeling RL Fine-Tuning 5D Embodied & Humanoid Intelligence

Full biography

My recent work includes geometry-aware VLA models, latent world models, and agentic approaches to robust robot manipulation, alongside earlier work in 6D pose estimation and 3D visual representation learning. It includes first-author work accepted at the British Machine Vision Conference (BMVC 2026) and a first-author publication in IEEE Access (SCIE-Q1), spanning geometry-aware spatial reasoning for VLA models and data-efficient RGB-D 6D pose estimation.

Before the M.Sc., I completed a one-year Korean Language and Literature program at Yeungnam University under the GKS program, and spent a year as a full-time Research Assistant at the WCSP Lab, National Tsing Hua University (NTHU), Taiwan, working on machine learning and privacy-preserving methods.

I earned my B.Tech in Electronics and Electrical Engineering from the Kalinga Institute of Industrial Technology (KIIT), India, as a recipient of the Government of India's Study in India scholarship, where I worked on image processing and machine learning. I also interned as a machine learning intern at IIT Roorkee, on image-processing and feature-extraction methods for biomedical signals.

News

Earlier
  • Sep 2025   Started the M.Sc. (Research) at Yeungnam University after one year of Korean Language and Literature under GKS.
  • Aug 2024   Moved to South Korea as a Global Korea Scholarship scholar.
  • Sep 2023   Joined the WCSP Lab, National Tsing Hua University, Taiwan, as a Research Assistant.
  • Jul 2023   Completed the B.Tech with distinction at KIIT, India.

Research

One arc, four stages: recover geometry, reason over it in a policy, predict the outcome in latent space, then act and measure.

flowchart LR
    P["Perceive<br/>3D vision · 6D pose"] --> R["Reason<br/>spatial tokens · VLAs"]
    R --> W["Predict<br/>latent world models"]
    W --> A["Act<br/>agentic manipulation"]
    A --> S(["Physical Super Intelligence"])
    classDef stage fill:#4f46e514,stroke:#4f46e5,stroke-width:1.5px
    classDef goal fill:#7c3aed22,stroke:#7c3aed,stroke-width:2px
    class P,R,W,A stage
    class S goal
Loading
Stage Question Work
Perceive How precisely can a robot recover object pose and geometry? C3G-VM6D · VLM6D · VFM-VLM · MSP
Reason How should 3D geometry enter a vision-language-action policy? GaussVLA · GST-VLA
Predict Can an agent imagine the outcome before it acts? Agentic-LWM · AGI
Act Can affordances drive zero-shot manipulation, and how do we measure it? AFFORD-X · PLUMB-Bench

Selected Publications

GaussVLA architecture

GaussVLA: Geometry-Aware Spatial Reasoning for Vision-Language-Action Model
Md Selim Sarowar, Md Tanvir Islam, Sungho Kim*, Sangtae Ahn
BMVC 2026, CORE A
Project page · OpenReview

BibTeX
@inproceedings{Sarowar_2026_BMVC,
  author    = {Md Selim Sarowar and Md Tanvir Islam and Sungho Kim and Sangtae Ahn},
  title     = {GaussVLA: Geometry-Aware Spatial Reasoning for Vision-Language-Action Model},
  booktitle = {37th British Machine Vision Conference 2026, {BMVC} 2026, Lancaster, UK, November 23-26, 2026},
  publisher = {BMVA},
  year      = {2026},
  url       = {https://bmva-archive.org.uk/bmvc/2026/assets/papers/Paper_121/paper.pdf}
}
HNH40K risk-weighted learning model

HNH40K: A Robust Dataset and Risk-Weighted Learning for Hand Filtering
Ashikuzzaman†, Md Selim Sarowar†, Md Tanvir Islam, Sangtae Ahn, Khan Muhammad
BMVC 2026, CORE A
Code · OpenReview

BibTeX
@inproceedings{Ashikuzzaman_2026_BMVC,
  author    = {Ashikuzzaman and Md Selim Sarowar and Md Tanvir Islam and Sangtae Ahn and Khan Muhammad},
  title     = {HNH40K: A Robust Dataset and Risk-Weighted Learning for Hand Filtering},
  booktitle = {37th British Machine Vision Conference 2026, {BMVC} 2026, Lancaster, UK, November 23-26, 2026},
  publisher = {BMVA},
  year      = {2026},
  url       = {https://bmva-archive.org.uk/bmvc/2026/assets/papers/Paper_652/paper.pdf}
}
C3G-VM6D architecture

C3G-VM6D: Data-Efficient C3G Vision Model Aided 6D Pose Estimation Based on RGB-D Data
Md Selim Sarowar, Manar Alnaasan, Sungho Kim*
IEEE Access 2026, SCIE Q1
Paper · Code

BibTeX
@article{11400555,
  author  = {Sarowar, Md Selim and Alnaasan, Manar and Kim, Sungho},
  journal = {IEEE Access},
  title   = {C3G-VM6D: Data-Efficient C3G Vision Model Aided 6D Pose Estimation Based on RGB-D Data},
  year    = {2026},
  volume  = {14},
  pages   = {34223-34238},
  doi     = {10.1109/ACCESS.2026.3666439}
}
Fusion VLM-Gait framework

Fusion VLM-Gait: RGB-D Fusion in Vision-Language Models for Explainable Multi-Task Parkinsonian Gait Analysis
Manar Alnaasan, Md Selim Sarowar, Sungho Kim*
Scientific Reports 2026, SCIE Q1
Paper · PDF

BibTeX
@article{alnaasan2026fusionvlmgait,
  title     = {Fusion VLM-Gait: RGB-D fusion in vision-language models for explainable multi-task Parkinsonian gait analysis},
  author    = {Alnaasan, Manar and Sarowar, Md Selim and Kim, Sungho},
  journal   = {Scientific Reports},
  year      = {2026},
  publisher = {Nature Publishing Group},
  doi       = {10.1038/s41598-026-67440-x},
  url       = {https://www.nature.com/articles/s41598-026-67440-x}
}
GST-VLA architecture

GST-VLA: Structured Gaussian Spatial Tokens for 3D Depth-Aware Vision-Language-Action Models
Md Selim Sarowar, Sungho Kim*
ICML 2026 Workshop, MusIML
OpenReview · PDF

BibTeX
@inproceedings{sarowar2026gstvla,
  title     = {{GSTVLA}: Structured Gaussian Spatial Tokens for 3D Depth-Aware Vision-Language-Action Models},
  author    = {Md Selim Sarowar and Sungho Kim},
  booktitle = {The 6th Muslims in ML (MusIML) Workshop at ICML 2026},
  year      = {2026},
  url       = {https://openreview.net/forum?id=J57nR3hyAd}
}

† Equal contribution  ·  * Corresponding author

More papers, preprints and patents

Papers

  • Dependency-Aware Reliable Orchestration for Smart-IoT Control with Reasoning-Aligned LLMs · IEEE ICECET 2026 · Paper · Code
  • VFM-VLM: Vision Foundation Model and Vision Language Model based Visual Comparison for 3D Pose Estimation · IEIE 2026 · arXiv
  • VLM6D: VLM based 6DoF Pose Estimation based on RGB-D Images · IEIE 2025 · arXiv
  • Hand Gesture Recognition Systems: A Review of Methods, Datasets, and Emerging Trends · IJCA 2025 · Paper
  • Improvement of Denoising in Images Using Generic Image Denoising Network (GID Net) · IEEE AESPC 2021 · Paper

Preprints

  • Vision-Language-Action and Vision Language Models for Robot Manipulation: A Comprehensive Review Towards Real-World Applications · Under review · Preprint
  • What Matters: Datasets or Robust Frameworks in Modern Robot Learning? · 2026 · Preprint

Granted patents

  • Compact 3D-Gaussian-Based 6D Object Pose Estimation Method for Physical AI · Korean Patent · 2026
  • AI-Enabled Coding and Algorithm Simulation Device · UK Design Patent · 2025
  • IoT-Based System and Design Approach for Health Monitoring of Wearable Devices · Indian Patent · 2023
  • An Experimental Special Technology in Emergency Management and Disaster Response · Indian Patent · 2022

Full list on the homepage and Google Scholar.

In Progress

AGI framework

AGI: Agentic Intelligence and Latent Lookahead for Robust Manipulation
Under review
Project page

AFFORD-X framework

AFFORD-X: Affordance-Grounded Agentic Policy for Zero-Shot Robot Manipulation
Under review
Project page · Code

MSP framework

MSP: Planning on a Residual-Based Pose Posterior Closes the Grasp-Success Gap Between Estimated and Annotated Object Poses
Under review
Project page · Code

Agentic-LWM framework

Agentic-LWM: Grounding Multimodal Policies in Hierarchical Latent World Model
Under review
Project page

PLUMB-Bench scenes

PLUMB-Bench: A Physically Verified Benchmark for Precise, Generalizable Robot Manipulation
Under review
Project page · Code

Toolbox

Deep learningPython PyTorch Hugging Face TensorFlow CUDA
3D visionOpenCV Open3D Blender Matplotlib
Robotics and simulationROS 2 Isaac Sim Isaac Lab
HardwareJetson AGX Orin RGB-D and LiDAR
Agentic systemsOpenClaw ClaudeClaw SpatialClaw
WritingLaTeX

Academic Journey

2025 – nowM.Sc. (Research), Computer Vision and AI · Yeungnam University, South Korea
Advanced Visual Intelligence Lab, advised by Prof. Sungho Kim · GKS scholar
2024 – 2025Korean Language and Literature program · Yeungnam University, South Korea
Global Korea Scholarship
2023 – 2024Research Assistant · National Tsing Hua University, Taiwan
WCSP Lab
2019 – 2023B.Tech, Electronics and Electrical Engineering · KIIT, India
Study in India scholar · ML intern at IIT Roorkee
Journey infographic
Academic journey: M.Sc. and graduate research at Yeungnam University (South Korea, GKS), research and teaching assistant at National Tsing Hua University (Taiwan), B.Tech at KIIT University (India), and St. Joseph Higher Secondary School (Bangladesh)

Honors & Awards

  • Global Korea Scholarship (GKS) · NIIED, Republic of Korea · 2024–2027 · full funding for the language program and the research M.Sc.
  • NTHU International Graduate Students Assistantship · the highest award for international students at National Tsing Hua University
  • Study in India Government Scholarship · Ministry of Education, India · 2019–2023 · fully funded B.Tech
  • KIIT University International Scholarship (KUISP) · 2019–2023 · academic excellence
More
  • Bangladesh Sweden Trust Fund (BSTF) Travel Grant · 2024 · research internship in Taiwan
  • Honorary Young Diplomat · Jeju Province · promoting Korean culture and international engagement
  • National Youth Inspiration Award · Citizens Open Forum · 2017 · Youth Leadership and Connectivity
  • Meritorious Scholarship, Secondary School Certificate · Bangladesh Ministry of Education · 2016–2018

Activity

Animated contribution graph

Contact

Tip

Open to research collaborations and PhD opportunities in world models, vision-language-action models, and agentic robotics. Reach me by email, on LinkedIn, or through my homepage.

Pinned Loading

  1. s-elim s-elim Public

    Python 1

  2. openpi openpi Public

    Forked from Physical-Intelligence/openpi

    Python

  3. PythonRobotics PythonRobotics Public

    Forked from AtsushiSakai/PythonRobotics

    Python sample codes and textbook for robotics algorithms.

    Python