2026
SlotVLA: Towards Modeling of Object-Relation Representations in Robotic Manipulation
A slot-attention-based VLA framework for compact object-relation representations in robotic manipulation.
Taisei Hanyu
M.S. Student in Computer Science · AICV Lab, University of Arkansas
I am a master's student in Computer Science at the University of Arkansas, where I am a member of the AICV Lab, supervised by Dr. Ngan Le. I received my Honors B.S. in Computer Science from the University of Arkansas.
Multimodal AIEmbodied AIRobot Learning
Recent News
- SlotVLA accepted to ICRA 2026.
- Memory-state paper published in the AAAI 2026 proceedings.
- SolarFormer++ published in IEEE Transactions on Industry Applications.
- AerialFormer published in Remote Sensing.
- Open-Fusion presented as an oral paper at ICRA 2024.
- Open-Fusion released for real-time open-vocabulary 3D mapping.
Latest Projects
All projectsSlotVLA
A slot-attention-based VLA framework for compact object-relation representations in robotic manipulation.
LIBERO-Mem
A non-Markovian manipulation benchmark and slot-centric VLA framework for memory-aware robot policies.
OBEYED-VLA
A VLA framework that separates perceptual grounding from action reasoning for clutter-resistant robot manipulation.
Open-Fusion
Real-time open-vocabulary 3D mapping and queryable scene representation from RGB-D observations.
Publications
All publications2026
CodeGraphVLP: Code-as-Planner Meets Semantic-Graph State for Non-Markovian Vision-Language-Action Models
A non-Markovian VLA framework that combines persistent semantic-graph state with executable code-as-planner reasoning.
2026
Rethinking Progression of Memory State in Robotic Manipulation: An Object-Centric Perspective
A memory-aware robotic manipulation benchmark and slot-centric VLA model for non-Markovian manipulation tasks.
2025
Clutter-Resistant Vision-Language-Action Models through Object-Centric and Geometry Grounding
A VLA framework that uses object-centric and geometry-grounded views for clutter-resistant robot manipulation.
2025
SolarFormer++: Multi-scale Transformer for Solar PV Profiling and Obstruction Localization for Degradation Mitigation
A multi-scale Transformer approach for solar PV profiling and obstruction localization for degradation mitigation.
2024
AerialFormer: Multi-resolution Transformer for Aerial Image Segmentation
A multi-resolution Transformer for semantic segmentation of aerial imagery.
2024
Open-Fusion: Real-time Open-Vocabulary 3D Mapping and Queryable Scene Representation
Real-time open-vocabulary 3D mapping and queryable scene representation using RGB-D observations.
2023
SolarFormer: Multi-scale Transformer for Solar PV Profiling
A multi-scale Transformer model for solar PV profiling from aerial imagery.