So Kuroki - 黒木 颯

So Kuroki is a research engineer at Sakana AI, focusing on developing LLM agents and enhancing their skill acquisition. He was working in robot learning at OMRON SINIC X and in Yutaka Matsuo's Lab at The University of Tokyo. When he was a student, he had the great opportunity to be mentored by Shixiang Shane Gu and Yujin Tang, Google Brain Team. He holds a bachelor's and master's degree from the University of Tokyo.

His research interests and experiences strongly align with collective Intelligence (robotics, learning) and AI collaboration with humans (HCI). He loves to see agents become smarter by interacting and sharing their knowledge, and in the same context, he loves to see AI empower humans through collaboration. 

KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI

So Kuroki, Yotaro Kubo, Takuya Akiba, Yujin Tang

arXiv

Reimagining Agent-based Modeling with Large Language Model Agents via Shachi

So Kuroki, Yingtao Tian, Kou Misaki, Takashi Ikegami, Takuya Akiba, Yujin Tang

arXiv

LAPPI: Interactive Optimization with LLM-Assisted Preference-Based Problem Instantiation

So Kuroki, Manami Nakagawa, Shigeo Yoshida, Yuki Koyama, Kozuno Tadashi

arXiv

Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search

Yuichi Inoue, Kou Misaki, Yuki Imajuku, So Kuroki, Taishi Nakamura, Takuya Akiba

NeurIPS 2025 (spotlight)

arXiv

Agent Skill Acquisition for Large Language Models via CycleQD

So Kuroki, Taishi Nakamura, Takuya Akiba, Yujin Tang

ICLR 2025

arXiv / github / blog: JP or EN

Multi-Agent Behavior Retrieval: Retrieval-Augmented Policy Training for Cooperative Push Manipulation by Mobile Robots

So Kuroki, Mai Nishimura, Tadashi Kozuno

IROS 2024

Project page / arXiv / paper / video

Language-Guided Pattern Formation for Swarm Robotics with Multi-Agent Reinforcement Learning

Hsu-Shen Liu, So Kuroki, Tadashi Kozuno, Wei-Fang Sun, Chun-Yi Lee

IROS 2024

Project page / paper / video

GenDOM: Generalizable One-shot Deformable Ojbect Manipulation with Parameter-Aware Policy

So Kuroki, Jiaxian Guo, Tatsuya Matsushima, Takuya Okubo, Masato Kobayashi, Yuya Ikeda, Ryosuke Takanami, Paul Yoo, Yutaka Matsuo, Yusuke Iwasawa

ICRA 2024

Project page / arXiv / github / paper / video

Swarm Body: Embodied Swarm Robots 

Sosuke Ichihashi, So Kuroki, Mai Nishimura, Kazumi Kasaura,                                 Kazutoshi Tanaka, Shigeo Yoshida

CHI 2024

Popular Choice Honorable Mention (Interactivity demo from accepted paper)

arXiv / video / paper / github / Blog (Japanese)

Collective Intelligence for 2D push Manipulations with Mobile Robots

So Kuroki, Tatsuya Matsushima, Jumpei Arima, Hiroki Furuta, Yutaka Matsuo, Shixiang Shane Gu, and Yujin Tang 

RA-L (presented at IROS 2023), 3/2023

Project page / arXiv / github / paper / slide / video

Predicting the presence of E. coli in tap water using machine learning in Nepal

So Kuroki, Ryuji Ogata, Maiko Samakoto

Water and Environment Journal, 1/2023

github / paper

Immersive Real World through Deep Billboards

Naruya Kondo, So Kuroki, Ryosuke Hyakuta, Yutaka Matsuo, Shixiang Shane Gu, Yoichi Ochiai

SIGGRAPH 2022 Immersive Pavilion

Project page / arXiv / github / paper

2D Rendering-based Pseudo 3D Models for Immersive VR Experiments

Naruya Kondo, So Kuroki, Ryosuke Hyakuta, Yutaka Matsuo, Shixiang Shane Gu, Yoichi Ochiai

The 36th Annual Conference of the Japanese Society for Artificial Intelligence, 2022

paper (Japanese)

Shape evaluation of activated sludge flocs under different operating conditions through quantitative image analysis

Hakoshima Suguru, So Kuroki, Tobino Tomohiro, Nakajima Fumiyuki

Excellent Poster Award at The 55nd Japan Society on Water Environment, 3/2021.

Project page

Lite 6 teleoperation

DYNAMIXEL

2025 New Year holiday project (Hobby) - half-day

Reference: GELLO (https://wuphilipp.github.io/gello_site/)

Drone creation from scratch 

Microcontroller programming (Arduino Nano) / PID / Frame design / 3D printer / Receiver+controller /

Project page / gitHub, Blog (in Japanese)

2024 New year holiday project (Hobby) - 10 days

Language-conditioned manipulation with xArm

Encode language commands with CLIP / xArm ROS control / teleoperated data collection / behavior cloning

A Language command "cut the left side of the tofu" control the left figure xArm. 

2024 spring - This turns out to be GenDOM (ICRA 2024)

Water Pipe Facility Ledger and Deterioration Diagnosis Business in Asago

@Fracta (intern from 6/2020-6/2021)

supported by Ministry of Health, Labour and Welfare, Japan

Published information (in Japanese)