Hiromu Taketsugu

Perception4PhysicalAI🤖 PhD @Ukita_Lab, DC1, Toyoaki
Intern @Woven_ToyotaJP, Discord @liveinvista, Dad🇻🇳
Ex: Intern @ToyotaCRDL, @TTIC_Connect, TA @Matsuo_Lab

[Hiromu Taketsugu @ImlntheMiddle]

先日のCVIM@徳島にて，主著「物理シミュレーションに基づく大規模事前学習による物理的妥当性を備えた歩行者軌跡予測」が奨励賞を受賞しました🙌
ありがとうございます！！

先月IEEE名古屋支部国際会議研究発表賞を受賞したCVPR論文とあわせて，大学ホームページでも取り上げていただいております！

[Paavo ✈️ ICML @PaavoParmas]

Our presentation with @nissymori1 at ICML is ongoing (Poster #111)!
Come find out about how retrying leads to the emergence of stochastic exploration in reinforcement learning!
#ICML2026

[Paavo ✈️ ICML @PaavoParmas]

At ICML, tomorrow (Thu 14:30) we’re presenting ReMax: a simple idea that leads to stochastic exploration when greedily maximizing rewards over retries.
What I find exciting is that this idea did not stay in one corner of RL. It now connects to LLM post-training, bandits, image diversity, continuous control, and distributional RL. I created a new thread about our body of follow-up works🧵
#ICML2026 @icmlconf

[Soichiro Nishimori @nissymori1]

@Matsuo_Lab @PaavoParmas @sotetsuk @Tdash_Koz @t_kitamura14 I’ll be presenting my poster at Hall A #111 on Wed, July 8, 2026, 10:30 PM–12:15 AM PDT.

Please stop by if you’re around — I’d be happy to chat!

Paper: https://t.co/jtlQbXZfON
Code: https://t.co/M0IR1eKWwk
See you in Seoul!! 🇰🇷

[Soichiro Nishimori @nissymori1]

Our paper "Emergence of Exploration in Policy Gradient Reinforcement Learning via Retrying" is accepted at #ICML2026 🇰🇷

Proposed a novel exploration objective called ReMax, evaluating best of multiple trials under uncertainty.

The objective comes from the basic question,

Why do RL agents need to explore?

We argue it is because
♻️ Agents are allowed to retry (otherwise, the rational choice is the current best action).
📈 Return is uncertain (otherwise, no point in trying suboptimal actions.)

ReMax naturally captures these intuitions by modeling the distribution of returns and evaluating the maximum over multiple retries, thereby encouraging agents to select actions that are currently suboptimal but highly uncertain.

The diagram is inspired by the Vector Policy Optimization (VPO) paper.

🧵1/n

[Claude @claudeai]

Introducing Claude Science, a new app designed with every stage of research in mind.

Artifacts traced to their code, environments managed on demand, and 60+ optional scientific databases that you can connect.

Available now in beta.

[Shumpei Takezaki @shumpeiMaxwell]

📢 Call for Papers @eccvconf

The LIMIT Workshop at #ECCV2026 is accepting submissions!! 🚀

How can we learn strong representations when resources are limited? 🤔

We invite submissions on representation learning when data, labels, compute, or other resources are scarce.

⏰ Deadline: July 6, 2026, 23:59 AoE
🔗 https://t.co/hRM8BshLUl

Please consider submitting your work! 🙌

[Soichiro Nishimori @nissymori1]

New preprint on statistically principled cheating detection of the Coding agent 🧑‍💻

The key idea is capped evaluation: create multiple equally valid outputs for a coding task, but evaluate against one randomly sampled choice. With two choices, e.g., 🍎 or 🍊, non-cheating performance is capped at 50% in expectation. Significant cap violations provide a statistical signal of possible test-gaming or leakage.

Led by @skydddoogg — happy to be a co-author!

[Jitendra MALIK @JitendraMalikCV]

I want to offer some unsolicited advice to computer vision researchers jumping into robotics. Don't focus too much on VLMs, VLAs etc. That's fine, but the real action is at the sensorimotor level. Most of the open problems in robotics are in manipulation, which is about hand-object interaction, and contacts and forces are central. Proprioception and tactile sensing are as important as vision. Don't get seduced by cherry-picked demos. You can't do robotics without doing robotics.

[トヨタ 未来創生センター | Toyota Frontier Research Center @TOYOTA_FRC]

6月9日15:30–15:45 人工知能学会 #JSAI2026 [2G5-OS-47b-01]において、#未来創生センター 松﨑らが「マルチモーダル自律移動基盤モデル」と題し、RGB画像・深度画像・LiDARの融合で高い汎化性と障害物回避、頑健性、柔軟性を目指したEnd-to-end自律移動モデルの研究を紹介します。

[Michael Black @Michael_J_Black]

I’ve been capturing 3D human motion for 30 years and today is maybe the biggest day in that history. We are presenting MAMMA at CVPR (oral session 2A). MAMMA is a markerless multi-camera system that has accuracy similar to marker-based systems.

[手塚 龍希(Ryuki Tezuka) @Ryuki86919418]

Say hi to our mascot, DINO-kun!🦖💚
(Paper & code on the project page ☝️)

[手塚 龍希(Ryuki Tezuka) @Ryuki86919418]

🎉Our paper will be presented at #CVPR2026 Findings!

"Group-DINOmics: Incorporating People Dynamics into DINO for Self-supervised Group Activity Feature Learning" by Ryuki Tezuka, Chihiro Nakatani @china64681791 , and Norimichi Ukita.

🦖Project: https://t.co/c0b06K8zwT

#CVPR

[Takayuki Shinohara @shnhrtkyk]

🎉 Thrilled to share that our paper received the Best Paper Award at the CV4AEC at CVPR 2026!
Huge thanks to my co-authors, the organizers, and the entire CVPR community! 🙏
#CVPR2026 #CV4AEC

[Ryosuke Korekata @kkrr10_]

#ICRA2026 にて主著含む3件，
#CVPR2026 にて1件の発表を行います！

ICRA in Vienna 🇦🇹
- [TuI1I.403] Affordance RAG: Hierarchical Multimodal Retrieval with Affordance-Aware Embodied Memory for Mobile Manipulation
🔗 https://t.co/zO1KH4Nv92

[Hironobu Fujiyoshi @hf149]

世界モデルのチュートリアルを作成しました。1. 世界モデルとは 2. 自動運転における世界モデル 3. LeWorldModelという構成です。
https://t.co/Q0hr2lCqSz

[arutema47 @arutema47]

CVPR発表のプレスリリースとなります。
自動運転LiDARではガラスや光沢面によって幻の人や壁のような「ゴースト」が生じてしまい、それを取り除くのは難しい問題でした。この研究ではLiDARの反射波形＋Transformerによって効果的にゴースト除去するフレームワークとデータセットを提案してます。

[トヨタ 未来創生センター | Toyota Frontier Research Center @TOYOTA_FRC]

6月5日、Workshop「Geometry in the Age of Data-Driven Robotics（#ICRA2026）」において、#未来創生センター 金井が、当センター最新のカメラ自己校正に関する研究成果を発表します。学習ベースのカメラ自己較正手法の脆弱性を再検討し、提案手法 ZeCNOBA によって対処できる可能性を示します。

[tokenbender @tokenbender]

i yearn for the ever-fleeting moment of excellence and some men just channel the inner divine.

[マエダ タカキ @takaki_maeda_99]

Gemma4をハッカソンでVLAにしてみました。
色々やっててめちゃくちゃおもろいのでぜひ見て！
初めてVLAアーキテクチャ作ってみて、超難しかったし大変でしたが、なんとか形にはできてよかったです。
クソ勉強になったし楽しかった！

https://t.co/UFX8GMei31

[Yuki Kondo @y_kondo_vision]

同研究室B4の前田君，松ケ谷君らが、#Kaggle Gemma 4 Good Hackathonで「GEM-4」を公開！
Gemmaをロボットアームと接続し、ユーザーの意図に応じて物理的に支援する身体性AIデモを短期間で実装しています。B4で短期間でここまで形にしていて本当にすごいです。
https://t.co/aTJSG3YicX