Shun Iwase

TRI's latest Large Behavior Model (LBM) paper landed on arxiv last night! Check out our project website: toyotaresearchinstitute.github.io/lbm1/ One of our main goals for this paper was to put out a very careful and thorough study on the topic to help people understand the state of the…

Presenting DemoDiffusion: An extremely simple approach enabling a pre-trained 'generalist' diffusion policy to follow a human-demonstration for a novel task during inference One-shot human imitation *without* requiring any paired human-robot data or online RL 🙂 1/n

Imagine if robots could fill in the blanks in cluttered scenes. ✨ Enter RaySt3R: a single masked RGB-D image in, complete 3D out. It infers depth, object masks, and confidence for novel views, and merges the predictions into a single point cloud. rayst3r.github.io

Imitation learning has a data scarcity problem. Introducing EgoDex from Apple, the largest and most diverse dataset of dexterous human manipulation to date — 829 hours of egocentric video + paired 3D hand poses across 194 tasks. Now on arxiv: arxiv.org/abs/2505.11709 (1/4)

Big shoutout to @NicholasEPfaff for this incredible work! If you’re interested in scalable conditional scene generation for robotics, you should check this out! 🤖🔥

My timeline is mixed with a lot of good news (e.g., promotion, job update) and April fool jokes since the Japanese academic year begins in April 😂

Aerospace is a reason enough for me to switch back to mac from arch linux... this is the best tiling manager on mac os I've ever used github.com/nikitabobko/Ae…

RL is notoriously sample inefficient. How can we scale RL on tasks much slower to simulate than rigid body physics, such as soft bodies? In our #ICLR2025 spotlight, we introduce both a new first-order RL algorithm, SAPO, and differentiable simulation platform, Rewarped. 1/n

always bet on people who spent their childhoods hacking video game ROMs

SFT Memorizes, RL Generalizes arxiv.org/abs/2501.17161

thrilled about this awesome initiative! super excited about reasoning models and building these datasets 🚀

8年間日本を見てきました。東大で学生も教えましたし客員准教授もやりました。日本の若いエンジニアや研究者、いろんな人とも話しました。日本に足りないのは、真なる競争です。空気を読む競争ではないです。

タリンまた行きたいなー

Does #CVPR2025 have a clear policy on concurrent work? An ECCV '24 paper which was published a month and half before the CVPR dealine could be considered concurrent? I know ICLR has explicit guidelines, but I couldn’t find anything similar for CVPR

Regarding the MS robotics program at CMU, students typically have a good chance to get some sort of financial support since it’s a research-based program. It’s not guaranteed and I agree that the fees are quite expensive though

The slide is bad, her response to an audience is even worse… “Maybe there is one, maybe they are common, who knows what. I hope it was an outlier."