⚡ In the last few weeks, I had the great chance to present our recent work on embodiment-aware reinforcement learning for robot control and design 🤖

🏆 Tim Faust received the Freudenberg award for his thesis on Model Simplification in Value-based Deep RL!! 🎉 Tim did his thesis at our lab supervised by @theo-vincent.bsky.social This award rewards the best master's thesis across all fields of research TU Darmstadt with 10k€

Congratulations to the whole team for this achievement!!

Théo Vincent@theo-vincent.bsky.social · 2w ago

🏅Gradient Iterated Temporal-Difference Learning received Best Paper Award for empirical RL research @rl-conference.bsky.social🥳

The RL4VLA Workshop has come to an end after an incredible Friday at #RSS2026 in Sydney 🇦🇺 We brought together researchers to discuss how RL can advance VLA models, the remaining challenges, and what it will take for robotics to have its own "RL moment" as we've seen with LLMs.

I had a lovely time in Seoul presenting Trust Region Inverse RL (arxiv.org/pdf/2605.11020) at @icmlconf.bsky.social! Also got a chance to visit and give a talk at the Robot Learning Lab at SNU (rllab.snu.ac.kr). They’re doing amazing work on RL for robotics 🦾

The progression. Three balls on the first attempt, four and five by the second.

Learning five-ball juggling on the second attempt, with two Barrett WAMs. Most humans take years of practice. Paper and videos: kai-ploeger.com/residual-juggling #robotics #juggling

I had a great time visiting Prof. Kevin S. Luck at @vuamsterdam.bsky.social and to present our recent works on embodiment-aware learning for robot control and design! So much cool work here in Amsterdam about Robot Co-Design! There is huge potential in the data-driven way of designing robots!

🚨 Less than 24 hours to submit your work to our RL4VLA workshop @rss #RSS2026

@ahmed-hendawy.bsky.social · 3mo ago

🚨 Deadline Extension Alert! The RL4VLA Workshop submission deadline has been extended by one week! 🗓️ New deadline: June 15 (AoE) Still time to submit your work. See you at #RSS2026

🚨 Deadline Extension Alert! The RL4VLA Workshop submission deadline has been extended by one week! 🗓️ New deadline: June 15 (AoE) Still time to submit your work. See you at #RSS2026

@ahmed-hendawy.bsky.social · 3mo ago

Enjoying #ICRA2026? 🤖 Submit to the 𝗥𝗟𝟰𝗩𝗟𝗔 𝗪𝗼𝗿𝗸𝘀𝗵𝗼𝗽 @ #RSS2026 🇦🇺 📅 Deadline: 𝗝𝘂𝗻𝗲 𝟴 (𝗔𝗼𝗘) 📍 Sydney 🗓️ 𝗝𝘂𝗹𝘆 𝟭𝟳 RL × VLA × Robotics 🤖 OpenReview: openreview.net/group?id=rob... Reviewers welcome: 🔗 docs.google.com/forms/d/e/1F... #RSS2026 #RL #Robotics #VLA

Submit your work to the RL4VLA Workshop @ #RSS2026 Deadline is June 8 (AoE)

@ahmed-hendawy.bsky.social · 3mo ago

Enjoying #ICRA2026? 🤖 Submit to the 𝗥𝗟𝟰𝗩𝗟𝗔 𝗪𝗼𝗿𝗸𝘀𝗵𝗼𝗽 @ #RSS2026 🇦🇺 📅 Deadline: 𝗝𝘂𝗻𝗲 𝟴 (𝗔𝗼𝗘) 📍 Sydney 🗓️ 𝗝𝘂𝗹𝘆 𝟭𝟳 RL × VLA × Robotics 🤖 OpenReview: openreview.net/group?id=rob... Reviewers welcome: 🔗 docs.google.com/forms/d/e/1F... #RSS2026 #RL #Robotics #VLA

Excited to be at @icra2026 this week in Vienna 👋 Looking forward to lots of interesting conversations and meeting great people throughout the week. Catch our work here:

See you at #ICRA2026! 🚀 Our lab will be presenting exciting new work across robotics, control, learning. Check out our papers below 👇

Check out the newest work by @nicobohlinger.bsky.social on using value gradients from multi-embodiment RL for robot design!

Nico Bohlinger@nicobohlinger.bsky.social · 3mo ago

⚡️What if we could design robots with gradients? 🤖 Introducing Shape Your Body: we train one multi-embodiment policy + value function, then optimize new robot designs through value gradients. 🔗 Try out our interactive demo here: nico-bohlinger.github.io/shape-your-b...

Are semi-gradient updates the way to go?🤔 Turns out new gradient TD methods can compete with semi-gradient methods on complex benchmarks!

Théo Vincent@theo-vincent.bsky.social · 4mo ago

Why do we keep using semi-gradient methods when they can diverge?🤨 Gradient TD methods are often overlooked, while they have convergence guarantees! @rl-conference.bsky.social, we will present the first gradient TD method shown to be competitive against semi-gradient methods on deep RL benchmarks🏆

📢 Excited to share that our paper 𝗧𝗿𝘂𝘀𝘁 𝗥𝗲𝗴𝗶𝗼𝗻 𝗜𝗻𝘃𝗲𝗿𝘀𝗲 𝗥𝗲𝗶𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 (𝗧𝗥𝗜𝗥𝗟) has been accepted at @icmlconf.bsky.social 2026! 🌐 Project page: lnkd.in/dqNi68Mk 📄 Paper: lnkd.in/dzbcCbjU #ICML2026 #ReinforcementLearning #InverseRL #ImitationLearning #Robotics

🚀 Excited to announce the 𝗥𝗟𝟰𝗩𝗟𝗔 Workshop @ #RSS2026 in Sydney, Australia 🇦🇺 Bringing together researchers working on reinforcement learning for Vision Language Action (VLA) systems, embodied AI, and scalable robot learning. 📅 Submission deadline: June 8th, 2026 (AoE)

The deadline for this workshop is coming up soon 👉 May 15th (AoE)

RL in Big Worlds@rlcbigworlds.bsky.social · 4mo ago

We are proud to have an amazing line-up of speakers! They will present their works, which incorporate the constraint that the world is bigger than the agent and impossible to anticipate, observe, or model perfectly. We are also looking forward to the panel discussion!

I had an amazing time today visiting Professor Luiz Chaimowicz's lab in Belo Horizonte! I really enjoyed discovering the research being done at UFMG. It seems to be an amazing place to do some research! See you @iclr-conf.bsky.social in Rio

I'm in Rio 🇧🇷 for #ICLR! Catch me presenting MINTO 🌿 this Friday, April 24 | Afternoon | Pavilion 4 #4603. MINTO 🌿 uses the online network when it helps speed up RL training without sacrificing stability. Let's chat about RL and your research. DM me or simply come say hi!

I will be presenting 3 papers @iclr-conf.bsky.social this week 🇧🇷 Looking forward to some interesting exchanges!

Interested in world models? Join the conversation @iclr-conf.bsky.social 👇

Théo Vincent@theo-vincent.bsky.social · 5mo ago

🌍World models can play an important role towards building general agents, but what should be their role in decision-making?🕹️ @joemwatson.bsky.social and I are organizing a Social @iclr-conf.bsky.social on this topic🎙️ 🗓️Feel free to join the conversation on Friday 24th April, at noon!!

Working with constrained agents in complex environments? Do not hesitate to submit your latest work to this workshop! See you in Montréal @rl-conference.bsky.social 🇨🇦

RL in Big Worlds@rlcbigworlds.bsky.social · 5mo ago

RL in Big Worlds is a workshop at @rl-conference.bsky.social about ideas that enable agents to achieve goals in environments vastly more complex than themselves. This requires giving agents the ability to learn continually and use approximate value functions, models, and policies effectively.

Benchmarking always takes a ton of time😮‍💨 and we often hear about it🗣️ But we rarely report the carbon footprint of experiments, which better reflects their weight! Here is the electricity emission of the experiments in each paper of my PhD👇

Excited to release FeLaN open-source: ✅ JAX training & evaluation pipelines ✅ Baselines: MJX, DeLaN, MLP ✅ Differentiable RNEA in MJX ✅ Open datasets (sim + real): Go2, Talos, Spot, HyQReal2 💻 github.com/schulze18/felan 🌐 schulze18.github.io/felan_website 📄 arxiv.org/abs/2510.17270

GitHub - Schulze18/felan: Open-source code of Floating-Base Deep Lagrangian Networks (FeLaN).

Open-source code of Floating-Base Deep Lagrangian Networks (FeLaN). - Schulze18/felan

github.com

Proud to share our latest work, accepted at @iclr-conf.bsky.social 2026: APPLE! 🍎 TL;DR: APPLE is a novel reinforcement learning framework for solving active perception problems. #ICLR2026 #Robotics #MachineLearning #ActivePerception #RL @ias-tudarmstadt.bsky.social

Should we use a target network in deep value-based RL?🤔 The answer has always been YES or NO, as there are pros and cons. @iclr-conf.bsky.social, I will present iS-QN, a method that lies in between this binary view, collecting the pros while reducing the cons🚀

🧵 Accepted at @iclr-conf.bsky.social! Target networks stabilize bootstrapping in RL 🛡️ But induce slow-moving targets 🐢 Online networks adapt fast ⚡ But can diverge with function approximation 💥 𝗠𝗜𝗡𝗧𝗢 🌿 uses the online network 𝗼𝗻𝗹𝘆 𝗶𝗳 𝗶𝘁 𝗰𝗮𝗻 — yielding faster 𝘢𝘯𝘥 more stable RL. Here’s how 👇

🥳 Accepted @iclr-conf.bsky.social 🥳 💡 This paper shows an easy way to halve the memory footprint of the critic of any temporal-difference learning algorithm while maintaining good performance 🪶

Théo Vincent@theo-vincent.bsky.social · 7mo ago

Should we use a target network in deep value-based RL?🤔 The answer has always been YES or NO, as there are pros and cons. @iclr-conf.bsky.social, I will present iS-QN, a method that lies in between this binary view, collecting the pros while reducing the cons🚀

🥳Our paper "Floating-Base Deep Lagrangian Networks (FeLaN)" has been accepted to #ICRA2026. FeLaN: a grey-box approach for physically consistent SysID of floating-base robots (humanoids, quadrupeds). 📄 arxiv.org/abs/2510.17270 💻 Soon! 🌐 schulze18.github.io/felan_website/