Vignesh Prasad

Trying to teach robots how to shitpost

Robot Learning Post-Doc w/ @GeorgiaChal
Ex: @ias_tudarmstadt @TCSResearch @iiit_hyderabad

(Also a comedian)

[Vignesh Prasad @FatAndFurious42]

@HumanoidsConf @serena_ivaldi @lsentis Just for context, the screenshot in the previous message was 6 mins before the deadline

[Vignesh Prasad @FatAndFurious42]

@HumanoidsConf The deadline was supposed to be 24th July AoE but the RAS papercept server shows all deadlines in Pacific Time and new submissions were closed 5 hours before 23:59 AoE . Could you reopen the portal so those affected by this bug can reupload??

@serena_ivaldi @lsentis

[Cockroach Janta Party - CJP @Cockroachisback]

Peak Dictatorship Happening At Safdarjung Hospital By Govt

Sonam sir’s wife is not allowed to take her phone inside hospital room. His medical reports are not shared with the family. No personal lawyers or doctors are allowed to meet

[Yaw Nyame, MD | Mind & Body @Yaw_nyame35]

The AfD's party leader Alice Weidel distinguishes between Germans and "passport Germans."

People like me. People who acquired citizenship rather than being born with it.

She was asked about this in 2019. She defended it again in 2025. The implication is clear: there are real Germans and there are legal Germans. One category is permanent. The other is conditional.

I did not need to hear this from a politician to know it exists. I have felt it in rooms, in silences, in the way some people recalibrate their warmth when they learn where I am from. But hearing it stated openly, from a party polling at 30%, is a different thing entirely.

If you are an immigrant in Germany and you are not paying attention to this, you should be.

[Georgia Chalvatzaki @GeorgiaChal]

Just returned from Vienna after attending the @ieee_ras_icra #ICRA2026—my very first one back after maternity leave!

What an incredible experience. I truly think this was one of the best ICRAs to date. Huge congratulations to Markus Vincze, Tamim Asfour, and the entire organizing team for pulling off such a monumental event! 🇦🇹🤖

The PEARL Lab—along with our youngest honorary member 👶—had a phenomenal time. Between presenting our latest work and engaging in deeply fruitful discussions with the community, it was incredibly rewarding. Traveling to a major conference with a toddler is a completely different kind of logistics challenge, but seeing the community so welcoming makes it all worth it!

I am also profoundly grateful to have co-organized a highly successful @ieeeras Women in Engineering (WiE) Luncheon. This was a bittersweet milestone, as it was my final one serving as Chair of the committee. I’m immensely proud of what we have achieved together so far and thrilled to hand the reins over to the brilliant new leadership.

Now, I’m shifting gears back to my duties as Program Co-Chair for the @corl_conf #CoRL2026 ⚙️

I am leaving Vienna fully recharged with fresh ideas and new perspectives for our ongoing research into Structured Robot Learning. We have some incredibly exciting work ahead for the SIREN project, generously funded by the @ERC_Research and the the Alfried Krupp von Bohlen und Halbach-Stiftung.

Stay tuned for what’s next from https://t.co/7ncjMayHwi! 🚀

#Robotics #RobotLearning #WomenInSTEM #WomenInRobotics #EmbodiedAI

[Pankhuri Vanjani @VanjaniPankhuri]

Most VLA models force high- and low-frequency signals into one synchronized stream. But in real manipulation, vision, contact, and proprioception evolve on very different time scales. Presenting our new work: DAM-VLA Decoupled Asynchronous Multimodal Vision-Language-Action Model

[yufeng jin @yjin_1118]

How long should it take to go from 🗣️ "Evaluate π0 on LIBERO" to a real-robot rollout?

We're betting on ⏱️ 1.5 hours.

Introducing 🐚 Nautilus — one prompt. Any policy. Any benchmark. Any robot.

An agentic harness for robot learning + @Anthropic Claude Code plugin.
🔗 https://t.co/CAwfBOlQPD

#Robotics #RobotLearning #LLMAgents #ClaudeCode #Harness

[Vignesh Prasad @FatAndFurious42]

What happens in VeGAS makes your embodied agents smarter...

Check out our latest CVPR findings work on how mistake-aware verifier training helps improve MLLM-based action planning

If you're at CVPR next week hit up @nishadsinghi who'll be presenting our work

[Nishad Singhi @nishadsinghi]

Embodied agents are getting better at reasoning… but they still make surprisingly brittle mistakes.
A robot that can “bring me a banana”
may completely fail at: “bring me a yellow curved fruit.”

Why?

Because current agents usually commit to the first action they come up with.

We introduce VeGAS: Verifier-Guided Action Selection for embodied agents. 🤖

Instead of acting immediately, the agent:
• samples multiple candidate actions
• verifies them
• executes the best one

Accepted as a #CVPR 2026 Findings paper.

🧵

[Vignesh Prasad @FatAndFurious42]

@aks1812 Heroes!!!

[Vignesh Prasad @FatAndFurious42]

Want to know how to learn effective multisensory representations that make your robot robust to partying?
Check out our latest RAL paper by @RKrohn_ : https://t.co/9MIj9wwaWV

[Rickmer Krohn @RKrohn_]

Solving contact-rich manipulation tasks even under disco light🪩?
Incorporate multiple sensors using MSDP and obtain a robust policy in under 55 min!
Excited to share our work accepted at RA-L: Self-Supervised Multisensory Pretraining for Contact-Rich Robot Reinforcement Learning

[Junyao Shi @JunyaoShi]

This is what happens when novelty, not results, becomes the North Star. The real work is often in the system details, but those rarely make it into the spotlight. Instead, we get unnecessary “novelty” that’s easier to package, pitch, and hype.

Will there ever be a time when system details are treated as first-class citizens in papers?

[Georgia Chalvatzaki @GeorgiaChal]

🚀 Hiring PhDs & Postdocs in Structured Robot Learning & Embodied AI @TUDarmstadt (PEARL Lab)

🤖 We study how structure in the robot–environment system can be exploited to learn robust, adaptive, and generalizable behaviors, beyond black-box policies

🔬 Topics:
• Grounding (language → perception → action)
• Structured world models
• VLA + control + memory
• RL (credit assignment, offline→online)
• Whole-body & bimanual mobile manipulation

🇪🇺 New EU Lighthouse Project on Generative AI for Robotics + ERC StG SIREN

👉 Full call: https://t.co/VxcornBuo9

Please repost 🙏

#RobotLearning #EmbodiedAI #Robotics #MachineLearning #PhDPositions #Postdoc

[Chenhao Li @breadli428]

Simplicity should be valued more. When a task can be solved equally with a simpler framework, one should not be blamed having “nothing new”.

Many unnecessary novelties are invented for the sake of novelty, while the effort of making simpler methods general is not appreciated.

[yufeng jin @yjin_1118]

🚀 New work accepted to ICRA 2026!
We introduce a flow-matching framework on the SE(3) manifold. Instead of just a single guess, our model captures the full pose distribution—perfect for handling tricky occlusions, symmetries, and partial views in real-world robotics.

Huge thanks to my amazing collaborators! @n_w_funk @softraeh @FatAndFurious42 @Jan_R_Peters @GeorgiaChal

📄 Paper: https://t.co/yygG8gAEav
🌐 Website: https://t.co/q4irmOBRZd

[Vignesh Prasad @FatAndFurious42]

Alma mater researching things that matter 🔥🔥🔥

[neural nets. @cneuralnetwork]

this is an actual research paper 😭

[Kosta Derpanis (back home, Toronto ❤️) @CSProfKGD]

Repeat after me: This is NOT a paper contribution, it’s an expected component!

[Abijit Ganguly @AbijitG]

Anyone with basic brains can figure out what Indigo did was very intentional. This was not sudden disruption, this was almost a planned strike to strongarm the government into giving it what it wants. What people went through doesn't matter to them, it's just collateral damage. And the fact that govt let this happen is on them. The fact that we essentially have monopolistic/duopolistic situations in so many major sectors is on this Government. But what to do, let's look for ways to blame RaGa on this too.

[Vignesh Prasad @FatAndFurious42]

How can we effectively integrate information from multiple sensors, such as cameras, force-torque, and joint proprioception, for real-world contact-rich manipulation?

Check out @RKrohn_'s work on multisensory representation learning for RL: https://t.co/kSFCzfkttY

[Rickmer Krohn @RKrohn_]

Multisensory Dynamic Pretraining (MSDP) is based on masked autoencoding and cross-sensor prediction, leading to rich sensor fusion. Usage of the FT-sensor boosts performance by 14%! Our approach is flexible regarding the number and type of sensors. More Details in the Preprint!

[Vignesh Prasad @FatAndFurious42]

@rz_california While I agree, RAL also has long timelines for conferences. Eg., to present at IROS in Oct. 2026, a RAL paper needs to be accepted and transferred by April 30th, so the paper should be submitted in Nov. 2025 in the most ideal case, almost a year before the conference

[Vignesh Prasad @FatAndFurious42]

Let's gooo!!!!

[VishalSid | CW: Jojos @VishalSid_1]

INDIAN BROS WE ARE SOO BACK
THIS LOOKS WAY TOO GOOD😭😭🔥

WE WILL BE SEATED🗣️🔥🔥
#BahubaliTheEternalWar