-
Tweets458
-
Followers4K
-
Following279
-
Likes657
Over the last few weeks we have seen some evidence of benchmark manipulation on RoboArena, in part flagged by members of the robotics community. We take this very seriously and have taken steps to protect the integrity of the benchmark and ensure fair, unbiased evaluations. Most importantly, going forward only organizations without an active policy on the leaderboard will be permitted to submit evaluations. We have applied these changes retroactively across the period of suspicious benchmark activity, to make sure all evals on the leaderboard can be trusted. A big thank you to all the RoboArena evaluators who volunteer their time to provide unbiased robot evaluations, and to the community members who helped audit the evaluation results and flagged potential wrong-doing.
We observed evidence of benchmark hacking on RoboArena since April. We have taken steps to prevent this in the future, and we have rolled back evals in accordance with these steps to retain the integrity of the benchmark. Read more about our changes here: robo-arena.github.io.
These choices are actually a bit orthogonal. Long vs short episodes mostly depends on your teleop success rate on long episodes --> if you frequently make mistakes in teleop, it's better to chunk a long task into shorter episodes. Re one task for everything vs subtask conditioning: the latter can potentially give you more generalization (see eg pi07 results), but requires training a high-level policy that picks subtasks at inference time, which can introduce one more source of errors. So I would suggest starting with simple one-task-for-all as baseline, and only after try the HL/LL design!
Excited to be in Rio this week to present RETAIN (w/ @zhiyuan_zhou_ , @ajwagenmaker, @KarlPertsch, and @svlevine) at #ICLR2026! 🇧🇷 Saturday 10:30 AM – 1:00 PM at P3-#1208. Project Page: retain.yajatyadav.com x.com/zhiyuan_zhou_/….
Do you ever find finetuning VLA overfits to the target task, to the point where generalist ability is lost and even minor deviations beyond the SFT data break the policy? We found an extremely simple solution: directly merge the base and finetuned policy in weight space 🤯 👇🧵
Happy to share some new results! π0.7 comes with memory, and algorithmic advances to pull out more performance and generalization from diverse training data! Check it out!
Our newest model, π0.7, has some interesting emergent capabilities: it can control a new robot to fold shirts for which we had no shirt folding data, figure out how to use an appliance with language-based coaching, and perform a wide range of dexterous tasks all in one model!
We developed an RL method for fine-tuning our models for precise tasks in just a few hours or even minutes. Instead of training the whole model, we add an “RL token” output to π-0.6, our latest model, which is used by a tiny actor and critic to learn quickly with RL.
late but: RETAIN will be presented at #ICLR2026 in Rio! The code is also out at github.com/yajatyadav/RET…, though all you really need is this one line
Do you ever find finetuning VLA overfits to the target task, to the point where generalist ability is lost and even minor deviations beyond the SFT data break the policy? We found an extremely simple solution: directly merge the base and finetuned policy in weight space 🤯 👇🧵
Jup, tho off the shelf VLMs today are often not well suited as HL policies for more complex tasks (many papers have shown this, they struggle with finegrained interaction understanding, failures etc) and robot fine tuned models so far need to be taught to remember explicitly. Agree tho that in the future this will hopefully be bridged
Many real-world tasks require memory to be successful. Yet, most robots don’t have any form of memory. Today, we are going to change that. We developed a system called MEM that introduces memory into VLAs on multiple scales
We’ve developed a memory system for our models that provides both short-term visual memory and long-term semantic memory. Our approach allows us to train robots to perform long and complex tasks, like cleaning up a kitchen or preparing a grilled cheese sandwich from scratch 👇
We equipped PI policies with memory! And taught our robots to do long-horizon real world tasks such as preparing the items for a recipe, cooking a grilled cheese and cleaning the kitchen!
We’ve developed a memory system for our models that provides both short-term visual memory and long-term semantic memory. Our approach allows us to train robots to perform long and complex tasks, like cleaning up a kitchen or preparing a grilled cheese sandwich from scratch 👇
We’ve developed a memory system for our models that provides both short-term visual memory and long-term semantic memory. Our approach allows us to train robots to perform long and complex tasks, like cleaning up a kitchen or preparing a grilled cheese sandwich from scratch 👇
This was one of the longest-running research projects at pi — adding memory to your models stretches all parts of your infra and needs innovation on the whole stack. The project started as @HomerWalke's internship project with @DannyDriess, but had lots of help from countless people at pi to get over the finish line. Special shoutout to @marceltornev who worked tirelessly to teach our models the long-horizon behaviors you saw in the videos above! For more details, check out our blog & paper: pi.website/research/memory
Finally, many prior works have reported that policies get *worse* on dexterous tasks when adding memory (because of spurious correlations, causal confusion etc). We find that by equipping pi06 with MEM and training it on our most diverse data mix, we can match pi06 performance on tasks that do not require memory (while clearly outperforming on memory tasks). This is IMO one of the biggest results here: we have a recipe for adding memory to VLAs without significant tradeoffs, both in terms of latency and performance!
This one has been a long time coming: today we’re introducing MEM, an approach for giving VLAs short-term and long-term memory. Memory is such an obvious capability, but adding it isn’t easy (most VLAs today are memory-less). A short thread on challenges, solutions, and the new capabilities MEM unlocks for us.
Use DROID data and PolaRiS sim evals to test your ideas on strong generalist policies! Congrats to the CoVer-VLA team!
🧵(6) DROID Eval CoVer-VLA achieves 14% gains in task progress and 9% in success rate on the challenging red-team PolaRiS benchmark. In the pan cleaning task, π₀.₅ shows incorrect intent, grasping the pan handle. In contrast, CoVer-VLA correctly uses sponge to scrub the pan.
Very exciting to see first steps of our models doing useful things in the world! Thanks to Ultra and Weave for being great partners in these deployments!
General-purpose AI models are behind some of the most exciting applications we now can't live without. We envision that an analogous “physical intelligence layer” built with models like π0.6 will similarly spur a new wave of applications for the physical world. We’ve recently
Ted Xiao @xiao_ted
71K Followers 798 Following Founding Member of Technical Staff @PrometheusInc. Previously Gemini, Robotics @GoogleDeepMind. Posts about frontier models, physical AGI, and scaling.
Sergey Levine @svlevine
131K Followers 143 Following Associate Professor at UC Berkeley Co-founder, Physical Intelligence
Danijar Hafner @danijarh
31K Followers 1K Following Building AI that autonomously understands and interacts with the world. Previous: @GoogleDeepMind @UCBerkeley @UofT
Jim Fan @DrJimFan
470K Followers 3K Following NVIDIA Director of Robotics & Distinguished Scientist. Co-Lead of GEAR lab. Solving Physical AGI, one motor at a time. Stanford Ph.D. OpenAI's 1st intern.
Chris Paxton @chris_j_paxton
33K Followers 4K Following Mostly posting about robots. currently AI @agilityrobotics prev embodied AI @AIatMeta, @NVIDIAAI. All views my own.
Jacky Liang @jackyliang42
7K Followers 793 Following Research Scientist @GoogleDeepMind working on foundation models for robotics. PhD @CMU_Robotics @iamlab_cmu
Eric Jang @ericjang11
135K Followers 4K Following
Xiaolong Wang @xiaolonw
22K Followers 1K Following Research Director, @Meta Superintelligence Labs Co-founder of ARI Associate Professor @UCSDJacobs Postdoc @berkeley_ai PhD @CMU_Robotics
Danfei Xu @danfei_xu
12K Followers 1K Following Faculty at Georgia Tech @ICatGT, Human Data for GR00T @nvidia | I like robots | all opinions are my own
Edward Johns @Ed__Johns
4K Followers 143 Following Associate Professor and Director of the Robot Learning Lab at Imperial College London.
Fei Xia @xf1280
12K Followers 838 Following Ex Research Scientist, TLM at @GoogleDeepMind, ✨♊, Gemini & Robotics, PhD from @StanfordAILab @StanfordSVL, previously @Tsinghua_Uni. #AGI through Embodiment
Dorsa Sadigh @DorsaSadigh
12K Followers 399 Following CS Faculty @Stanford, @StanfordAILab, @StanfordHAI Research scientist @GoogleDeepMind PhD and BS from @Berkeley_EECS
Danny Driess @DannyDriess
4K Followers 337 Following Research Scientist @physical_int. Formerly Google DeepMind
Markus Wulfmeier @m_wulfmeier
13K Followers 2K Following Chief Scientist #Nomagic - Prev. AI & Robotics @GoogleDeepMind - @UniOfOxford @UCBerkeley @ETH @MIT - European @ELLISforEurope
Shane Gu @shaneguML
50K Followers 2K Following Gemini/Omni Thinking, Senior Staff AE @GoogleDeepMind. 🇯🇵-born 🇨🇳🇨🇦. ex: Gemini 1.5-3, GPT-4 @OpenAI, Google Robotics (JP: @shanegJP). Personal opinions.
Abhishek Gupta @abhishekunique7
11K Followers 984 Following Assistant Professor at University of Washington. I like robots, and reinforcement learning. Previously: post-doc at MIT, PhD at Berkeley
Soumith Chintala @soumithchintala
309K Followers 1K Following Building new things @thinkymachines. Also dabble in robotics at NYU. Cofounded @PyTorch. AI is delicious when it is accessible and open-source.
Joseph Lim @JosephLim_AI
1K Followers 431 Following PI of the CLVR lab. Associate Professor @KAIST_ai. Prev: assist. prof. @USC, postdoc @Stanford, CS PhD @MIT, BA @UCBerkeley. AI, Robotics, Vision, ML.
Jesse Thomason @_jessethomason_
3K Followers 1K Following Incoming Associate Prof in the GaTech IC. Prev: assistant Prof at USC Viterbi CS. Leading the GLAMOR lab https://t.co/VQhcMiCG7c (he/him; 💖💜💙)
13604 @136048338q24378
0 Followers 117 Following
Yibin (Leon) Liu @yibinleonliu
21 Followers 399 Following Incoming PhD at TODO Research Intern at Robbyant (part of Ant Group)
Nishant @nishantchandna_
32 Followers 641 Following DARPA Triage Challenge | Robotics | IROS 2025 Workshop | UAVs
Nicky @Nickyrobotics
0 Followers 320 Following
Christina Gilbert @ChristinaG325
1K Followers 333 Following Visiting Partner @ycombinator, founder of @oneschema_co
Iris| Tech Job Search @ScrenaAI
27 Followers 105 Following The Hiring Insider Tech job search, minus the BS. Referrals, recruiter strategy, real interview prep. Building @ScrenaAI_ to make it easier.
Rishi @rishi_d_j
0 Followers 73 Following
Tim @tim404x
3K Followers 3K Following Co-founder @hubxyz | YC P26. Real-world data for frontier labs.
hty @hty45819296
1 Followers 62 Following
OffsidePlayer @offsider777
0 Followers 28 Following
Yuwei Niu @purshow04
449 Followers 1K Following Building on Multimodal Intelligence. Incoming MSc @PKU1898 ('26).
Ivan M @med_1v
1 Followers 4K Following
Xavier Puig @xavierpuigf
1K Followers 975 Following Founder at robotics stealth co (hiring!) | Prev Research Scientist at FAIR @AIatMeta on EmbodiedAI | PhD @MIT_CSAIL
Rasool Fakoor @rasoolfa
472 Followers 1K Following Building agents that reason, adapt, and act with RL and friends!
Dongjun Lee @dongjunlie
59 Followers 169 Following Second year Ph.D student @KAIST_AI | Prev: Research intern @microsoft
Solar Plexus @SolarPlexus007
2 Followers 341 Following
Kevin Lee @jeffren949
0 Followers 72 Following
yao @yao_o1o
1 Followers 42 Following
Kevin Wood | Robotics... @KWRoboticsAI
586 Followers 830 Following 10+ years in robotics, AI/Ml, computer vision and controls. 64k+ subscribers on YouTube. Subscribe to learn more! Link down below 👇
Andrew Lyubovsky @a_andlyu
291 Followers 345 Following Idk, testing something (automating robotic evals)
_ @_prmd_
146 Followers 5K Following Product Developer @ SAP, Product Security, AppSec, Data Privacy, GDPR. Tweets are personal. RTs aren't endorsements.
Amin Marashi @marashiai
6 Followers 223 Following
Yifan Chou @YifanChouYF
2 Followers 375 Following
Khang Trần Nhật @KhangTrnNh57285
0 Followers 18 Following
Xingxin HE @iamhxx
17 Followers 110 Following PhD student in Robotics @hkust ; Ex software engineer @autodesk
Eric Space @erictian830831
0 Followers 2K Following
Guy Freund @guyfreund1712
10 Followers 163 Following
Manish Shah @mhshah273
243 Followers 3K Following Influenced by The Secret. Take the life easy......
choohaay @choohaay
25 Followers 400 Following
Yuckey @YuckeyLab
400 Followers 922 Following
Ilia @IliaLarchenko
2K Followers 103 Following Data Science Director at Agoda | Ex-TOP-20 on Kaggle | Playing with robots for fun
Suhan N Shetty @Suhan_N_Shetty
28 Followers 250 Following A humble student of robotics. Believer in optimal control and reinforcement learning for robotics :)
Shengyang Wang @wangche35134045
1 Followers 16 Following
Nathan Saylor @NSaylor17086
1 Followers 7 Following Building Robotics & physical AI conference. July 16, San Francisco
autodidac @autodidaclzfm
0 Followers 5K Following
YEdawu66 @PFo32udr
4 Followers 727 Following
Animesh Garg @animesh_garg
35K Followers 1K Following Foundation Models for Generalizable Autonomy in Robotics. Reinforcement Learning. Faculty in AI Robotics @GeorgiaTech. Prev @nvidia @UCberkeley @stanford
Ted Xiao @xiao_ted
71K Followers 798 Following Founding Member of Technical Staff @PrometheusInc. Previously Gemini, Robotics @GoogleDeepMind. Posts about frontier models, physical AGI, and scaling.
Sergey Levine @svlevine
131K Followers 143 Following Associate Professor at UC Berkeley Co-founder, Physical Intelligence
Danijar Hafner @danijarh
31K Followers 1K Following Building AI that autonomously understands and interacts with the world. Previous: @GoogleDeepMind @UCBerkeley @UofT
Jim Fan @DrJimFan
470K Followers 3K Following NVIDIA Director of Robotics & Distinguished Scientist. Co-Lead of GEAR lab. Solving Physical AGI, one motor at a time. Stanford Ph.D. OpenAI's 1st intern.
Google DeepMind @GoogleDeepMind
1.5M Followers 278 Following The engine room of @Google. Building AI safely and responsibly to solve the world’s most complex problems. Join us: https://t.co/jUHQA27iBL
Yuke Zhu @yukez
22K Followers 475 Following Associate Professor @UTCompSci | Director @NVIDIAAI Co-Leading GEAR | CS PhD @Stanford | Building generalist robot autonomy in the wild | Opinions are my own
Lerrel Pinto @LerrelPinto
9K Followers 211 Following Making robots more dexterous, robust and general. Co-founder of Assured Robot Intelligence (ARI)
Eric Jang @ericjang11
135K Followers 4K Following
Xiaolong Wang @xiaolonw
22K Followers 1K Following Research Director, @Meta Superintelligence Labs Co-founder of ARI Associate Professor @UCSDJacobs Postdoc @berkeley_ai PhD @CMU_Robotics
Danfei Xu @danfei_xu
12K Followers 1K Following Faculty at Georgia Tech @ICatGT, Human Data for GR00T @nvidia | I like robots | all opinions are my own
Deepak Pathak @pathak2206
29K Followers 415 Following Co-Founder & CEO @SkildAI, Faculty @CarnegieMellon. PhD @UCBerkeley; BTech @IITKanpur I study topics in AI (robotics, machine learning & computer vision).
Edward Johns @Ed__Johns
4K Followers 143 Following Associate Professor and Director of the Robot Learning Lab at Imperial College London.
Yilun Du @du_yilun
12K Followers 329 Following Assistant Professor at Harvard @KempnerInst + CS. PhD @MIT_CSAIL, BS MIT. Generative Models, Compositionality, Embodied Agents, Robot Learning.
Fei Xia @xf1280
12K Followers 838 Following Ex Research Scientist, TLM at @GoogleDeepMind, ✨♊, Gemini & Robotics, PhD from @StanfordAILab @StanfordSVL, previously @Tsinghua_Uni. #AGI through Embodiment
Dorsa Sadigh @DorsaSadigh
12K Followers 399 Following CS Faculty @Stanford, @StanfordAILab, @StanfordHAI Research scientist @GoogleDeepMind PhD and BS from @Berkeley_EECS
Danny Driess @DannyDriess
4K Followers 337 Following Research Scientist @physical_int. Formerly Google DeepMind
Anastasis Germanidis @agermanidis
5K Followers 447 Following Simple ideas, pursued maximally. Co-Founder & Co-CEO @runwayml.
Nicolas Carion @alcinos26
2K Followers 249 Following In a stealth startup. Former research Scientist at Meta. Previously, post-Doc @nyuniversity, PhD student @FacebookAI Paris. RL, Transformers, Computer Vision.
XuDong Wang @XDWang101
1K Followers 784 Following MTS at Physical Intelligence @physical_int · Incoming Asst. Prof. @DukeU | PhD from @Berkeley_AI @UCBerkeley | Prev: @MetaAI (MSL/FAIR) · @GoogleDeepMind
Tobias Springenberg @jtspringenberg
85 Followers 14 Following
Evan Morikawa @E0M
17K Followers 2K Following VPE @Generalistai. GP ZSF. Prev: led eng at @openai. Launched & scaled ChatGPT, & APIs. @OlinCollege alum.
General Intuition @gen_intuition
9K Followers 18 Following We build foundation models and general agents for environments that require deep spatial and temporal reasoning.
Hang Zhao @zhaohang0124
5K Followers 1K Following Asst. Prof@ Tsinghua University, Co-founder @GalaxeaDynamics. Former Scientist@Waymo, MIT PhD’19. I research on multimodal learning, self-driving and robotics!
Weijia Shi @WeijiaShi2
10K Followers 2K Following PhD student @uwnlp | Prev @allen_ai @MetaAI @CS_UCLA
Matthew Mason @mastertoadster
323 Followers 48 Following Professor Emeritus, Carnegie Mellon Robotics Institute
Neuracore @Neuracore_AI
2K Followers 404 Following Robot Learning Infrastructure to Scale and Deploy Faster | Record asynchronous data, train VLA models, and ship solutions faster than competitors.
Andrea Bajcsy @andrea_bajcsy
2K Followers 230 Following Assistant Professor @SCSatCMU, @CMU_Robotics | PhD from @Berkeley_EECS | Robots, humans, learning, and safety
Jason Ma @JasonMa2020
59K Followers 984 Following Co-founder @DynaRobotics Prev: @GoogleDeepMind, @NVIDIAAI, @MetaAI, @Penn, @Harvard.
Astribot @Astribot_Inc
2K Followers 4 Following
Aishwarya Kamath @ashkamath20
8K Followers 633 Following Staff Research Scientist @GoogleDeepMind on the Gemini team. Vision Lead on Gemma 3 & Gemma 4. PhD at NYU with @ylecun Masters at UMass Amherst
Yuzhe Qin @QinYuzhe
5K Followers 736 Following Co-founder of @DexmateAI prev: Robotics Ph.D. @ UC San Diego
Zsolt Kira @zsoltkira
2K Followers 61 Following Associate Professor @ Georgia Tech artificial intelligence | machine learning | computer vision | robotics
Roei Herzig @roeiherzig
2K Followers 916 Following Postdoc @berkeley_ai. Researcher @IBMResearch. PhD @TelAvivUni. Working on Multimodal Robot Foundation Models and Structured Physical AI.
Project Aria @Meta @meta_aria
2K Followers 162 Following Project Aria is a research tool from Meta Reality Labs, to help build the future responsibly.
Joey Hejna @JoeyHejna
1K Followers 637 Following CS PhD @ Stanford, prev Berkeley AI, @GoogleDeepMind, @physical_int
Aleksei Petrenko @petrenko_ai
1K Followers 565 Following Research Scientist @ Apple Deep Learning & AI. CS PhD. I do fast open-source RL, see https://t.co/S6VEcLFia7
Nathan Ratliff @robot_trainer
2K Followers 411 Following Director of Robotic Systems @NVIDIA. Isaac Cortex, cobots, geometric methods; PhD CMU, research Max Planck, TTI-C, co-founder Lula Robotics, eng Google, Amazon
Yuki @y_m_asano
5K Followers 784 Following Full Professor and head of Fundamental AI (FunAI) Lab @funailab at University of Technology Nuremberg. previously @UvA_Amsterdam, @Oxford_VGG
Jiayi Pan @jiayi_pirate
14K Followers 2K Following Research | Prev @xAI @Berkeley_AI | Views Are My Own
Ross Taylor @rosstaylor90
12K Followers 1K Following Co-founder and CEO @GenReasoning. Previously lots of other things like: reasoning lead Meta AI, Llama 3/2, Galactica, Papers with Code.
Paril Jain @pariljain
9K Followers 1K Following Founder/CTO @thebotcompany . Built Tesla FSD. Former Head of AI+Planning @Tesla
David Luan @jluan
12K Followers 1K Following retired!! former cofounder of @adeptailabs, vp engineering @openai and @amazon, and @google LLMs lead. all about type II fun.
Archit Sharma @archit_sharma97
8K Followers 370 Following RL, post-training, reasoning research @GoogleDeepMind | co-created: Gemini Deep Think series, DPO | prev: @Stanford @Google Brain @IITKanpur @MILAMontreal
Charlie Snell @sea_snell
8K Followers 6K Following PhD student @berkeley_ai; research @cursor_ai; prev @GoogleDeepMind. My friend told me to tweet more. I stare at my computer a lot and make things
Ajay Jain @ajayj_
8K Followers 5K Following Co-founder @genmoai. Co-created denoising diffusion (DDPM), DreamFusion, Dream Fields. Ex Ph.D. @berkeley_ai, @googleai, @facebookai, @nvidiaai, @mit
Perceptron AI @perceptroninc
3K Followers 0 Following The layer of intelligence that bridges the physical and digital worlds. https://t.co/Mmo3044h89
Oleg Sinavski @SinavskiOleg
175 Followers 183 Following Just trying to solve AI like everyone else. LLMs, VLMs, RL and self-driving researcher, Python developer, sometimes a manager
Scott Reed @scott_e_reed
17K Followers 667 Following
Caitlin Kalinowski @kalinowski007
39K Followers 2K Following Builder-strategist in physical AI. Boards: Axon, Institute of Contemporary Art SF.
Yang Gao @gao_young
2K Followers 209 Following Co-founder @ Spirit AI Robotics. Assistant Professor @ Tsinghua University. Curious about vision and robotics
Russ Tedrake @RussTedrake
3K Followers 97 Following Professor at MIT, studying robotics. Founder of a stealth Physical AI startup.
Yuandong Tian @tydsh
44K Followers 939 Following Co-founder of @Recursive_SI. ex-Meta FAIR Director. ex-Google. Reasoning, Optimization and Understanding LLM. Novelist in spare time. PhD in @CMU_Robotics.
World Labs @theworldlabs
71K Followers 49 Following World Labs is a spatial intelligence company, building frontier models that can perceive, generate, and interact with the 3D world.
Stone Tao @Stone_Tao
7K Followers 1K Following Simulation/Robot Learning research @physical_int @UCSanDiego, previously @allen_ai @NVIDIA @sudo_robotics
Armen Aghajanyan @ArmenAgha
16K Followers 300 Following Co-founder & CEO @perceptroninc; ex-RS FAIR/MSFT
Murtaza Dalal @mihdalal
22K Followers 623 Following AI @ Tesla Optimus | Robotics/AI PhD @CMU_Robotics, EECS @UCBerkeley | Opinions entirely and unequivocally my ownTrends for United States
You might like
