
AI/ML leader and independent robotics researcher. NeurIPS BEHAVIOR '25 winner · ICRA LeHome '26: 1st sim / 2nd real · Kaggle top-20 · Director @Agoda
Bangkok, Thailand
Joined September 2019
- Tweets481
- Following113
- Followers2.2K
- Likes225
Pinned Tweet
I placed 🥈 2nd in the LeHome Challenge (at @ieee_ras_icra 2026) earlier this month, and before that I was 🥇 1st of 62 teams in the simulation round. Now I am sharing my solution — with a detailed logic walkthrough and open-source code.
The task was to teach a cheap two-armed robot to fold different garments in simulation and on a real robot.
I trained a VLA policy with an RL loop to make it work. Let's break it down 👇
Thank you RoboEval team for organizing it!
🏆 Announcing the RoboEval Challenge winners at #CoRL2026!
🥇 @IliaLarchenko — 0.988
🥈 Metric AI Lab (@daniel7an, @HrantDavtian) — 0.982
🥉 LocalModels — 0.978
Across 8 bimanual manipulation tasks, we evaluated not just task success, but also safety, efficiency, and coordination.
Huge congratulations to our winners! 🎉🤖
1
7
395
Our team finished 9th in the 2026 BEHAVIOR Challenge and received the Open Source award 🏆
Huge thanks to my teammates @zaringleb, @atindranair, @keertimanney, @vkleban, @mirkokiefer and @ayoda1234 !
As you may guess from the prize name, we've open-sourced our solution: code, weights and a detailed tech report. Here is a summary 👇
Talking about other teams, it was super pleasant to see that many of them built on our last year's solution, including the top 2:
- 1st place used our 2025 solution as the base for their VLA policies: behavior1k-champion.com
- 2nd place adopted our stage scheme, task/stage tokens, delta actions and gripper rules: arxiv.org/abs/2611.13818
- many other teams forked our code or started from our checkpoint
This is exactly why we open-source everything.
1
3
94
Learn more in our:
📝 Blog post: robot-learning-collective.gi…
📄 Tech report: arxiv.org/abs/2611.14432
🔧 Code: github.com/Robot-Learning-Co…
🤗 Weights: huggingface.co/IliaLarchenko…
Unfortunately, I couldn't make it to #CoRL2026 because all tickets were sold out. But I still have a couple of things to present remotely! 🤖
🏆 RoboEval Challenge: I competed solo and finished among the top teams (official results aren't public yet). My solution will be presented twice:
Mon, Nov 9, 15:30-16:00 at the @bitrobot booth (#408)
Thu, Nov 12, 11:40-12:00 at the corl2026-realworld-robot-lea… workshop (JW Marriott, Room 212)
🏆 BEHAVIOR Challenge 2026: Our team took 9th place this year! We'll present our solution on Wed, Nov 11, at around 10:30 AM at the @bitrobot booth (#408). @keertimanney will represent our team in person!
If you're attending CoRL, feel free to stop by! Would love to hear your thoughts and questions! 🚀
I tagged the wrong @BitRobotNetwork
xitter.cf/sanatem/status/2119906…
Lots of lightning talks at our @BitRobotNetwork booth today!
Big kudos to @DJiafei, @YiruHelenWang and @hq_fang for the great presentations!
1
161
We took 9th place in the BEHAVIOR Challenge 2026! 🎉
Last year we won the competition. This year we significantly improved our previous solution, explored two different approaches (our Pi0.5-based VLA and Astra), and achieved much higher results!
I'm really proud of our progress, but the competition was much tougher this year. The top teams did even better, which is very impressive! I'm looking forward to their presentations at CoRL next week and learning more about their approaches.
As usual, we'll share our findings and open-source our solution so others can build on it.
It was also a bigger team effort this year! Huge thanks to @zaringleb, @atindranair, @keertimanney, @vkleban, @mirkokiefer and @ayoda1234. It was great working with you all!
We were the only independent team in the top 10, and it's interesting to see how the competition is evolving. More and more well-funded companies and research labs are participating, and competing at this level seems to be getting increasingly expensive.
Still, I strongly believe that competitions are much better benchmarks for robotics than polished lab demos. They provide a much clearer picture of the actual progress in the field.
Great initiative! Nice to see more real life evals.
Introducing RoboColosseum
An open robot eval initiative by @huggingface @BitRobotNetwork @eaic_sg @DJiafei
Access:
- 100+ robots (@UnitreeRobotics G1s, @LeRobotHF SO-101s, YAMs, Franka and more)
- 140+ in-the-wild scenes
- Open rollout data
- 10,000 FREE eval hrs
Thread ↓










