Ilia

@IliaLarchenko • 2,160 subscribers

AI/ML leader and independent robotics researcher. NeurIPS BEHAVIOR '25 winner · ICRA LeHome '26: 1st sim / 2nd real · Kaggle top-20 · Director @Agoda

Shorts

I just released a first video deep dive into my LeHome Challenge solution (🥇 sim round, 🥈 real-robot final at ICRA 2026). I already shared what I did and the video series is about why: the reasoning, the alternatives I rejected, and the intuition behind each choice. I split the whole video explanation into 3 parts: Part 1 is about RL for VLAs 👇

I just released a first video deep dive into my LeHome Challenge solution (🥇 sim round, 🥈 real-robot final at ICRA 2026). I already shared what I did and the video series is about why: the reasoning, the alternatives I rejected, and the intuition behind each choice. I split the whole video explanation into 3 parts: Part 1 is about RL for VLAs 👇

32,748 views

The final Part 3 of my LeHome Challenge deep dive is out (🥈 real-robot final at ICRA 2026). This one is about the real world: the sim-to-real gap, my home data-collection rig, domain randomization, and the human-in-the-loop DAgger recovery data collection.

The final Part 3 of my LeHome Challenge deep dive is out (🥈 real-robot final at ICRA 2026). This one is about the real world: the sim-to-real gap, my home data-collection rig, domain randomization, and the human-in-the-loop DAgger recovery data collection.

10,128 views

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

I placed 🥈 2nd in the LeHome Challenge (at IEEE 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 👇

I placed 🥈 2nd in the LeHome Challenge (at IEEE 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 👇

29,735 views • 4 months ago

Pi-05 after just 30 minutes of fine-tuning. Feels like the best VLA I tried so far

Pi-05 after just 30 minutes of fine-tuning. Feels like the best VLA I tried so far

49,006 views • 1 year ago

A couple of days ago, I presented our 1st place solution for the 2025 BEHAVIOR Challenge at NeurIPS Conference . Now, we've open-sourced our solution: code, model weights, and a detailed tech report. Let me unpack what we did 👇

A couple of days ago, I presented our 1st place solution for the 2025 BEHAVIOR Challenge at NeurIPS Conference . Now, we've open-sourced our solution: code, model weights, and a detailed tech report. Let me unpack what we did 👇

28,213 views • 11 months ago

1/ In my recent robotics experiments, I started exploring state-of-the-art (SOTA) behavioral cloning (BC) models. Initially, this was just a way to familiarize myself with the LeRobot library and understand the existing approaches. However unexpectedly I developed a very simple model that outperformed SOTA on two benchmark environments. This thread covers my findings, the model, and why I believe some current approaches might be overcomplicating things.

1/ In my recent robotics experiments, I started exploring state-of-the-art (SOTA) behavioral cloning (BC) models. Initially, this was just a way to familiarize myself with the LeRobot library and understand the existing approaches. However unexpectedly I developed a very simple model that outperformed SOTA on two benchmark environments. This thread covers my findings, the model, and why I believe some current approaches might be overcomplicating things.

39,408 views • 1 year ago

I ran a quick test of Gemini Robotics ER 1.5 controlling my LeKiwi using the MCP I developed before. No extra code required - just specified a new model. Unfortunately, it doesn't work well, even though I had expected it to be the best, as it is the first major VLM model trained specifically for robotics applications. The video below is a relatively good attempt. It does understand what is going on and generates a reasonable action plan, but it is not doing well with the low-level control. I guess it was not trained for this and was used together with the VLA, so it is trained to generate more high-level commands like "grab this object and put it there" rather than doing low-level control specifying every robot movement step by step. Did anybody try to use it in a similar setup? What is your experience?

I ran a quick test of Gemini Robotics ER 1.5 controlling my LeKiwi using the MCP I developed before. No extra code required - just specified a new model. Unfortunately, it doesn't work well, even though I had expected it to be the best, as it is the first major VLM model trained specifically for robotics applications. The video below is a relatively good attempt. It does understand what is going on and generates a reasonable action plan, but it is not doing well with the low-level control. I guess it was not trained for this and was used together with the VLA, so it is trained to generate more high-level commands like "grab this object and put it there" rather than doing low-level control specifying every robot movement step by step. Did anybody try to use it in a similar setup? What is your experience?

15,658 views • 1 year ago

No more content to load