Détails

Source

TAO Daily

Publication

27 août à 10h17

Lien direct

https://taodaily.io/openrobotos-first-month-is-already-forcing-a-new-robotics-benchmark/

Contenu source (brut)

<p class="wp-block-paragraph">OpenRoboto may have entered Bittensor only recently, but its first month has already produced a result that changed the subnet’s roadmap.</p> <p class="wp-block-paragraph">In a live conversation with <a href="https://www.youtube.com/live/eETePHcENvs?si=jVZVxdrRWBy0v0J0">Shizzy</a>, OpenRoboto co-founder Cameron said miners have taken the subnet’s starting π0.5 model from roughly <strong>50% to 86.9% success on the LIBERO-Pro benchmark in about one month</strong>. The team had initially estimated that reaching this level could take <strong>three to five months</strong>.</p> <p class="wp-block-paragraph">That acceleration has created an unusual problem: <strong>the benchmark is becoming too easy.</strong></p> <p class="wp-block-paragraph">Cameron said the team is now developing its own benchmark with changing tasks to make overfitting harder and create more headroom for continued improvement.</p> <p class="wp-block-paragraph">More importantly, OpenRoboto plans to introduce <strong>physical robots into validation</strong>, moving beyond simulation to test whether a miner’s improvements can transfer to hardware.</p> <p class="wp-block-paragraph">This is significant because the <strong>sim-to-real gap remains <a href="https://www.nature.com/articles/s42256-022-00573-6?utm_source=chatgpt.com">a major obstacle</a> in robotics</strong>. Simulators mostly simplify physical reality, and research has shown they can perform substantially worse when transferred to real hardware.</p> <p class="wp-block-paragraph">In <a href="https://www.nature.com/articles/s41467-024-50131-4?utm_source=chatgpt.com">one study</a>, controllers that ranked highly in simulation experienced a <strong>57% drop in performance on physical robots</strong>, illustrating why simulation scores alone are not enough to establish real-world capability.</p> <p class="wp-block-paragraph">The subnet is also building out the data side of this loop. OpenRoboto <a href="https://taodaily.io/axis-robotics-partners-with-openroboto-to-advance-decentralized-robotics-ai/">recently partnered</a> with Axis Robotics, which is supplying <strong>more than 3 million multimodal robotics trajectories</strong> to its Open Data Pool and supporting its benchmarking infrastructure.</p> <p class="wp-block-paragraph">There is an interesting research angle behind that partnership as well. An AXIS research paper reports that continual pretraining on its data improved π0.5&#8217;s overall success rate by <strong>5.8%</strong>, with particularly strong gains under layout, sensor-noise, and camera perturbations.</p> <p class="wp-block-paragraph">Cameron also revealed that OpenRoboto is thinking beyond research benchmarks. When asked what success would look like a year from now, he pointed to two outcomes: materially better robotics foundation models <strong>validated on real robots</strong>, and commercial relationships with factories and data providers that generate actual revenue.</p> <p class="wp-block-paragraph">The takehome from the conversation is therefore not simply that OpenRoboto is training robotics models on Bittensor. <strong>Its competitive training loop is improving models quickly enough that the team is having to accelerate the next layer of the system for eventual physical-robot validation.</strong></p> <figure class="wp-block-image size-large"><img fetchpriority="high" decoding="async" width="1024" height="729" src="https://taodaily.io/wp-content/uploads/2026/08/image-183-1024x729.png" alt="" class="wp-image-24248" srcset="https://taodaily.io/wp-content/uploads/2026/08/image-183-1024x729.png 1024w, https://taodaily.io/wp-content/uploads/2026/08/image-183-300x214.png 300w, https://taodaily.io/wp-content/uploads/2026/08/image-183-767x546.png 767w, https://taodaily.io/wp-content/uploads/2026/08/image-183.png 1180w" sizes="(max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption">Roadmap on the <a href="https://www.openroboto.ai/">subnet&#8217;s website</a></figcaption></figure> <p class="wp-block-paragraph">OpenRoboto&#8217;s public roadmap now explicitly lists its own benchmark and a real-robot gate as the next stages.</p> <p class="wp-block-paragraph">Watch the full conversation on Shizzy Unchained:</p> <figure class="wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio"><div class="wp-block-embed__wrapper"> <iframe title="Building the Robot Brain on Bittensor | Cameron, OpenRoboto" width="500" height="281" src="https://www.youtube.com/embed/eETePHcENvs?feature=oembed" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe> </div></figure>