https://dibbla.space/posts/matrices/Recent content in Many Matrices on dibbla.spaceHugoen-usby Dibbla with ❤Thu, 21 May 2026 00:19:20 -0700https://dibbla.space/posts/matrices/automated_science/Thu, 21 May 2026 00:19:20 -0700https://dibbla.space/posts/matrices/automated_science/<p>When I first write down the hook of this blog last month, I was thinking about something narrow: AI models have become very good at coding, and this is changing how machine learning research gets done. My thought was that the field might be shifting from a <em>workflow</em> problem to an <em>evaluation</em> problem — where the hard part is no longer doing the work, but judging which results are advancing human understanding.</p>https://dibbla.space/posts/matrices/iil_family/Sat, 21 Oct 2023 11:37:48 +0800https://dibbla.space/posts/matrices/iil_family/<p><em>Update Nov 2025:</em> I am surprised by PI integrates IIL method into the <a href="https://arxiv.org/abs/2511.14759">$\pi$*-0.6 model</a>, and I firmly believe that human / end-user will be integrated into the post-post-training of robotic foundation models in certain ways.</p> <hr> <p>In this blog, we discuss the imitation learning in an online fashion with human from DAgger, to HG-DAgger and more recent advances</p> <h2 id="dagger">DAgger</h2> <p>Dataset Aggregation (DAgger) is an imitation learning algorithm proposed in the AISTATS11 paper <em>A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning</em> by Stéphane Ross, Geoffrey J. Gordon and J. Andrew Bagnell. It is a simple yet effective algorithm that has been widely used in imitation learning, and as you can tell from the title, it&rsquo;s not related to human-in-the-loop RL.</p>