<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Action Models for Robot Learning</title><description>New sections of the open-access textbook Action Models for Robot Learning, by Pavan Kumar Kandapagari, as they are drafted.</description><link>https://action-models-book.vercel.app/</link><language>en-us</language><item><title>§4.1 — Symbolic actions: STRIPS, PDDL, and action schemas</title><link>https://action-models-book.vercel.app/chapters/4/1/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/4/1/</guid><description>Section 4.1 of Action Models for Robot Learning. Prereqs: §1.2 (the three-slot anatomy), §1.3 (STRIPS situated in Era 1); high-school logic; the willingness to read a Lisp-flavored syntax for half an hour. ~2,000 target words.</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.x — Hands-on exercise + chapter references</title><link>https://action-models-book.vercel.app/chapters/3/x/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/x/</guid><description>Section 3.x of Action Models for Robot Learning. Prereqs: §3.1–§3.6; the SmallPolicy training loop from §3.3; the debugging checklist from §3.5; PyTorch installed; about two hours of laptop CPU time (no GPU required). ~2,000 target words.</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.6 — Summary</title><link>https://action-models-book.vercel.app/chapters/3/6/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/6/</guid><description>Section 3.6 of Action Models for Robot Learning. Prereqs: §3.1–§3.5; the SmallPolicy training loop, the three loss families, and the debugging checklist. ~2,000 target words.</description><pubDate>Tue, 26 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.5 — Debugging a model that will not train</title><link>https://action-models-book.vercel.app/chapters/3/5/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/5/</guid><description>Section 3.5 of Action Models for Robot Learning. Prereqs: §3.1 (gradients, chain rule, vanishing/exploding), §3.2 (cross-entropy, KL divergence), §3.3 (PyTorch training loop), §3.4 (three loss families and their diagnostic signatures). ~2,000 target words.</description><pubDate>Mon, 25 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.4 — Three loss families: supervised, RL, self-supervised</title><link>https://action-models-book.vercel.app/chapters/3/4/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/4/</guid><description>Section 3.4 of Action Models for Robot Learning. Prereqs: §3.1 (gradients, chain rule), §3.2 (cross-entropy, KL divergence), §3.3 (PyTorch training loop). ~2,000 target words.</description><pubDate>Sun, 24 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.3 — A 50-line PyTorch training loop, annotated</title><link>https://action-models-book.vercel.app/chapters/3/3/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/3/</guid><description>Section 3.3 of Action Models for Robot Learning. Prereqs: §3.1 (vectors, gradients, chain rule), §3.2 (cross-entropy, KL divergence); a working Python/PyTorch installation. ~2,000 target words.</description><pubDate>Sat, 23 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.2 — Random variables, expectations, KL divergence</title><link>https://action-models-book.vercel.app/chapters/3/2/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/2/</guid><description>Section 3.2 of Action Models for Robot Learning. Prereqs: §3.1 (vectors, gradients, chain rule); familiarity with summation notation and basic set theory. ~2,000 target words.</description><pubDate>Fri, 22 May 2026 00:00:00 GMT</pubDate></item><item><title>§3.1 — Vectors, matrices, gradients, and why the chain rule rules robotics</title><link>https://action-models-book.vercel.app/chapters/3/1/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/3/1/</guid><description>Section 3.1 of Action Models for Robot Learning. Prereqs: Chapter 1 (vocabulary of action models), Chapter 2 (a running OpenVLA loop); high-school calculus and a passing acquaintance with linear algebra. ~2,000 target words.</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.x — Hands-on exercise + chapter references</title><link>https://action-models-book.vercel.app/chapters/2/x/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/x/</guid><description>Section 2.x of Action Models for Robot Learning. Prereqs: §2.1–§2.6; a working OpenVLA-on-LIBERO loop; the three silent failures from §2.4; willingness to break the loop on purpose. ~2,000 target words.</description><pubDate>Wed, 20 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.6 — Summary</title><link>https://action-models-book.vercel.app/chapters/2/6/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/6/</guid><description>Section 2.6 of Action Models for Robot Learning. Prereqs: §2.1–§2.5; a working OpenVLA-on-LIBERO loop, the three named silent failures from §2.4, and the chapter map from §2.5. ~1,900 target words.</description><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.5 — What is left for the rest of the book</title><link>https://action-models-book.vercel.app/chapters/2/5/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/5/</guid><description>Section 2.5 of Action Models for Robot Learning. Prereqs: §2.1 (the four hidden commitments of a VLA), §2.3 (the inference loop), §2.4 (the failure taxonomy); a working mental picture of OpenVLA driving LIBERO; willingness to treat the rest of the book as commentary on this one demo. ~2,000 target words.</description><pubDate>Tue, 19 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.4 — When it works and when it does not</title><link>https://action-models-book.vercel.app/chapters/2/4/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/4/</guid><description>Section 2.4 of Action Models for Robot Learning. Prereqs: §2.3 (the inference loop with `predict_action`, `unnorm_key`, and the `agentview_image` flip); willingness to break the loop in three specific ways and read the resulting behavior; LIBERO suite vocabulary (`libero_object`, `libero_spatial`, `libero_goal`, `libero_10`). ~2,000 target words.</description><pubDate>Sat, 16 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.3 — Walking through the inference loop one line at a time</title><link>https://action-models-book.vercel.app/chapters/2/3/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/3/</guid><description>Section 2.3 of Action Models for Robot Learning. Prereqs: §2.1 (the end-to-end picture), §2.2 (working GPU, pinned weights, LIBERO smoke test); comfort with PyTorch tensors and Hugging Face Auto* APIs. ~2,000 target words.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.2 — Setting up the environment (OpenVLA weights, LIBERO simulator)</title><link>https://action-models-book.vercel.app/chapters/2/2/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/2/</guid><description>Section 2.2 of Action Models for Robot Learning. Prereqs: §2.1 (what the running model looks like end-to-end); Python, pip, comfort reading tracebacks; an NVIDIA GPU. ~2,000 target words.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>§2.1 — What we are going to build, and what is hidden inside</title><link>https://action-models-book.vercel.app/chapters/2/1/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/2/1/</guid><description>Section 2.1 of Action Models for Robot Learning. Prereqs: §1.2 (the three-slot anatomy), §1.4 (the foundation/VLA family); basic Python. ~2,000 target words.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.x — Hands-on exercise + chapter references</title><link>https://action-models-book.vercel.app/chapters/1/x/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/x/</guid><description>Section 1.x of Action Models for Robot Learning. Prereqs: §1.1–§1.6. ~2,000 target words.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.6 — Summary</title><link>https://action-models-book.vercel.app/chapters/1/6/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/6/</guid><description>Section 1.6 of Action Models for Robot Learning. Prereqs: §1.1–§1.5. ~1,800 target words.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.5 — What you will and will not find in this book</title><link>https://action-models-book.vercel.app/chapters/1/5/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/5/</guid><description>Section 1.5 of Action Models for Robot Learning. Prereqs: §1.1–§1.4 (the action problem, anatomy, history, four families). ~2,000 target words.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.4 — The four families of action models</title><link>https://action-models-book.vercel.app/chapters/1/4/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/4/</guid><description>Section 1.4 of Action Models for Robot Learning. Prereqs: §1.2 (the three-slot anatomy), §1.3 (the six-era history). ~2,000 target words.</description><pubDate>Thu, 14 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.3 — A short history, from STRIPS to π0</title><link>https://action-models-book.vercel.app/chapters/1/3/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/3/</guid><description>Section 1.3 of Action Models for Robot Learning. Prereqs: §1.1 (why action is hard), §1.2 (the three-slot anatomy). ~2,200 target words.</description><pubDate>Wed, 13 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.2 — Anatomy of an action model: inputs, outputs, training signal</title><link>https://action-models-book.vercel.app/chapters/1/2/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/2/</guid><description>Section 1.2 of Action Models for Robot Learning. Prereqs: §1.1 (the four reasons action is hard). ~2,000 target words.</description><pubDate>Wed, 13 May 2026 00:00:00 GMT</pubDate></item><item><title>§1.1 — Why &quot;action&quot; is the hard part of robotics</title><link>https://action-models-book.vercel.app/chapters/1/1/</link><guid isPermaLink="true">https://action-models-book.vercel.app/chapters/1/1/</guid><description>Section 1.1 of Action Models for Robot Learning. Prereqs: linear algebra, basic Python. ~2,000 target words.</description><pubDate>Wed, 13 May 2026 00:00:00 GMT</pubDate></item></channel></rss>