Calvin Luo

Excited to share with everyone an accessible, intuitive tutorial on diffusion models! If you're curious about the math behind diffusion models and how their different interpretations can be unified, please check it out! Stay tuned for a blog post soon! arxiv.org/abs/2208.11970

This is joint first-author work with Zilai Zeng @zilaizeng, collaborating with Mingxi Jia @mingxijiaa, advised by Yilun Du @du_yilun and Chen Sun @jesu9. Site: diffusion-supervision.github.io/silvr arXiv: arxiv.org/abs/2506.06658 Come visit us at #ICLR2026 on April 24th 10:30am, Poster #1907!

SILVR is robust to choices of filtering strategies; using VLMs to evaluate success can also enable successful improvement. Furthermore, under some cases we observe that SILVR can still sample-efficiently learn from self-collected experience even 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐟𝐢𝐥𝐭𝐞𝐫𝐢𝐧𝐠.

How can visual planning agents 𝙨𝙚𝙡𝙛-𝙞𝙢𝙥𝙧𝙤𝙫𝙚 from their own collected experience? We present 𝗦𝗜𝗟𝗩𝗥🩶, a framework that combines offline data with online experience for concurrent zero-shot generalization and sample-efficient self-improvement capabilities!#ICLR2026

We find that RL post-training can substantially improve BC policies without teaching them anything fundamentally new. So what is RL doing? In DICE-RL, it contracts a broad behavior prior toward high-value modes. (1/n) zhanyisun.github.io/dice.rl.2026/

For video generation in robotic applications, looking pretty is usually not enough. Robot manipulation requires understanding how visual observations and 3D geometry evolve over time under agent actions, with temporal coherence and geometric consistency across camera views. We study this challenge in our work (recently accepted by @iclr_conf ), 4D Video Generation for Robot Manipulation, which enforces multi-view 3D consistency via geometric supervision to generate spatio-temporally aligned videos.

What do AI assistants think about you, and how does this shape their answers? Because assistants are trained to optimize human feedback, how they model users drives issues like sycophancy, reward hacking, and bias. We provide data + methods to extract & steer these user models.

Skills are useful abstractions for transferring useful behavior across settings, but they often need subtle tweaks for new problems. How can we learn such flexible skills? Check out @vedant_gupta_16 's thread on our end-to-end discovery of these skills! 🤖

Excited to introduce DEPS (Discovery of GenEralizable Parameterized Skills) at #NeurIPS2025! DEPS learns interpretable parameterized skills that drastically improve generalisation to unseen tasks, especially in data-constrained settings and on out-of-distribution tasks. (1/n)

💡Can we trust synthetic data for statistical inference? We show that synthetic data (e.g. LLM simulations) can significantly improve the performance of inference tasks. The key intuition lies in the interactions between the moments of synthetic data and those of real data

Excited to introduce Dreamer 4, an agent that learns to solve complex control tasks entirely inside of its scalable world model! 🌎🤖 Dreamer 4 pushes the frontier of world model accuracy, speed, and learning complex tasks from offline datasets. co-led with @wilson1yan

1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).

A mental model I find useful: all data acquisition (web scrapes, synthetic data, RL rollouts, etc.) is really an exploration problem 🔍. This perspective has some interesting implications for where AI is heading. Wrote down some thoughts: yidingjiang.github.io/blog/post/expl…

Ever wish you could turn your video generator into a controllable physics simulator? We're thrilled to introduce Force Prompting! Animate any image with physical forces and get fine-grained control, without needing any physics simulator or 3D assets at inference. 🧵(1/n)

This is joint first-author work with Zilai Zeng @zilaizeng, and advised by Yilun Du @du_yilun and Chen Sun @jesu9. Project: diffusion-supervision.github.io/adapt2act Code: github.com/brown-palm/ada… arXiv: arxiv.org/abs/2504.15369 Check out our #ICLR2025 poster on April 25th at 3:30pm (Poster #418)!

We also discover that internet-scale pretraining can bridge the suboptimality gap through probabilistic adaptation and its inverse. Even for an in-domain model trained only on failed trajectories, successful video plans can be synthesized through adaptation even for novel tasks.

Internet-scale datasets of videos and natural language are a rich training source! But can they be used to facilitate novel downstream robotic behaviors across embodiments and environments? Our new #ICLR2025 paper, Adapt2Act, shows how.