A fun, cute, festive cartoon by cAPSLOCK! 🤖🐦🎃🍬

How can robots perform a wide variety of novel tasks from natural language? Execited to present Code as Policies - using language models to directly write robot policy code from language instructions. See paper, colabs, blog, and demos at code-as-policies.github.io long 🧵👇

My second post is up now (as promised, updates are very infrequent). "Paper Writing: A View from the Trenches" ofirnachum.github.io/posts/view-fro…

📢 Code, checkpoints and a colab for multitask (object detection, semantic segmentation, keypoint detection and captioning) Pix2Seq are now available! github.com/google-researc… github.com/google-researc… x.com/_akhaliq/statu…

An elegant paper that proposes a novel compression objective inspired by minimum description length to discover and learn skills with maximally common structure. Really exciting implications for hierarchical reinforcement learning - I #LOVE this work!

Join us at the WiML Un-Workshop breakout session on "Role of Mentorship and Networking"! Do not miss the chance to talk with leading researchers Samy Bengio, Susan Zhang Hugo Larochelle Sharon Y. Li Pablo Samuel Castro John Langford and others! #ICML2023 WiML

Sneak peek: - Paper: drive.google.com/file/d/1jILc6c… - Slides: drive.google.com/file/d/1rlee-a…

High-quality human feedback for RLHF is expensive 💰. AI feedback is emerging as a scalable alternative, but are we using AI feedback effectively? Not yet; RLAIF improves perf *only* when LLMs are SFT'd on a weak teacher. Simple SFT on a strong teacher can outperform RLAIF! 🧵->

Excited to share our latest preprint: “Self-Correcting Self-Consuming Loops for Generative Model Training”. It's a step towards generative AI models that can learn from the universe of data they generate!! 🤖(1/n)

Presenting LexC-Gen that generates data for extremely low-resource languages. 🤗 You only need 7B-LLMs and bilingual lexicons. 🔥 Our synthetic data are competitive with expert-translated data on sentiment and topic classification. Paper + Code: batsresearch.github.io/lexcgen/ [1/n]

Announcing Score Entropy Discrete Diffusion (SEDD) w/ Chenlin Meng Stefano Ermon. SEDD challenges the autoregressive language paradigm, beating GPT-2 on perplexity and quality! Arxiv: arxiv.org/abs/2310.16834 Code: github.com/louaaron/Score… Blog: aaronlou.com/blog/discrete-… 🧵1/n

Excited to present LAST: A new framework for discovering skills from demonstrations! By combining LLM initial segmentation+temporal variational inference+MDL, an agent is able to discover reusable skills from long-horizon task trajectories that helps downstream task learning.