🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

You can now train agents to communicate in your browser using DIAL! Thanks to Moksh Jain for making Minqi Jiang's implementation more accessible and available 🙏

First blog post, where I discuss disentangled representations and basics of VAEs. The post is accompanied by a notebook which helps you get started with disentanglement_lib from Google AI Intelligent Systems for @disen_challenge mj10.github.io/blog/2019/07/1…

Pretty cool to see this blog post + notebook that illustrates how to use disentanglement_lib (github.com/google-researc…) for the NeurIPS 2019 competition on learning disentangled representations (aicrowd.com/challenges/neu…)!

Also, we recently posted this work arxiv.org/abs/1908.00261 on the theory of policy gradients for reinforcement learning! A long time in the works, this paper finally gets a handle on function approximation with policy gradient methods.

Happy to share our great collaborative effort “RL: Generic reinforcement learning codebase in TensorFlow” - joss.theoj.org/papers/10.2110…. Bryan M. Li @AidanNGomez Alexander Imani Cowen-Rivers. David Tao Sid Nitarshan Rajkumar Hariharan Sezhiyan Sicong (Sheldon) Huang

Excited to attend my first NeurIPS Conference this year, thanks to travel grants from NeurIPS Conference and MineRL Project! Looking forward to presenting my work and learning from the vast ocean of amazing work and talented people present there!

FOR.ai NeurIPS Dinner 2019 ❤️

Looking forward to discussing our work at #ICML2020 poster session!

Amazing blog post surveying adversarial attacks for Deep RL. Great work!

C.R. Rao, the doyen of Statistics, who populated his field (and thereby #AI/#ML) with such pleasing Rao-isms as Rao-Blackwell theorem, Cramer-Rao lower bound--turns 100 today. 🙏(en.wikipedia.org/wiki/C._R._Rao ) thehindu.com/opinion/open-p…

Excited to share our method "DROCC: Deep Robust One-Class Classification" for solving anomaly detection problem in any domain (proceedings.icml.cc/book/2020/file…) Work with awesome collaborators: Sachin Goyal , Moksh Jain , Aditi Raghunathan, Harsha Simhadri. 1/n

The Microsoft Turing team & Microsoft Research have collaborated to create the Turing Universal Representation Language Model, which now leads the XTREME public leaderboard for cross-lingual transfer learning. Learn how T-ULRv2 sets new state of the art: aka.ms/AAa0r9z

Excited to share this post on the M2D2 blog accompanying our recent paper arxiv.org/abs/2302.00615, which outlines the potential of GFlowNets to accelerate scientific discovery.

New work on scalable off-policy (MaxEnt) RL for LLM fine-tuning led by Brian Bartoldson We learned a lot about effective RL training of LLMs, so checkout the thread and paper for more details! Excited about more work in this direction!

New work on learning abstractions for amortized samplers led by Oussama Boussif and other brilliant collaborators! Even with simple tokenization schemes like BPE we can discover meaningful abstractions! Check out the paper and come chat with us ICLR 2026

Bringing hierarchical structure for search-based inference in LLM. Sharing our recent work "Search-Based Correction of Reasoning Chains for LMs". Work done with Jean-Pierre Falet, Oliver Richard, Xiaoyin Chen , Moksh Jain , Sungjin Ahn , Sungsoo Ahn, Yoshua Bengio