Linux/PyTorch Foundation Workshop w. Meta, HuggingFace, and Unsloth: Agentic RL and Environments | AgentX-AgentBeats Competition
Join us for an exclusive AgentX–AgentBeats Competition workshop with Meta, HuggingFace, and Unsloth, in partnership with the Linux/PyTorch Foundation!
📅 Wed, Feb 19, 2026 | 10:00 AM PT
Learn how to go from “zero to hero” with agentic reinforcement learning and environment-driven evaluation using OpenEnv—with updated notebooks that reflect the latest features and best practices.
You’ll see how to run and iterate quickly, how to align your work with modern inference/runtime flows (including Llama + vLLM notebook patterns), and how to package and submit through the Hugging Face Hub for competitions or structured evaluation.
What to Expect:
A guided walkthrough of the latest OpenEnv features and notebook workflows
A practical “submission notebook” example and submission-ready packaging approach
A step-by-step Hub push + submission process (best practices and common pitfalls)
A closing showcase of related ecosystem work: torchtitan, torchforge, and more
Live Q&A and an on-ramp to the OpenEnv Discord for ongoing support
Don’t miss this chance to learn from three of the leading AI research organizations and advance your AgentBeats OpenEnv Challenge progress.
The OpenEnv Challenge is sponsored by the PyTorch team at Meta, Hugging Face, and Unsloth, with a $10,000 prize pool in Hugging Face credits (with more prizes/resources to be announced), plus an invitation to publish on the PyTorch.org blog about what you built!
What You’ll Build: The goal of the challenge is to create a real-world RL environment and then attempt to solve it using your favorite LLM.
To enter, participants will:
Add their environment to the Hugging Face Hub
Publish training notebooks/scripts on GitHub
Write a submission blog on Hugging Face
OpenEnv Challenge Submission Deadline: March 30, 2026
Evaluation Criteria
The main evaluation criteria will be based on the submission blog. We will have a judging panel that will grade submissions based on:
Creative and Robust use of OpenEnv
Technical Excellence
Story-telling
Open Source Demo
Green Agent wrapper for the environment
We can't wait to see what you've built!
And don’t forget to sign up for the AgentX-AgentBeats Competition, hosted by Berkeley RDI, where over $1M in prizes and resources are up for grabs! This two-phase competition invites participants to: First build or enhance benchmarks for agentic AI (Phase 1), and then develop AI agents that excel on them (Phase 2). Together, these phases aim to advance the field by creating high-quality, broad-coverage, and realistic agent evaluations as shared public goods—building a unified, community-driven ecosystem for agent evaluation benchmarks that are compatible, standardized, reproducible, collaborative, and discoverable.
✅ Sign Up for the Competition:
Participant Sign‑Up Form – Each participant must register individually.
Team Sign‑Up Form – After completing your individual registration, please submit a team registration form. All participants must complete this step — even if competing solo (a team may consist of just one participant).
Let’s shape the future of Agentic AI — together!