AI Radar
Support
LiveUpdated 2026-09-19 18:00 UTC

justrl

justrl — [ICLR 2026 Blogpost Track Poster] JustRL: Scaling a 1.5B LLM with a Simple RL Recipe -…

justrl is [ICLR 2026 Blogpost Track Poster] JustRL: Scaling a 1.5B LLM with a Simple RL Recipe - thunlp/JustRL. It is ranked #784 on the AI Radar, in AI research, first seen 8 days ago and shared in 2 posts (10.2k views).

Visit github.com

What people said about justrl on X

JustRL Set a New 1.5B Reasoning SOTA With Plain GRPO Two 1.5B models scored 54.87% and 64.32% across nine math benchmarks. One beat a nine-stage system using only half its training-token budget. The surprising part: adding more “optimization” made the results worse. As JustRL II extends this research line toward 128K…

@ZhihuFrontier, 8 days ago · 37 likes · see the post

JustRL II Takes AIME 2025 From 61% to 81% by Giving Long-CoT RL a Critic The original post showed that a remarkably simple RL recipe could push 1.5B reasoning models to SOTA. This analysis is its direct follow-up, tackling a harder problem: how to train small models for much longer chains of thought. Zhihu…

@ZhihuFrontier, 8 days ago · 30 likes · see the post

Alternatives to justrl

justrl in numbers

FAQ

What is justrl?

[ICLR 2026 Blogpost Track Poster] JustRL: Scaling a 1.5B LLM with a Simple RL Recipe - thunlp/JustRL It was first shared on X 8 days ago and is ranked #784 on the AI Radar.

Is justrl free?

It is open source.

Who shared justrl?

1 account on X, including @ZhihuFrontier, in 2 posts totalling 10.2k views.

Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 27.1k posts from 4.8k X accounts over the last 14 days, 1.2k tools, 19 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-19 18:00 UTC. Full method.