Show HN: RULER – Easily apply RL to any agent

Hacker News (score: 11)
Found: July 11, 2025
ID: 249

Description

Other
Show HN: RULER – Easily apply RL to any agent Hey HN, Kyle here, one of the co-founders of OpenPipe.

Reinforcement learning is one of the best techniques for making agents more reliable, and has been widely adopted by frontier labs. However, adoption in the outside community has been slow because it's so hard to implement.

One of the biggest challenges when adapting RL to a new task is the need for a task-specific "reward function" (way of measuring success). This is often difficult to define, and requires either high-quality labeled data and/or significant domain expertise to generate.

RULER is a drop-in reward function that works across different tasks without any of that complexity.

It works by showing N trajectories to an LLM judge and asking it to rank them relative to each other. This sidesteps the calibration issues that plague most LLM-as-judge approaches. Combined with GRPO (which only cares about relative scores within groups), it just works (surprisingly well!).

We have a full writeup on the blog, including results on 4 production tasks. On all 4 tasks, small Qwen 2.5 models trained with RULER+GRPO beat the best prompted frontier model, despite being significantly smaller and cheaper to run. Surprisingly, they even beat models trained with hand-crafted reward functions on 3/4 tasks! https://openpipe.ai/blog/ruler

Repo: https://github.com/OpenPipe/ART

More from Hacker

Show HN: Smart Switcher - data driven tool to improve the window switching

Show HN: Smart Switcher - data driven tool to improve the window switching Hello, my name is Andrew. I&#x27;m an indie developer and I&#x27;m excited to release Smart Switcher for Windows 10&#x2F;11. I&#x27;m looking for feedback on the overall project and the application itself.<p>I built this because I couldn&#x27;t find a window switching&#x2F;management solution that worked for me. I tried all kinds of different solutions, virtual desktop extensions, obscure GUI window managers, you name it. Overtime I realized I wanted something that prioritizes one window at a time, is keyboard driven with has minimal if no GUI elements. I figured this part out, but knew something was missing. I had my eureka moment when I realized I could combine my switching method with a prediction algorithm. This led to the creation of Smart Switcher.<p>Smart Switcher is a data driven window switcher aimed at improving the overall window switching experience. It logs data on your windows switching, then a prediction algorithm analyzes this data and uses it to predict which window you would want to switch to next. When you need to switch windows, you press the switch shortcut to switch to the next predicted window. If this isn&#x27;t the window you wanted, press the override shortcut to switch to the next most likely window. You can press the override shortcut as many times as needed until you arrive at your desired window.<p>It’s a paid app with a demo and trial version. There is a introductory discount and some additional discount tiers for early adopters.<p>Any feedback is appreciated! Thanks!

Show HN: Virby, a vfkit-based Linux builder for Nix-Darwin

Show HN: Virby, a vfkit-based Linux builder for Nix-Darwin Virby is a module for nix-darwin that configures a lightweight linux VM as a remote build machine for nix, allowing linux packages to be built on macOS.

No other tools from this source yet.