Trending

AI Agents Break Rules Under Everyday Pressure

RL is more information inefficient than you thought

CS234: Reinforcement Learning Winter 2025

Agent Design Is Still Hard

SIMA 2: An Agent That Plays, Reasons, and Learns with You in Virtual 3D Worlds

Jasmine: A Simple, Performant and Scalable Jax-Based World Modeling Codebase

Show HN: Cancer diagnosis makes for an interesting RL environment for LLMs

Cursor Composer: Building a fast frontier model with RL

Agent Lightning: Train agents with RL (no code changes needed)

OpenAI at QCon AI NYC: Fine Tuning the Enterprise

At QCon AI NYC 2025, Will Hang from OpenAI unveiled Agent RFT—a cutting-edge reinforcement fine-tuning approach for tool-using agents. By optimizing prompts and tasks before model adjustments, Hang showcased effective strategies to enhance decision-making and efficiency, emphasizing a balanced grading system. The session revealed a future where smarter agents reduce latency and improve outcomes. By Andrew Hoblitzell

Researchers show a robot learning 1,000 tasks in 24 hours

How do you keep agents aligned when tasks get messy?

You can now do FP8 reinforcement learning locally! (<5GB VRAM)

[P] Training RL agent to reach #1 in Teamfight Tactics through 100M simulated games

Thinking of creating an agent, need ideas

[P] SDLArch-RL is now compatible with Citra!!!! And we'll be training Street Fighter 6!!!

[P] RLHF (SFT, RM, PPO) with GPT-2 in Notebooks

[R] My RL agent taught itself a complete skill progression using only a “boredom” signal (no rewards)

[Deep Dive] How We Solved Poker: From Academic Bots to Superhuman AI (1998-2025)

RL ReSearch

Save item

DR-Tulu-8B

Xiaomi MiMo

Save item

MiMo-7B-RL