[ICML 2026] RLAnything & DemyAgent: General and scalable agentic RL algorithms across terminal, GUI, SWE, and tool-call settings

Scalable and extensible reinforcement learning for LM agents.

Apprentissage par renforcement pour la régulation des feux de signalisation sur un boulevard typique de Kinshasa

A curated list of training & evaluation environments for LLM/VLM agents (SWE-Gym, GEM, RAGEN, AgentGym, WebArena, OSWorld, ToolBench…). Updated weekly.

Train SLM to use Tools with RL

Compiler feedback as process reward for coding agent RL training (Junhao Fu, 2025)