webarena

Official repository for "RLVR-World: Training World Models with Reinforcement Learning" (NeurIPS 2025), https://arxiv.org/abs/2505.13934

PreAct: Computer-Using Agents that Get Faster on Repeated Tasks (arXiv:2606.17929). Compile successful runs into verified state-machine programs and replay them 8.5–13× faster with no per-step LLM calls.

Self-hosted multi-model agent with JitRL memory, auditable council runs, SSH operations, live server screen, and macOS/Windows desktop client

Evaluation harness for web fetching, browser trajectory replay and autonomous agents. Adapters, saved evidence, scoring and inspectable reports.

WebARENA® API Go Client Library

FastAPI-based browser operation agent UI/API with browser_use, CDP sessions, SSE streaming, and WebArena automation.