-
Why SUMO’s Rendered Videos Should Never Be Used as RL Training Data
A examination of why the visual output of SUMO. Despite being clean and intuitive, it cannot serve as learning data for RL agents, and why this limitation is inherent in how the simulator is built.
-
Re-running an RL Experiment and Getting a Different Answer
A engineering reflection on why two RTX 4090 machines produced diverging RL curves despite identical code, seeds, and configurations. And what this reveals about RL’s numerical sensitivity.
-
Using Local v2rayN Proxy for Cloud Servers via SSH Reverse Tunnel
A practical record of troubleshooting outbound network restrictions on Chinese cloud servers and enabling stable access to foreign academic resources.