System Switch: When Should a Fast Decision Model Stop and Think?

Key points

• In real-time settings the slow model usually runs continuously; in turn-based agents and robot planners it is invoked on events such as uncertainty or a detected failure.1
• We study a fast learned actor that takes every decision and hands control to a reasoning vision-language model only when a gate opens, while the game keeps running.1
• We use closed-loop Doom and the new open "System One" typed-decision models, served through a common llama.cpp interface.1
• We release code, prompts, data and logs.1

Sources (2)

• [1]System Switch: When Should a Fast Decision Model Stop and Think?
  Hugging Face Daily Papers · Oct 7, 12:00 AM

  Dual-process agents pair a fast policy with a slow deliberative model.

  In real-time settings the slow model usually runs continuously; in turn-based agents and robot planners it is invoked on events such as uncertainty or a detected failure.

• [2]System Switch: When Should a Fast Decision Model Stop and Think?
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 08:44 AM · same content

Extractive summary: sentences quoted from the sources.