AGENTIC TEST-TIME SCALING

Summary from the TLDR AI newsletter

AutoTTS explores automated strategy discovery for test-time scaling by using coding agents to iteratively refine controller logic inside a replay environment, avoiding gradient updates and online LLM calls.