Selection Is Not Learning: Choosing Model Futures

Key points

• Current loops such as R-Zero, Agent0, and Skill Self-Play make this decision through the task, admitting tasks that look learnable and inheriting every resulting update.1
• We identify a selection-to-update gap in these loops, where a task that looks learnable need not produce a useful update.1
• EvoHeir separates parameter inheritance from experience inheritance, using current and archived experience to repair failures and preserve successes before final confirmation.1
• Our analysis characterizes selection-induced optimism and decomposes regret into selection, absorption, and confirmation errors.1

Sources (2)

• [1]Selection Is Not Learning: Choosing Model Futures
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 03:07 PM

  Self-evolving language models train on tasks they generate, so every round must decide which updated model the next round inherits.

  Current loops such as R-Zero, Agent0, and Skill Self-Play make this decision through the task, admitting tasks that look learnable and inheriting every resulting update.

• [2]Selection Is Not Learning: Choosing Model Futures
  Hugging Face Daily Papers · Oct 15, 12:00 AM · same content

Extractive summary: sentences quoted from the sources.