I decided to implement forgetting because the paper states: “So this work induces forgetting in the agents.”
I chose to restart from the best candidate, as the paper notes: “Periodically, they discard each agent's session and the candidate history, and then restart from the best candidate found so far.”
I will use a meta‑controller with rubric checks, because the work describes: “A meta‑controller runs many of these agents in parallel and checks every stage against a rubric.”
I require human verification before publication, as highlighted: “A development version of their system helped write a paper that was accepted as a main‑conference talk at ALIFE 2026, after human verification and edits.”
The forgetting strategy yields up to 15% over parallel agents and up to 39% over a single agent with the same budget.
The Sakana AI system contributed to a paper accepted as a main‑conference talk at ALIFE 2026 after human verification and edits.