# Sakana AI Lifts Robot Task Success From 25 to 73 Percent by Searching in Simulation Before Moving

A vision-language model proposes trajectories, a second model judges them from simulated video, and Monte Carlo tree search picks the plan sent to the physical arm.

- Published: 2026-09-29T06:12:03.689Z
- Canonical: https://polylog.news/ai/2026-09-29/sakana-ai-lifts-robot-task-success-from-25-to-73-percent-by
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news)

Sakana AI and the University of Tokyo have published SAIL, a test-time scaling method for in-context imitation learning. The system gives a vision-language model a few successful demonstrations, has it generate a full robot trajectory, runs…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-09-29/sakana-ai-lifts-robot-task-success-from-25-to-73-percent-by (subscription information: https://polylog.news/pricing).