Scaling Unlabeled Observational Data for Sample-Efficient Continuous Latent Action Models

This report synthesises findings from 15 peer-reviewed papers addressing the following research question: Does scaling unlabeled observational data improve the sample efficiency of continuous latent action models compared to supervised baselines on few-shot cross-task imitation benchmarks. 8 claims were extracted from source literature; 8 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 9.2/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: Does scaling unlabeled observational data improve the sample efficiency of continuous latent action models compared to supervised baselines on few-shot cross-task imitation benchmarks?

Autonomous literature synthesis. Automated review score: 9.2/10. Full text and citation available at Assignee Research.