Google DeepMind, UMD, UVA (authors) · 2026-09-14
Dream-RSI: Recursive Self-Improvement through Evolving Worlds
“A lightweight orchestration layer makes exploration explicit and programmable while leaving the underlying base agent unchanged.”
- Who
- Google DeepMind, UMD, UVA (authors)
- Date
- 2026-09-14
- Evidence
- paper or code what this means
- Replication
- None recorded
- Checked
- 2026-10-14, against the source page
- Links
- Source Code
Also on the page
- The method improves the exploration policy; model weights are not updated.
- The 162x figure compares against another method's reported numbers, not against a fixed baseline run.
What got repeated wrong
- contradictedGoogle cracked RSI / Google open-sourced Dream-RSI.It is a research paper with a personal-account repo, not a Google release.