Independent, unofficial Dream-RSI-inspired Python SDK. Alpha: model-agnostic adapters, recorded-tree replay and policy search. Not affiliated with Google.
-
Updated
Sep 30, 2026 - Python
#
dream-rsi
Here are 4 public repositories matching this topic...
Self-improvement research testbed for LLMs: Dream-RSI-style policy evolution (arXiv:2609.14858) + STaR-style weight upgrades, with a Rust rootless sandbox. Apache-2.0.
-
Updated
Sep 19, 2026 - Python
Dream-RSI recursive self-improvement as a DeepSeek Harness plugin: discovery history becomes a replay simulator; candidate exploration policies are dream-evaluated off-policy and redeployed under a no-regression guarantee.
dsh circle-packing cordis recursive-self-improvement alphaevolve agent-plugin deepseek-harness dream-rsi
-
Updated
Oct 2, 2026 - TypeScript
A research harness for your next breakthrough: Dream-RSI-inspired search-policy evolution with replay, budgets and noise-aware promotion. Research prototype · Python · MIT.
python bootstrap program-synthesis experiment-design research-automation agent-skills autoresearch dream-rsi search-policy
-
Updated
Sep 29, 2026 - Python
Add this topic to your repo
To associate your repository with the dream-rsi topic, visit your repo's landing page and select "manage topics."