MMLA-RSI: Recursive Self-Improvement through RTT and RTMA
We propose MMLA-RSI, a recursive self-improvement scheme with two implementations in Memory-Mediated Learning Architecture: Reasoning-Time Training (RTT) changes the policy state $Φ$, while Reasoning-Time Memory Adaptation (RTMA) changes authoritative memory $M$.
ProofPaper ↗
Key points
- An agent can improve its behavior within an unresolved problem by changing how it acts and which experience it retains.
- We give bounded executable rule languages, shared task/proposal interfaces, and one algorithm for either route or their composition.
- The RTT instance changes scalar schedules in a clipped-momentum update.
- In chronological continuous-integration replay, feedback binding is positive while retention increases loss.
Sources (1)
- [1]MMLA-RSI: Recursive Self-Improvement through RTT and RTMAarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 02:47 PM
We propose MMLA-RSI, a recursive self-improvement scheme with two implementations in Memory-Mediated Learning Architecture: Reasoning-Time Training (RTT) changes the policy state $Φ$, while Reasoning-Time Memory Adaptation (RTMA) changes authoritative memory $M$.
An agent can improve its behavior within an unresolved problem by changing how it acts and which experience it retains.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 14, 2026mistralai/Timestral-209M
- Oct 13, 2026HERO: Harnessing Evidence for Test-time Refinement and Prompt Optimization in Program Generation
- Oct 13, 2026Run fast, private agents on-device with LFM2.5-24B-A2B
- Oct 13, 2026Gemini 4 Argon for Robotics
- Oct 13, 2026inclusionAI/Ling-3.1-flash
- Oct 13, 2026Fish Audio Drama 3 Technical Report