Learning to Compose: Recursive Self-Improvement of Shared Executable Stacks for Zero-Shot Long-Horizon Robot Manipulation
We introduce Recursia, a recursive self-improvement framework that treats the shared executable stack, rather than individual task programs, as the persistent unit of improvement.
ProofPaper ↗
Key points
- Individually competent robot skills can fail when composed: tools may lack a required capability, completion signals may misrepresent physical progress, and one skill may leave a state that violates its successor's prerequisites.
- Source-side execution traces guide failure attribution and the generation of missing tools or repairs to existing shared components.
- Analyses of constituent execution, candidate rejection, and inference drivers distinguish local mechanism improvements from complete task success and characterize the cost of feedback-conditioned orchestration.
- These results examine shared executable interfaces as a concrete mechanism for compositional transfer in robot programming.
Sources (1)
- [1]Learning to Compose: Recursive Self-Improvement of Shared Executable Stacks for Zero-Shot Long-Horizon Robot ManipulationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 17, 05:10 AM
We introduce Recursia, a recursive self-improvement framework that treats the shared executable stack, rather than individual task programs, as the persistent unit of improvement.
Individually competent robot skills can fail when composed: tools may lack a required capability, completion signals may misrepresent physical progress, and one skill may leave a state that violates its successor's prerequisites.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 17, 2026EdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge Backends
- Oct 17, 2026ECHO: Efficient Causal Self-Recursion Beyond Auxiliary Multi-Token Prediction
- Oct 16, 2026Momentum: Persistent Dynamics for Video World Models
- Oct 16, 2026Language Models Learn to Improve Language Models
- Oct 15, 2026On the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction
- Oct 15, 2026The Density of Thought: Predicting and Steering Reasoning in Large Language Models