Recursive Self-Improvement of Visuomotor Policies through Local Recovery Supervision
cs.RO
Submitted: 2026-10-04
Updated: 2026-10-04
Comments: 19 pages, 4 figures, 2 tables
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- Deep Reinforcement Learning at the Edge of the Statistical Precipice
- MAGMA-GEN: Validated Recovery Supervision from Ambiguous Failures via Counterfactual Re-Execution
- RoboCat: A Self-Improving Generalist Agent for Robotic Manipulation
- RoboDojo: A Unified Sim-and-Real Benchmark for Comprehensive Evaluation of Generalist Robot Manipulation Policies
- RACER: Rich Language-Guided Failure Recovery Policies for Imitation Learning
- ProTracer: Proprioception-Guided Failure Diagnosis in Robot Manipulation
- IntervenGen: Interventional Data Generation for Robust and Data-Efficient Robot Imitation Learning
- LoRA: Low-Rank Adaptation of Large Language Models
- RaC: Robot Learning for Long-Horizon Tasks by Scaling Recovery and Correction
- Agent as Policy for Robotic Manipulation
- HG-DAgger: Interactive Imitation Learning with Human Experts
- Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planning
- EmbodiRSI: Recursive Self-Improvement for Data-Efficient Robot Adaptation
- Learning Actionable Manipulation Recovery via Counterfactual Failure Synthesis
- RoboRecover: Benchmarking Robot Policy Recovery under Execution Deviations
- Code as Policies: Language Model Programs for Embodied Control
- LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning
- LIBERO-RECOVER: Beyond Task Success Towards Failure Recovery in Robotic Manipulation Models
- AERIS: Aerial-Edge Role-Driven Intelligence at Runtime via Orchestrated Language-Model Swarm
- AeroWeaver: An Embodied-Agent Harness for Weaving Aerial Skills into Distributed, Adaptive Swarm Execution
Related papers
- FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
- MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
- RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
- ARC: A Reasoning Recipe for Robot Foundation Models
- HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
- APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies