Bimanual Robot Manipulation via Multi-Agent In-Context Learning
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917428032700416 |
|---|---|
| author | Palma, Alessio Spinelli, Indro Prasad, Vignesh Scofano, Luca Jin, Yufeng Chalvatzaki, Georgia Galasso, Fabio |
| author_facet | Palma, Alessio Spinelli, Indro Prasad, Vignesh Scofano, Luca Jin, Yufeng Chalvatzaki, Georgia Galasso, Fabio |
| contents | Language Models (LLMs) have emerged as powerful reasoning engines for embodied control. In particular, In-Context Learning (ICL) enables off-the-shelf, text-only LLMs to predict robot actions without any task-specific training while preserving their generalization capabilities. Applying ICL to bimanual manipulation remains challenging, as the high-dimensional joint action space and tight inter-arm coordination constraints rapidly overwhelm standard context windows. To address this, we introduce BiCICLe (Bimanual Coordinated In-Context Learning), the first framework that enables standard LLMs to perform few-shot bimanual manipulation without fine-tuning. BiCICLe frames bimanual control as a multi-agent leader-follower problem, decoupling the action space into sequential, conditioned single-arm predictions. This naturally extends to Arms' Debate, an iterative refinement process, and to the introduction of a third LLM-as-Judge to evaluate and select the most plausible coordinated trajectories. Evaluated on 13 tasks from the TWIN benchmark, BiCICLe achieves up to 71.1% average success rate, outperforming the best training-free baseline by 6.7 percentage points and surpassing most supervised methods. We further demonstrate strong few-shot generalization on novel tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_20348 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Bimanual Robot Manipulation via Multi-Agent In-Context Learning Palma, Alessio Spinelli, Indro Prasad, Vignesh Scofano, Luca Jin, Yufeng Chalvatzaki, Georgia Galasso, Fabio Robotics Artificial Intelligence Multiagent Systems I.2.9; I.2.11; I.2.6 Language Models (LLMs) have emerged as powerful reasoning engines for embodied control. In particular, In-Context Learning (ICL) enables off-the-shelf, text-only LLMs to predict robot actions without any task-specific training while preserving their generalization capabilities. Applying ICL to bimanual manipulation remains challenging, as the high-dimensional joint action space and tight inter-arm coordination constraints rapidly overwhelm standard context windows. To address this, we introduce BiCICLe (Bimanual Coordinated In-Context Learning), the first framework that enables standard LLMs to perform few-shot bimanual manipulation without fine-tuning. BiCICLe frames bimanual control as a multi-agent leader-follower problem, decoupling the action space into sequential, conditioned single-arm predictions. This naturally extends to Arms' Debate, an iterative refinement process, and to the introduction of a third LLM-as-Judge to evaluate and select the most plausible coordinated trajectories. Evaluated on 13 tasks from the TWIN benchmark, BiCICLe achieves up to 71.1% average success rate, outperforming the best training-free baseline by 6.7 percentage points and surpassing most supervised methods. We further demonstrate strong few-shot generalization on novel tasks. |
| title | Bimanual Robot Manipulation via Multi-Agent In-Context Learning |
| topic | Robotics Artificial Intelligence Multiagent Systems I.2.9; I.2.11; I.2.6 |
| url | https://arxiv.org/abs/2604.20348 |