goal2skill-long-horizon-manipulation-with-adaptive

---
name: goal2skill-long-horizon-manipulation-with-adaptive
description: 'Research paper: Goal2Skill: Long-Horizon Manipulation with Adaptive Planning and Reflection'
metadata:
  source: arXiv
  arxiv_id: 2604.13942
  published: 2026-04-15
  utility_score: 0.91
  keywords: agentic, memory, long-horizon, reasoning, planning, reflection
---

# Goal2Skill: Long-Horizon Manipulation with Adaptive Planning and Reflection

**arXiv ID:** 2604.13942  
**Published:** 2026-04-15  
**Utility Score:** 0.91  
**URL:** http://arxiv.org/abs/2604.13942

## Authors
Zhen Liu, Xinyu Ning, Zhe Hu

## Categories
cs.RO

## Abstract
Recent vision-language-action (VLA) systems have demonstrated strong capabilities in embodied manipulation. However, most existing VLA policies rely on limited observation windows and end-to-end action prediction, which makes them brittle in long-horizon, memory-dependent tasks with partial observability, occlusions, and multi-stage dependencies. Such tasks require not only precise visuomotor control, but also persistent memory, adaptive task decomposition, and explicit recovery from execution failures. To address these limitations, we propose a dual-system framework for long-horizon embodied manipulation.   Our framework explicitly separates high-level semantic reasoning from low-level motor execution. A high-level planner, implemented as a VLM-based agentic module, maintains structured task memory and performs goal decomposition, outcome verification, and error-driven correction. A low-level executor, instantiated as a VLA-based visuomotor controller, carries out each sub-task through diffusion-based action generation conditioned on geometry-preserving filtered observations. Together, the two systems form a closed loop between planning and execution, enabling memory-aware reasoning, adaptive replanning, and robust online recovery. Experiments on representative RMBench tasks show that the proposed framework substantially outperforms representative baselines, achieving a 32.4% average success rate compared with 9.8% for the strongest baseline. Ablation studies further confirm the importance of structured memory and closed-loop recovery for long-horizon manipulation.

## Matched Keywords
agentic, memory, long-horizon, reasoning, planning, reflection

## Relevance to AI Agents
This paper is highly relevant to AI agent systems research with focus on:
- agentic, memory, long-horizon, reasoning, planning

## Quick Reference
```bash
# View paper
open http://arxiv.org/abs/2604.13942

# Download PDF
open http://arxiv.org/pdf/2604.13942.pdf
```

---
*Auto-generated from arXiv on 2026-04-17*

## Activation Keywords

- "goal2skill-long-horizon-manipulation-with-adaptive"
- "goal2skill long horizon manipulation with adaptive"
- "use goal2skill long horizon manipulation with adaptive"
- "goal2skill long horizon manipulation with adaptive help"
- "goal2skill long horizon manipulation with adaptive tool"

## Tools Used

- `Read` - Read existing files and documentation
- `Write` - Create new files and documentation
- `Bash` - Execute commands when needed

## Instructions for Agents

1. Identify user's intent and specific requirements
2. Gather necessary context from files or user input
3. Execute appropriate actions using available tools
4. Provide clear results and suggest next steps

## Examples

### Basic Goal2Skill Long Horizon Manipulation With Adaptive usage
```
User: "Help me with goal2skill long horizon manipulation with adaptive"
→ Understand requirements → Execute actions → Provide results
```

### Advanced usage
```
User: "I need detailed goal2skill long horizon manipulation with adaptive assistance"
→ Clarify scope → Provide comprehensive solution → Follow up
```