skills.rest
Gonglitian
GonglitianCOMMUNITY
Explore all skills by this author →
GitHub

paper_rob__ibrl

Select actions between RL and BC policies using a dual-Q critic.

2|Updated Mar 20, 2026
Search Tags:#robotics#reinforcement-learning#imitation-learning#ibrl#dual-critic#sample-efficiency#policy-selection
One-click install
npx skills add https://github.com/Gonglitian/agent-skills --skill paper-rob-ibrl
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: paper_rob__ibrl
Source: https://github.com/Gonglitian/agent-skills/tree/main/skills/papers/paper_rob__ibrl
Command: npx skills add https://github.com/Gonglitian/agent-skills --skill paper-rob-ibrl

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

IBRL provides a sample-efficient approach to robotic manipulation by dynamically selecting between a reinforcement learning (RL) policy and a behavior cloning (BC) policy using a dual-Q critic. This enables faster learning from demonstrations and online interactions, reducing data requirements and improving performance on complex tasks.

Core Features & Use Cases

  • Q-based action selection between RL and BC actions for exploration and bootstrapping.
  • Supports multiple modalities: pixel-based Robomimic tasks and state-based Meta-World tasks, with configurable encoders and critics.
  • Training paradigms: IBRL, RLPD, and RFT, across both simulation and real-like settings, with pretraining and fine-tuning workflows.
  • Real-world use case: a robot learns from demonstrations and gradually takes over with RL as it improves, reducing annotation costs.

Quick Start

Load the IBRL config and start a training run to observe improved sample efficiency.

Frequently Asked Questions about paper_rob__ibrl

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How does Q-based action selection improve sample efficiency in robotic manipulation?▼

You can apply this approach to pixel-based Robomimic tasks and state-based Meta-World tasks, with configurable encoders like ViT, ResNet, or DrQ supporting both simulation and real-like robotic manipulation settings.

How do I start training a reinforcement learning policy with behavior cloning bootstrapping?▼

To start training, load the IBRL configuration and initiate a run to observe improved sample efficiency, utilizing a PyTorch-based setup with a dual-Q critic, an RL actor, a BC policy, and a custom replay buffer to coordinate learning.

Do I need a PyTorch setup with specific components to use dual-critic reinforcement learning for robots?▼

This Q-guided RL approach distinguishes itself by dynamically selecting between RL and BC actions for exploration and bootstrapping, reducing data requirements compared to standard reinforcement learning without behavior cloning integration.

What training paradigms are supported for imitation learning and reinforcement learning integration?▼

This approach is limited to robotic manipulation tasks compatible with Robomimic or Meta-World environments, requiring precise configuration of encoders, critics, and replay buffers to function correctly.

Can I use a modulated encoder like ViT or ResNet for state-based Meta-World robotic tasks?▼

Yes, you can use configurable modulated encoders including ViT, ResNet, and DrQ to process inputs for both pixel-based Robomimic tasks and state-based Meta-World robotic manipulation tasks.

Related Skills

Explore More Education & Research Skills→
nahtonajnahtonaj

agentdb-learning-plugins

Create, train, and deploy AgentDB reinforcement learning plugins via CLI.

Community
Advanced
RyanJarvRyanJarv

AgentDB Learning Plugins

Create and train reinforcement learning plugins for autonomous agents.

Community
Advanced
bensonlee5bensonlee5

smolvla-metaworld

Execute robotic manipulation tasks from natural language descriptions using a Sawyer robot.

Community
Advanced
Marcus-Mok-GHMarcus-Mok-GH

AgentDB Learning Plugins

Train AI learning plugins using nine reinforcement learning algorithms.

Community
Advanced