Action Correction

Action correction in artificial intelligence focuses on refining agent actions to improve performance, safety, and efficiency across various tasks, from robotic control to conversational AI. Current research emphasizes methods like reinforcement learning with regularization techniques to smooth jerky movements in robots and multi-task learning approaches that decompose policies to mitigate task conflicts. These advancements leverage model architectures such as transformers and incorporate human feedback or curated datasets to enhance learning and address limitations of existing methods, ultimately leading to more robust and adaptable intelligent systems. The impact spans improved robotic dexterity, safer human-robot interaction, and more reliable conversational agents.

Papers

October 8, 2024

Reinforcement Learning From Imperfect Corrective Actions And Proxy Rewards
Zhaohui Jiang, Xuening Feng, Paul Weng, Yifei Zhu, Yan Song, Tianze Zhou, Yujing Hu, Tangjie Lv, Changjie Fan
Reinforcement Learning Proxy Reward Sparse Semantic Value Based Deep Reinforcement Learning Agent Alignment Action Correction

July 5, 2024

Gradient-based Regularization for Action Smoothness in Robotic Control with Reinforcement Learning
I Lee, Hoang-Giang Cao, Cong-Tinh Dao, Yu-Cheng Chen, I-Chen Wu
Reinforcement Learning Deep Reinforcement Learning Reinforcement Learning Algorithm Robot Control Smooth Animation Action Correction

April 9, 2024

Efficient Multi-Task Reinforcement Learning via Task-Specific Action Correction
Jinyuan Feng, Min Chen, Zhiqiang Pu, Tenghai Qiu, Jianqiang Yi
Multi Objective Sparse Reward Multi Task Reinforcement Learning Action Correction

May 17, 2023

Scalable and Safe Remediation of Defective Actions in Self-Learning Conversational Systems
Sarthak Ahuja, Mohammad Kachuee, Fateme Sheikholeslami, Weiqing Liu, Jaeyoung Do
Potential Scalability Policy Evaluation Policy Reinforcement Learning Conversational System Policy Improvement Conversational Artificial Intelligence Remediation Strategy Action Correction

November 18, 2022

Language-Conditioned Reinforcement Learning to Solve Misunderstandings with Action Corrections
Frank Röder, Manfred Eppe
Reinforcement Learning Language Conditioned Robot Goal Action Correction

June 25, 2022

Guided Exploration in Reinforcement Learning via Monte Carlo Critic Optimization
Igor Kuznetsov
Reinforcement Learning Exploration Method Guided Exploration Policy Deep Reinforcement Learning Modern Reinforcement Learning Action Correction Unsupervised Calibration

June 22, 2022

Behavior Transformers: Cloning $k$ modes with one stone
Nur Muhammad Mahi Shafiullah, Zichen Jeff Cui, Ariuntuya Altanzaya, Lerrel Pinto
Behavior Cloning STONE Deterioration Pattern Behaviour Learning Behavior Transformer Action Correction K Mode

January 19, 2022

Improving Behavioural Cloning with Human-Driven Dynamic Dataset Augmentation
Federico Malato, Joona Jehkonen, Ville Hautamäki
Simulation Study Behavior Cloning Human in the Loop Expert Trajectory Action Correction