Awesome Multi-Modal Reinforcement Learning

This is a collection of research papers for Multi-Modal reinforcement learning (MMRL). And the repository will be continuously updated to track the frontier of MMRL. Some papers may not be relevant to RL, but we include them anyway as they may be useful for the research of MMRL.

Welcome to follow and star!

Introduction

Multi-Modal RL agents focus on learning from video (images), language (text), or both, as humans do. We believe that it is important for intelligent agents to learn directly from images or text, since such data can be easily obtained from the Internet.

Table of Contents

• A Taxonomy of Multi-Modal Reinforcement Learning
• Papers
  • ICLR 2024(New!!!)
  • ICLR 2023
  • ICLR 2022
  • ICLR 2021
  • ICLR 2019
  • Neurips 2023
  • Neurips 2022
  • Neurips 2021
  • Neurips 2018
  • ICML 2022
  • ICML 2019
  • ICML 2017
  • CVPR 2022
  • CoRL 2022
  • Arxiv
• Contributing

Papers

format:
- [title](paper link) [links]
  - authors.
  - key words.
  - experiment environment.

ICLR 2024

• DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization

  • Guowei Xu, Ruijie Zheng, Yongyuan Liang, Xiyao Wang, Zhecheng Yuan, Tianying Ji, Yu Luo, Xiaoyu Liu, Jiaxin Yuan, Pu Hua, Shuzhen Li, Yanjie Ze, Hal Daumé III, Furong Huang, Huazhe Xu
  • Keyword: Visual RL; Dormant Ratio
  • ExpEnv: DeepMind Control Suite,Meta-world,Adroit

• Revisiting Data Augmentation in Deep Reinforcement Learning

  • Jianshu Hu, Yunpeng Jiang, Paul Weng
  • Keyword: Reinforcement Learning, Data Augmentation
  • ExpEnv: DeepMind Control Suite

• Revisiting Plasticity in Visual Reinforcement Learning: Data, Modules and Training Stages

  • Guozheng Ma, Lu Li, Sen Zhang, Zixuan Liu, Zhen Wang, Yixin Chen, Li Shen, Xueqian Wang, Dacheng Tao
  • Keyword: Plasticity, Visual Reinforcement Learning, Deep Reinforcement Learning, Sample Efficiency
  • ExpEnv: DeepMind Control Suite,Atari

• Entity-Centric Reinforcement Learning for Object Manipulation from Pixels

  • Dan Haramati, Tal Daniel, Aviv Tamar
  • Keyword: deep reinforcement learning, visual reinforcement learning, object-centric, robotic object manipulation, compositional generalization
  • ExpEnv: IsaacGym

ICLR 2023

• PaLI: A Jointly-Scaled Multilingual Language-Image Model(notable top 5%)

  • Xi Chen, Xiao Wang, Soravit Changpinyo, AJ Piergiovanni, Piotr Padlewski, Daniel Salz, Sebastian Goodman, Adam Grycner, Basil Mustafa, Lucas Beyer, Alexander Kolesnikov, Joan Puigcerver, Nan Ding, Keran Rong, Hassan Akbari, Gaurav Mishra, Linting Xue, Ashish Thapliyal, James Bradbury, Weicheng Kuo, Mojtaba Seyedhosseini, Chao Jia, Burcu Karagol Ayan, Carlos Riquelme, Andreas Steiner, Anelia Angelova, Xiaohua Zhai, Neil Houlsby, Radu Soricut
  • Keyword: amazing zero-shot, language component and visual component
  • ExpEnv: None

• VIMA: General Robot Manipulation with Multimodal Prompts

  • Yunfan Jiang, Agrim Gupta, Zichen Zhang, Guanzhi Wang, Yongqiang Dou, Yanjun Chen, Li Fei-Fei, Anima Anandkumar, Yuke Zhu, Linxi Fan. NeurIPS Workshop 2022
  • Key Words: multimodal prompts, transformer-based generalist agent model, large-scale benchmark
  • ExpEnv: VIMA-Bench, VIMA-Data

• MIND ’S EYE: GROUNDED LANGUAGE MODEL REASONING THROUGH SIMULATION

  • Ruibo Liu, Jason Wei, Shixiang Shane Gu, Te-Yen Wu, Soroush Vosoughi, Claire Cui, Denny Zhou, Andrew M. Dai
  • Keyword: language2physical-world, reasoning ability
  • ExpEnv: MuJoCo

ICLR 2022

• How Much Can CLIP Benefit Vision-and-Language Tasks?
  • Sheng Shen, Liunian Harold Li, Hao Tan, etc. ICLR 2022
  • Key Words: Vision-and-Language, CLIP
  • ExpEnv: None

ICLR 2021

• Grounding Language to Entities and Dynamics for Generalization in Reinforcement Learning

  • Austin W. Hanjie, Victor Zhong, Karthik Narasimhan. ICML 2021
  • Key Words: Multi-modal Attention
  • ExpEnv: Messenger

• Mastering Atari with Discrete World Models

  • Danijar Hafner, Timothy Lillicrap, Mohammad Norouzi, etc.
  • Key Words: World models
  • ExpEnv: Atari

• Decoupling Representation Learning from Reinforcement Learning

  • Adam Stooke,Kimin Lee,Pieter Abbeel, etc.
  • Key Words: representation learning, unsupervised learning
  • ExpEnv: DeepMind Control, Atari, DMLab

ICLR 2019

• Learning Actionable Representations with Goal-Conditioned Policies
  • Dibya Ghosh, Abhishek Gupta, Sergey Levine.
  • Key Words: Actionable Representations Learning
  • ExpEnv: 2D navigation(2D Wall, 2D Rooms, Wheeled, Wheeled Rooms, Ant, Pushing)

NeurIPS 2023

• Language Is Not All You Need: Aligning Perception with Language Models
  • Shaohan Huang, Li Dong, Wenhui Wang, Yaru Hao, etc.
  • Key Words: Multimodal Perception, World Modeling
  • ExpEnv: IQ50

NeurIPS 2022

• MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge

  • Linxi Fan, Guanzhi Wang, Yunfan Jiang, etc.
  • Key Words: multimodal dataset, MineClip
  • ExpEnv: Minecraft

• Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos

  • Bowen Baker, Ilge Akkaya, Peter Zhokhov, etc.
  • Key Words: Inverse Dynamics Model
  • ExpEnv: minerl

NeurIPS 2021

• SOAT: A Scene-and Object-Aware Transformer for Vision-and-Language Navigation

  • Abhinav Moudgil, Arjun Majumdar,Harsh Agrawal, etc.
  • Key Words: Vision-and-Language Navigation
  • ExpEnv: Room-to-Room, Room-Across-Room

• Pretraining Representations for Data-Efficient Reinforcement Learning

  • Max Schwarzer, Nitarshan Rajkumar, Michael Noukhovitch, etc.
  • Key Words: latent dynamics modelling, unsupervised RL
  • ExpEnv: Atari

NeurIPS 2018

• Recurrent World Models Facilitate Policy Evolution
  • David Ha, Jürgen Schmidhuber.
  • Key Words: World model, generative RNN, VAE
  • ExpEnv: VizDoom, CarRacing

ICML 2022

• Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents

  • Wenlong Huang, Pieter Abbeel, Deepak Pathak, etc.
  • Key Words: large language models, Embodied Agents
  • ExpEnv: VirtualHome

• Reinforcement Learning with Action-Free Pre-Training from Videos

  • Younggyo Seo, Kimin Lee, Stephen L James, etc.
  • Key Words: action-free pretraining, videos
  • ExpEnv: Meta-world, DeepMind Control Suite

• History Compression via Language Models in Reinforcement Learning

  • Fabian Paischer, Thomas Adler, Vihang Patil, etc.
  • Key Words: Pretrained Language Transformer
  • ExpEnv: Minigrid, Procgen

ICML 2019

• Learning Latent Dynamics for Planning from Pixels
  • Danijar Hafner, Timothy Lillicrap, Ian Fischer, etc.
  • Key Words: latent dynamics model, pixel observations
  • ExpEnv: DeepMind Control Suite

ICML 2017

• Zero-Shot Task Generalization with Multi-Task Deep Reinforcement Learning
  • Junhyuk Oh, Satinder Singh, Honglak Lee, Pushmeet Kohli
  • Key Words: unseen instruction, sequential instruction
  • ExpEnv: Minecraft

CVPR 2022

• End-to-end Generative Pretraining for Multimodal Video Captioning

  • Paul Hongsuck Seo, Arsha Nagrani, Anurag Arnab, Cordelia Schmid
  • Key Words: Multimodal video captioning, Pretraining using a future utterance, Multimodal Video Generative Pretraining
  • ExpEnv: HowTo100M

• Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks

  • Wenhui Wang, Hangbo Bao, Li Dong, Johan Bjorck, Zhiliang Peng, Qiang Liu, Kriti Aggarwal, Owais Khan Mohammed, Saksham Singhal, Subhojit Som, Furu Wei
  • Key Words: backbone architecture, pretraining task, model scaling up
  • ExpEnv: ADE20K, COCO, NLVR2, Flickr30K

• Think Global, Act Local: Dual-scale Graph Transformer for Vision-and-Language Navigation

  • Shizhe Chen, Pierre-Louis Guhur, Makarand Tapaswi, Cordelia Schmid, Ivan Laptev
  • Keyword: dual-scale graph transformer, dual-scale graph transformer, affordance detection
  • ExpEnv: None

• Masked Visual Pre-training for Motor Control

  • Tete Xiao, Ilija Radosavovic, Trevor Darrell, etc. ArXiv 2022
  • Key Words: self-supervised learning, motor control
  • ExpEnv: Isaac Gym

CoRL 2022

• LM-Nav: Robotic Navigation with Large Pre-Trained Models of Language, Vision, and Action

  • Dhruv Shah, Blazej Osinski, Brian Ichter, Sergey Levine
  • Key Words: robotic navigation, goal-conditioned, unannotated large dataset, CLIP, ViNG, GPT-3
  • ExpEnv: None

• [Real-World Robot Learning with Masked Visual Pre-training](https://arxiv.org/abs/2210.03109）

  • Ilija Radosavovic, Tete Xiao, Stephen James, Pieter Abbeel, Jitendra Malik, Trevor Darrell
  • Key Words: real-world robotic tasks，
  • ExpEnv: None

• R3M: A Universal Visual Representation for Robot Manipulation

  • Suraj Nair, Aravind Rajeswaran, Vikash Kumar, etc.
  • Key Words: Ego4D human video dataset, pre-train visual representation
  • ExpEnv: MetaWorld, Franka Kitchen, Adroit

Other

• Language Conditioned Imitation Learning over Unstructured Data RSS 2021

  • Corey Lynch, Pierre Sermanet
  • Keyword: open-world environments
  • ExpEnv: None

• Learning Generalizable Robotic Reward Functions from “In-The-Wild” Human Videos RSS 2021

  • Annie S. Chen, Suraj Nair, Chelsea Finn.
  • Key Words: Reward Functions, “In-The-Wild” Human Videos
  • ExpEnv: None

• Offline Reinforcement Learning from Images with Latent Space Models L4DC 2021

  • Rafael Rafailov, Tianhe Yu, Aravind Rajeswaran, etc.
  • Key Words: Latent Space Models
  • ExpEnv: DeepMind Control, Adroit Pen, Sawyer Door Open, Robel D’Claw Screw

• Is Cross-Attention Preferable to Self-Attention for Multi-Modal Emotion Recognition? ICASSP 2022

  • Vandana Rajan, Alessio Brutti, Andrea Cavallaro.
  • Key Words: Multi-Modal Emotion Recognition, Cross-Attention
  • ExpEnv: None

ArXiv

• Spatialvlm: Endowing vision-language models with spatial reasoning capabilities

  • Boyuan Chen, Zhuo Xu, Sean Kirmani, Brian Ichter, Danny Driess, Pete Florence, Dorsa Sadigh, Leonidas Guibas, Fei Xia
  • Key Words: Visual Question Answering, 3D Spatial Reasoning
  • ExpEnv: spatial VQA dataset

• M2CURL: Sample-Efficient Multimodal Reinforcement Learning via Self-Supervised Representation Learning for Robotic Manipulation

  • Fotios Lygerakis, Vedant Dave, Elmar Rueckert
  • Key Words: Robotic Manipulation, Self-supervised representation
  • ExpEnv: Gym

• On Time-Indexing as Inductive Bias in Deep RL for Sequential Manipulation Tasks

  • M. Nomaan Qureshi, Ben Eisner, David Held
  • Key Words: Multimodality of policy output, Action head switching
  • ExpEnv: MetaWorld

• Parameterized Decision-making with Multi-modal Perception for Autonomous Driving

  • Yuyang Xia, Shuncheng Liu, Quanlin Yu, Liwei Deng, You Zhang, Han Su, Kai Zheng
  • Key Words: Autonomous driving, GNN in RL
  • ExpEnv: CARLA

• A Contextualized Real-Time Multimodal Emotion Recognition for Conversational Agents using Graph Convolutional Networks in Reinforcement Learning

  • Fathima Abdul Rahman, Guang Lu
  • Key Words: Emotion Recognition, GNN in RL
  • ExpEnv: IEMOCAP

• Reinforced UI Instruction Grounding: Towards a Generic UI Task Automation API

  • Zhizheng Zhang, Wenxuan Xie, Xiaoyi Zhang, Yan Lu
  • Key Words: LLM, generic UI task automation API
  • ExpEnv: RicoSCA, MoTIF

• Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Driving

  • Long Chen, Oleg Sinavski, Jan Hünermann, Alice Karnsund, Andrew James Willmott, Danny Birch, Daniel Maund, Jamie Shotton
  • Key Words: LLM in Autonomous Driving, object-level multimodal LLM
  • ExpEnv: RicoSCA, MoTIF

• Nonprehensile Planar Manipulation through Reinforcement Learning with Multimodal Categorical Exploration

  • Juan Del Aguila Ferrandis, João Moura, Sethu Vijayakumar
  • Key Words: multimodal exploration approach
  • ExpEnv: KUKA iiwa robot arm

• End-to-End Streaming Video Temporal Action Segmentation with Reinforce Learning

  • Wujun Wen, Jinrong Zhang, Shenglan Liu, Yunheng Li, Qifeng Li, Lin Feng
  • Key Words: Temporal Action Segmentation, RL in Video Analysis
  • ExpEnv: EGTEA

• Do as I can, not as I get:Topology-aware multi-hop reasoningon multi-modal knowledge graphs

  • Shangfei Zheng, Hongzhi Yin, Tong Chen, Quoc Viet Hung Nguyen, Wei Chen, Lei Zhao
  • Key Words: Multi-hop reasoning, multi-modal knowledge graphs, inductive setting, adaptive reinforcement learning
  • ExpEnv: None

• Multimodal Reinforcement Learning for Robots Collaborating with Humans

  • Afagh Mehri Shervedani, Siyu Li, Natawut Monaikul, Bahareh Abbasi, Barbara Di Eugenio, Milos Zefran
  • Key Words: robust and deliberate decisions, end-to-end training, importance enhancement, similarity, improve IRL training process multimodal RL domains
  • ExpEnv: None

• See, Plan, Predict: Language-guided Cognitive Planning with Video Prediction

  • Maria Attarian, Advaya Gupta, Ziyi Zhou, Wei Yu, Igor Gilitschenski, Animesh Garg
  • Keyword: cognitive planning, language-guided video prediction
  • ExpEnv: None

• Open-vocabulary Queryable Scene Representations for Real World Planning

  • Boyuan Chen, Fei Xia, Brian Ichter, Kanishka Rao, Keerthana Gopalakrishnan, Michael S. Ryoo, Austin Stone, Daniel Kappler
  • Key Words: Target Detection, Real World, Robotic Tasks
  • ExpEnv: Say Can

• Do As I Can, Not As I Say: Grounding Language in Robotic Affordances

  • Michael Ahn, Anthony Brohan, Noah Brown, Yevgen Chebotar, Omar Cortes, Byron David, Chelsea Finn, Chuyuan Fu, Keerthana Gopalakrishnan, Karol Hausman, Alex Herzog, Daniel Ho, Jasmine Hsu, Julian Ibarz, Brian Ichter, Alex Irpan, Eric Jang, Rosario Jauregui Ruano, Kyle Jeffrey, Sally Jesmonth, Nikhil J Joshi, Ryan Julian, Dmitry Kalashnikov, Yuheng Kuang, Kuang-Huei Lee, Sergey Levine, Yao Lu, Linda Luu, Carolina Parada, Peter Pastor, Jornell Quiambao, Kanishka Rao, Jarek Rettinghouse, Diego Reyes, Pierre Sermanet, Nicolas Sievers, Clayton Tan, Alexander Toshev, Vincent Vanhoucke, Fei Xia, Ted Xiao, Peng Xu, Sichun Xu, Mengyuan Yan, Andy Zeng
  • Key Words: real world, natural language
  • ExpEnv: Say Can

Contributing

Our purpose is to make this repo even better. If you are interested in contributing, please refer to HERE for instructions in contribution.

License

Awesome Multi-Modal Reinforcement Learning is released under the Apache 2.0 license.