mlsanityData to 7 Oct 2026

Hierarchical Decision Making by Generating and Following Natural Language Instructions

arXiv 1906.00744 ↗First posted 3 Jun 2019cs.AI cs.CLcited by 30

Not on the map. Why?

Abstract (arXiv)

We explore using latent natural language instructions as an expressive and compositional representation of complex actions for hierarchical decision making. Rather than directly selecting micro-actions, our agent first generates a latent plan in natural language, which is then executed by a separate model. We introduce a challenging real-time strategy game environment in which the actions of a large number of units must be coordinated across long time scales. We gather a dataset of 76 thousand pairs of instructions and executions from human play, and train instructor and executor models. Experiments show that models using natural language as a latent variable significantly outperform models that directly imitate human actions. The compositional structure of language proves crucial to its effectiveness for action representation. We also release our code, models and data.

Key facts. Cited by 30 papers in the map (data through 7 Oct 2026). First posted on arXiv 3 Jun 2019.

Builds on

No reference list yet.

Built on by

Cited by year

Citing papers per year: 2021: 7; 2022: 12; 2023: 7; 2024: 1; 2025: 3.

30 papers.

Sort:

By area:

  1. 2021Skill Induction and Planning with Latent Languagecited by 99
  2. 2022Improving Intrinsic Exploration with Language Abstractionscited by 49
  3. 2023KITE: Keypoint-Conditioned Policies for Semantic Manipulationcited by 39
  4. 2021Reason first, then respond: Modular Generation for Knowledge-infused Dialoguecited by 32
  5. 2022LISA: Learning Interpretable Skill Abstractions from Languagecited by 30
  6. 2023Thought Cloning: Learning to Think while Acting by Imitating Human Thinkingcited by 30
  7. 2022Interactive Grounded Language Understanding in a Collaborative Environment: IGLU 2021cited by 23
  8. 2022EAGER: Asking and Answering Questions for Automatic Reward Shaping in Language-guided RLcited by 23
  9. 2022Language Control Diffusion: Efficiently Scaling through Space, Time, and Taskscited by 20
  10. 2021Boosting Search Engines with Interactive Agentscited by 19
  11. 2021NeurIPS 2021 Competition IGLU: Interactive Grounded Language Understanding in a Collaborative Environmentcited by 15
  12. 2022IGLU 2022: Interactive Grounded Language Understanding in a Collaborative Environment at NeurIPS 2022cited by 14
18 more
  1. 2022MMKGR: Multi-hop Multi-modal Knowledge Graph Reasoningcited by 13
  2. 2022Perceiving the World: Question-guided Reinforcement Learning for Text-based Gamescited by 10
  3. 2023LLM Augmented Hierarchical Agentscited by 9
  4. 2021Multitasking Inhibits Semantic Driftcited by 8
  5. 2023Semantically Aligned Task Decomposition in Multi-Agent Reinforcement Learningcited by 8
  6. 2023Improving Long-Horizon Imitation Through Instruction Predictioncited by 8
  7. 2022One-Shot Learning from a Demonstration with Hierarchical Latent Languagecited by 6
  8. 2023SAGE: Bridging Semantic and Actionable Parts for GEneralizable Manipulation of Articulated Objectscited by 5
  9. 2022Teachable Reinforcement Learning via Advice Distillationcited by 2
  10. 2023Interactive Text Generationcited by 2
  11. 2021Guided Policy Search for Parameterized Skills using Adverbscited by 1
  12. 2024Instructional Text Across Disciplines: A Survey of Representations, Downstream Tasks, and Open Challenges Toward Capable AI Agentscited by 1
  13. 2025Language and Experience: A Computational Model of Social Learning in Complex Taskscited by 1
  14. 2021Procedures as Programs: Hierarchical Control of Situated Agents through Natural Languagecited by 0
  15. 2022Learning Invariable Semantical Representation from Language for Extensible Policy Generalizationcited by 0
  16. 2022Structured Exploration Through Instruction Enhancement for Object Navigationcited by 0
  17. 2025elsciRL: Integrating Language Solutions into Reinforcement Learning Problem Settingscited by 0
  18. 2025HERAKLES: Hierarchical Skill Compilation for Open-ended LLM Agentscited by 0