# Improving Policy Learning via Language Dynamics Distillation

arXiv: https://arxiv.org/abs/2210.00066 · HTML: https://mlsanity.com/paper/2210.00066/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 26 of its 34 references that are in the map.
- First posted on arXiv 30 Sep 2022.

## Builds on (most cited first, 26 of 26)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [ALFWorld: Aligning Text and Embodied Environments for Interactive Learning](https://mlsanity.com/paper/2010.03768/) (2020)
- [Vision-and-Language Navigation: Interpreting visually-grounded navigation instructions in real environments](https://mlsanity.com/paper/1711.07280/) (2017)
- [Exploration by Random Network Distillation](https://mlsanity.com/paper/1810.12894/) (2018)
- [IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures](https://mlsanity.com/paper/1802.01561/) (2018)
- [Recurrent World Models Facilitate Policy Evolution](https://mlsanity.com/paper/1809.01999/) (2018)
- [ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Tasks](https://mlsanity.com/paper/1912.01734/) (2019)
- [Behavioral Cloning from Observation](https://mlsanity.com/paper/1805.01954/) (2018)
- [Room-Across-Room: Multilingual Vision-and-Language Navigation with Dense Spatiotemporal Grounding](https://mlsanity.com/paper/2010.07954/) (2020)
- [Reinforcement Learning with Unsupervised Auxiliary Tasks](https://mlsanity.com/paper/1611.05397/) (2016)
- [Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration](https://mlsanity.com/paper/1802.08802/) (2018)
- [Touchdown: Natural Language Navigation and Spatial Reasoning in Visual Street Environments](https://mlsanity.com/paper/1811.12354/) (2018)
- [The NetHack Learning Environment](https://mlsanity.com/paper/2006.13760/) (2020)
- [Decoupling Strategy and Generation in Negotiation Dialogues](https://mlsanity.com/paper/1808.09637/) (2018)
- [RIDE: Rewarding Impact-Driven Exploration for Procedurally-Generated Environments](https://mlsanity.com/paper/2002.12292/) (2020)
- [Third-Person Imitation Learning](https://mlsanity.com/paper/1703.01703/) (2017)
- [Imitating Latent Policies from Observation](https://mlsanity.com/paper/1805.07914/) (2018)
- [Learning human behaviors from motion capture by adversarial imitation](https://mlsanity.com/paper/1707.02201/) (2017)
- [Language Understanding for Text-based Games Using Deep Reinforcement Learning](https://mlsanity.com/paper/1506.08941/) (2015)
- [Semantic Exploration from Language Abstractions and Pretrained Representations](https://mlsanity.com/paper/2204.05080/) (2022)
- [Imitation Learning from Observations by Minimizing Inverse Dynamics Disagreement](https://mlsanity.com/paper/1910.04417/) (2019)
- [Grounding Language to Entities and Dynamics for Generalization in Reinforcement Learning](https://mlsanity.com/paper/2101.07393/) (2021)
- [TorchBeast: A PyTorch Platform for Distributed RL](https://mlsanity.com/paper/1910.03552/) (2019)
- [Hybrid Reinforcement Learning with Expert State Sequences](https://mlsanity.com/paper/1903.04110/) (2019)
- [Internal Model from Observations for Reward Shaping](https://mlsanity.com/paper/1806.01267/) (2018)
- [SILG: The Multi-environment Symbolic Interactive Language Grounding Benchmark](https://mlsanity.com/paper/2110.10661/) (2021)

## Built on by (most cited first, 9 of 9)

- [Learning to Model the World with Language](https://mlsanity.com/paper/2308.01399/) (2023)
- [Distilling Internet-Scale Vision-Language Models into Embodied Agents](https://mlsanity.com/paper/2301.12507/) (2023)
- [Language Control Diffusion: Efficiently Scaling through Space, Time, and Tasks](https://mlsanity.com/paper/2210.15629/) (2022)
- [Semantic HELM: A Human-Readable Memory for Reinforcement Learning](https://mlsanity.com/paper/2306.09312/) (2023)
- [NetHack is Hard to Hack](https://mlsanity.com/paper/2305.19240/) (2023)
- [WebWISE: Web Interface Control and Sequential Exploration with Large Language Models](https://mlsanity.com/paper/2310.16042/) (2023)
- [Assessing the Zero-Shot Capabilities of LLMs for Action Evaluation in RL](https://mlsanity.com/paper/2409.12798/) (2024)
- [diff History for Neural Language Agents](https://mlsanity.com/paper/2312.07540/) (2023)
- [A Survey on Explainable Reinforcement Learning: Concepts, Algorithms, Challenges](https://mlsanity.com/paper/2211.06665/) (2022)

## Abstract (arXiv)

Recent work has shown that augmenting environments with language descriptions improves policy learning. However, for environments with complex language abstractions, learning how to ground language to observations is difficult due to sparse, delayed rewards. We propose Language Dynamics Distillation (LDD), which pretrains a model to predict environment dynamics given demonstrations with language descriptions, and then fine-tunes these language-aware pretrained representations via reinforcement learning (RL). In this way, the model is trained to both maximize expected reward and retain knowledge about how language relates to environment dynamics. On SILG, a benchmark of five tasks with language descriptions that evaluate distinct generalization challenges on unseen environments (NetHack, ALFWorld, RTFM, Messenger, and Touchdown), LDD outperforms tabula-rasa RL, VAE pretraining, and methods that learn from unlabeled demonstrations in inverse RL and reward shaping with pretrained experts. In our analyses, we show that language descriptions in demonstrations improve sample-efficiency and generalization across environments, and that dynamics modelling with expert demonstrations is more effective than with non-experts.