Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents (2022-01-18T00:00:00.000000Z)

TL;DR

This paper investigates the possibility of grounding high-level tasks, expressed in natural language, to a chosen set of actionable steps and proposes a procedure that conditions on existing demonstrations and semantically translates the plans to admissible actions.

Abstract

Can world knowledge learned by large language models (LLMs) be used to act in interactive environments? In this paper, we investigate the possibility of grounding high-level tasks, expressed in natural language (e.g."make breakfast"), to a chosen set of actionable steps (e.g."open fridge"). While prior work focused on learning from explicit step-by-step examples of how to act, we surprisingly find that if pre-trained LMs are large enough and prompted appropriately, they can effectively decompose high-level tasks into mid-level plans without any further training. However, the plans produced naively by LLMs often cannot map precisely to admissible actions. We propose a procedure that conditions on existing demonstrations and semantically translates the plans to admissible actions. Our evaluation in the recent VirtualHome environment shows that the resulting method substantially improves executability over the LLM baseline. The conducted human evaluation reveals a trade-off between executability and correctness but shows a promising sign towards extracting actionable knowledge from language models. Website at https://huangwl18.github.io/language-planner

Authors

P. Abbeel

35 papers

Igor Mordatch

6 papers

Deepak Pathak

2 papers

TL;DR

Abstract

Authors

References55 items

Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models

Pre-Trained Language Models for Interactive Decision-Making

Synchromesh: Reliable code generation from pre-trained language models

Learning To Retrieve Prompts for In-Context Learning

Training Verifiers to Solve Math Word Problems

Skill Induction and Planning with Latent Language

Generate & Rank: A Multi-task Framework for Math Word Problems

On the Opportunities and Risks of Foundation Models

Embodied BERT: A Transformer Model for Embodied, Language-guided Visual Task Completion

Evaluating Large Language Models Trained on Code

Multimodal Few-Shot Learning with Frozen Language Models

Learning From Stories: Using Crowdsourced Narratives to Train Virtual Agents

Implicit Representations of Meaning in Neural Language Models

Measuring Coding Challenge Competence With APPS

ELLA: Exploration through Learned Language Abstraction

Pretrained Transformers as Universal Computation Engines

What Makes Good In-Context Examples for GPT-3?

Making Pre-trained Language Models Better Few-shot Learners

ALFWorld: Aligning Text and Embodied Environments for Interactive Learning

Visually-Grounded Planning without Vision: Language Models Infer Detailed Plans from High-level Instructions

Language Models are Few-Shot Learners

Human Instruction-Following with Deep Reinforcement Learning via Transfer-Learning from Text

Language Conditioned Imitation Learning Over Unstructured Data

Grounding Language in Play

Probing Text Models for Common Ground with Visual Representations

Improving Vision-and-Language Navigation with Image-Text Pairs from the Web

How Much Knowledge Can You Pack into the Parameters of a Language Model?

oLMpics-On What Language Model Pre-training Captures

ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Tasks

How Can We Know What Language Models Know?

Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer

HuggingFace's Transformers: State-of-the-art Natural Language Processing

Language Models as Knowledge Bases?

Commonsense Knowledge Mining from Pretrained Models

Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

RoBERTa: A Robustly Optimized BERT Pretraining Approach

Synthesizing Environment-Aware Activities via Activity Sketches

The Curious Case of Neural Text Degeneration

Habitat: A Platform for Embodied AI Research

From Language to Goals: Inverse Reinforcement Learning for Vision-Based Instruction Following

Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation

VirtualHome: Simulating Household Activities Via Programs

Speaker-Follower Models for Vision-and-Language Navigation

AI2-THOR: An Interactive 3D Environment for Visual AI

Vision-and-Language Navigation: Interpreting Visually-Grounded Navigation Instructions in Real Environments

SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Crosslingual Focused Evaluation

Attention is All you Need

Environment-Driven Lexicon Induction for High-Level Instructions

GloVe: Global Vectors for Word Representation

Tell me Dave: Context-sensitive grounding of natural language to manipulation instructions

Weakly Supervised Learning of Semantic Parsers for Mapping Instructions to Actions

Understanding and executing instructions for everyday manipulation tasks from the World Wide Web

Long Short-Term Memory

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Language Models are Unsupervised Multitask Learners

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names