Learning to Act without Actions

arXiv 2023 39.7 method

TLDR

Introduces LAPO, a method to recover latent actions from videos, enabling training of policies and world models without action labels, and pre-training on web videos.

Reasoning

The paper presents a novel method for extracting latent action information from videos, which is a significant step toward pre-training RL agents on unlabeled data. Strengths include the innovative approach and potential for scaling to web-scale video. Weaknesses are that experiments are only on procedurally-generated environments, and no real-world validation is mentioned.

Read-first score

Read-first score 39.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 29.

Recency 8%

65.1

Uses a gentle age decay so recent papers surface without erasing older foundations. 2023

Topical relevance 42%

41.4

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=dataset

Field roles

Candidate

Rank sensitivity

Stability: volatile; rank range: 253.

Keyword Scores

world model

model-based reinforcement learning world model

video world model

world dynamics prediction

world simulator

generative world model

interactive world model

Resources

Paper