LaST-HD: Learning Latent Physical Reasoning from Scalable Human Data for Robot Manipulation

arXiv 2026 39.1 method, application

TLDR

LaST-HD aligns human and robot demonstrations in a shared latent space using an action-conditioned world model for robot manipulation learning.

Reasoning

The paper introduces a novel paradigm for human-to-robot action learning by leveraging a world model to align cross-embodiment dynamics, which is a strength. However, the abstract lacks explicit details on real-world evaluation results and does not directly address several of the specified keywords beyond 'world model'.

Read-first score

Read-first score 39.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 11.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

92.1

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.92117111

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=none

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Topical relevance 29%

15.7

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

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 321.

Keyword Scores

world model

world dynamics prediction

world simulator

generative world model

interactive world model

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Extending reasoning-before-acting VLA by aligning human-hand and robot demonstrations in a shared latent reasoning space
Training an auxiliary action-conditioned world model on unpaired human-hand and robot trajectories to synthesize unified latent targets
Developing Out-of-Lab (OOL) Glove, a low-cost motion-capture glove for human-hand data collection
Progressive mixed-to-human training recipe comprising mixed human-robot co-training and human-hand online correction post-training

Methodology

LaST-HD is a human-to-robot action learning paradigm that extends reasoning-before-acting VLA by aligning human-hand and robot demonstrations in a shared latent reasoning space. It trains an auxiliary action-conditioned world model on unpaired human-hand and robot trajectories to synthesize unified latent targets, then aligns cross-embodiment representations in this forward-dynamics space. The method uses the OOL Glove for data collection and employs a progressive mixed-to-human training recipe with mixed co-training and online correction.

Key Results

LaST-HD improves generalization to novel objects, scenes, and positions using only human-hand demonstrations. With online correction, it achieves over 90% accuracy using only 20 minutes of OOL glove data.

Resources

Paper