Data-Asymmetric Latent Imagination and Reranking for 3D Robotic Imitation Learning

arXiv 2026 48.9 method, application

TLDR

DALI-R uses a latent world model and reranking to improve 3D robotic imitation learning from mixed-quality trajectories, achieving 6.8% success rate improvement.

Reasoning

The paper introduces a novel framework that leverages a latent world model for imagined rollouts and a task completion scorer for reranking, effectively using suboptimal demonstrations. Strengths include clear methodology and empirical gains on two benchmarks, but weaknesses are limited evaluation to simulated environments and modest inference overhead.

Read-first score

Read-first score 48.9, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 34.

Recency 6%

100

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

Citation impact 18%

72.7

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.7268605

Methodology quality 18%

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

Topical relevance 29%

48.6

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 18%

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

Citation velocity 12%

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

Field roles

FrontierBridge

Rank sensitivity

Stability: volatile; rank range: 260.

Keyword Scores

world model

world dynamics prediction

generative world model

world simulator

model-based reinforcement learning world model

interactive world model

video world model

Deep Analysis

Innovations

Data-Asymmetric Latent Imagination and Reranking (DALI-R) framework for 3D robotic imitation learning from mixed-quality trajectories
Latent World Model over 3D point clouds for imagined rollouts
Task Completion Scorer that reranks candidate action chunks
Integration with both diffusion and efficient flow-matching policies

Methodology

DALI-R learns a Latent World Model over 3D point clouds to generate imagined rollouts and a Task Completion Scorer to rerank candidate action chunks. It is instantiated with diffusion and flow-matching policies and evaluated on Adroit and MetaWorld benchmarks.

Key Results

Across two 3D base policies, DALI-R achieves an average 6.8% improvement in success rate while incurring less than 0.7× additional inference overhead.

Resources

Paper