Plans Are Not Executions: Adapting Frozen World-Model Planners to Controller-Induced Execution Shift

arXiv 2026 33.6 method

TLDR

Proposes ExecutionBridge to adapt frozen world-model planners to controller-induced execution shift, improving planning across unseen controllers on NAVSIM.

Reasoning

The paper clearly identifies a novel planning-execution gap and proposes a lightweight adaptation that keeps the world model frozen, with empirical validation on NAVSIM. Weaknesses include limited scope (simulation benchmark) and no discussion of alternative methods beyond execution substitution.

Read-first score

Read-first score 33.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 40.

Recency 5%

100

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

Methodology quality 15%

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

Topical relevance 25%

57.1

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 15%

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

Graph prestige 15%

5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 10%

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

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 298.

Keyword Scores

world model

world dynamics prediction

world simulator

model-based reinforcement learning world model

generative world model

interactive world model

video world model

Deep Analysis

Innovations

Identifies a planning-execution gap caused by deploying a frozen world-model planner with a different downstream controller than during training.
Shows that better execution fidelity does not necessarily improve decisions: directly substituting predicted executed trajectories yields only a 0.048 PDMS improvement across 10 unseen controller configurations.
Introduces ExecutionBridge, which infers controller-specific execution characteristics from planned-executed trajectory pairs and adapts the native ego latent rollout while keeping the planned trajectory unchanged.
Keeps the planner, world model, and reward head frozen while adapting to controller-induced execution shift.

Methodology

ExecutionBridge infers controller-specific execution characteristics from planned-executed trajectory pairs and uses them to adapt the native ego latent rollout, leaving the original planned trajectory, planner, world model, and reward head frozen. It is evaluated on NAVSIM across unseen controller configurations and multiple frozen world-model planners, and compared against direct substitution of predicted execution trajectories.

Key Results

Directly substituting predicted execution trajectories yields only a 0.048 PDMS improvement across 10 unseen controller configurations. ExecutionBridge consistently improves planning across unseen controller configurations and multiple frozen world-model planners, with larger gains in challenging driving scenes where execution mismatch more strongly affects candidate evaluation.

Resources

Paper