Awesome World Model Hub Papers · Datasets · Projects
← Back to papers

Stealthy World Model Manipulation via Data Poisoning

arXiv 2026 54.5 method

TLDR

Proposes SWAAP, a two-stage data poisoning framework to stealthily manipulate learned world models in model-based RL, causing performance degradation.

Reasoning

The paper introduces a novel attack framework with clear methodology and evaluation across continuous-control tasks, demonstrating effectiveness and stealthiness. However, it focuses only on fine-tuning attacks and may not address broader training scenarios or defenses comprehensively.

Read-first score

Read-first score 54.5, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 35.

Recency 6%
100

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

Citation impact 18%
92.4

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.92422945

Methodology quality 18%
70

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

Topical relevance 29%
50

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%
30

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

Citation velocity 12%
0

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

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 377.

Keyword Scores

world model
10
model-based reinforcement learning world model
9
world dynamics prediction
8
world simulator
3
generative world model
2
interactive world model
2
video world model
1

Deep Analysis

Innovations

  • First two-stage data poisoning framework for learned world models (SWAAP)
  • First stage uses first-order bilevel optimization with a transition-gradient theorem to identify a harmful target world model that induces low-return behavior while staying close to clean dynamics
  • Second stage employs stealth-constrained gradient matching to modify fine-tuning transition targets, with a prediction-error regularizer to keep poisoned targets close to the world model's natural approximation error
  • Evaluation of attack stealthiness across three stages of the poisoning pipeline: pre-training detection, robust training during fine-tuning, and test-time monitoring of the resulting world model

Methodology

SWAAP operates in two stages: first, it identifies a harmful target world model via first-order bilevel optimization enabled by a transition-gradient theorem, ensuring the target induces low-return behavior under planning while remaining close to clean dynamics. Second, it realizes this target through stealth-constrained gradient matching, modifying only a limited fraction of fine-tuning transition targets so that training gradients steer the victim model toward the adversarial target, while a prediction-error regularizer keeps poisoned targets close to the world model's natural approximation error. The attack is evaluated on diverse continuous-control tasks against non-adaptive residual, CUSUM, and TRIM-style defenses across three detection stages.

Key Results

SWAAP causes substantial performance degradation in downstream planning while keeping poisoned transitions close to clean data and evading the evaluated non-adaptive residual/CUSUM/TRIM-style defenses, revealing a practical vulnerability in world-model adaptation pipelines.

Limitations

  • Only non-adaptive defenses (residual, CUSUM, TRIM-style) were evaluated; adaptive defenses may detect the attack
  • The attack assumes access to the fine-tuning data and the ability to modify transition targets, which may not always be feasible
  • The framework may require knowledge of the victim's world model architecture or training procedure, though not explicitly stated
  • Computational cost of the bilevel optimization in the first stage is not discussed and could be a practical limitation

Tags

data poisoningworld modelsmodel-based reinforcement learningadversarial attacksbilevel optimizationstealthy attacksLGCR