# Same World, Different Knowledge: When Isolated Audits Misjudge World-Model Repairs

arXiv: https://arxiv.org/abs/2609.21155 · HTML: https://mlsanity.com/paper/2609.21155/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 6 of its 27 references that are in the map.
- First posted on arXiv 17 Sep 2026.

## Placed in

- World models

## Builds on (most cited first, 6 of 6)

- [A Unified View on Solving Objective Mismatch in Model-Based Reinforcement Learning](https://mlsanity.com/paper/2310.06253/) (2023)
- [SkyJEPA: Learning Long-Horizon World Models for Zero-Shot Sim-to-Real Control of Quadrotors](https://mlsanity.com/paper/2606.23444/) (2026)
- [Adaptive Model Predictive Control for High-Accuracy Trajectory Tracking in Changing Conditions](https://mlsanity.com/paper/1807.05290/) (2018)
- [Decision-Metric Alignment in Latent World Models: Diagnostics and Action-Conditioned Objectives for MPC Planning](https://mlsanity.com/paper/2608.18746/) (2026)
- [Predicting Closed-Loop Performance of Latent World Models: Offline Checkpoint Selection for MPC and Model-Based RL Under Non-Markovian Rewards in LunarLander](https://mlsanity.com/paper/2607.01736/) (2026)
- [Fixed-time Disturbance Observer-Based MPC Robust Trajectory Tracking Control of Quadrotor](https://mlsanity.com/paper/2408.15019/) (2024)

## Abstract (arXiv)

A repair favored under an isolated input fault can be inferior when deployed modules share the faulty information. We introduce an information-interface audit for world models, distinguishing fidelity gaps, where exact inputs become estimates, from availability gaps, where inputs are missing. Fixed-weight interventions measure prediction error, input dependence, and paired closed-loop benefit, including dependencies introduced by reconstruction. In simulated quadrotor model predictive control, coupled, opposite-sign 10% mass/thrust calibration errors reduce a physics-anchored model's success from 69% to 8%; uncertainty training restores 65%. Wind reconstruction recovers control benefit but inherits calibration dependence. For a positive calibration offset, reconstruction-only corruption favors uncertainty-trained reconstruction, whereas shared corruption favors the baseline. Acceleration diagnostics reveal compensation between reconstruction bias and nominal-model error, also observed with a disturbance observer. Repair selection therefore depends on the information paths used in deployment.