# Paired Exact-Reset Evaluation of a Prediction-Derived Medium-to-Full World-Model Cascade

arXiv: https://arxiv.org/abs/2608.14650 · HTML: https://mlsanity.com/paper/2608.14650/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Embodied and robot world models, one of 11 directions in World models.
- Builds on 18 of its 25 references that are in the map.
- First posted on arXiv 30 Jul 2026.

## Placed in

- World models / [Embodied and robot world models](https://mlsanity.com/wm/embodied-and-robot-world-models/)

## Builds on (most cited first, 18 of 18)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning](https://mlsanity.com/paper/2411.04983/) (2024)
- [Fast-WAM: Do World Action Models Need Test-time Future Imagination?](https://mlsanity.com/paper/2603.16666/) (2026)
- [Consistent Estimators for Learning to Defer to an Expert](https://mlsanity.com/paper/2006.01862/) (2020)
- [Melding the Data-Decisions Pipeline: Decision-Focused Learning for Combinatorial Optimization](https://mlsanity.com/paper/1809.05504/) (2018)
- [Adaptive Neural Networks for Efficient Inference](https://mlsanity.com/paper/1702.07811/) (2017)
- [Calibrating Predictions to Decisions: A Novel Approach to Multi-Class Calibration](https://mlsanity.com/paper/2107.05719/) (2021)
- [When Does Confidence-Based Cascade Deferral Suffice?](https://mlsanity.com/paper/2307.02764/) (2023)
- [AHA-WAM:Asynchronous Horizon-Adaptive World-Action Modeling with Observation-Guided Context Routing](https://mlsanity.com/paper/2606.09811/) (2026)
- [When to Trust Imagination: Adaptive Action Execution for World Action Models](https://mlsanity.com/paper/2605.06222/) (2026)
- [Right Decisions from Wrong Predictions: A Mechanism Design Alternative to Individual Calibration](https://mlsanity.com/paper/2011.07476/) (2020)
- [SANTS: A State-Adaptive Scheduler for World Action Models](https://mlsanity.com/paper/2605.27947/) (2026)
- [Test-Time Scaling for World Action Models via Zero-Shot Geometric Evaluation](https://mlsanity.com/paper/2607.17454/) (2026)
- [ACID: Action Consistency via Inverse Dynamics for Planning with World Models](https://mlsanity.com/paper/2607.02403/) (2026)
- [Le Cam meets LeCun: Deficiency and Generic Feature Learning](https://mlsanity.com/paper/1402.4884/) (2014)
- [The Variational Deficiency Bottleneck](https://mlsanity.com/paper/1810.11677/) (2018)
- [Calibrating conditional risk](https://mlsanity.com/paper/2604.20409/) (2026)
- [Adaptive Compute in Latent World Models: When Depth Helps, Hurts, or Doesn't Matter](https://mlsanity.com/paper/2607.10203/) (2026)

## Abstract (arXiv)

Existing adaptive-inference and world-action-model systems use cheap-stage outputs or predicted futures to allocate additional computation. We study a narrower question: under paired exact-reset physical outcomes, can a Medium-derived interface predict when switching to a separately frozen Full predictor improves task-specific decision loss enough to justify sequential overhead? Our contribution is a paired evaluation and audit protocol, not a new generic routing rule: all candidate actions are executed from the same reset state, Medium and Full act on the same candidate set and task, and their paired physical-loss difference defines the routing target. On a fresh PushT bank (V106; 1,600 states, 39 tasks, three checkpoint pairs), a frozen prediction-interface router lowers overhead-inclusive decision cost relative to standalone Medium, standalone Full, and a latency-advantaged task-only router. We then prospectively seal a second 1,600-state PushT confirmation (V107) against a stronger current-state control using the task, a dimension-matched projection of current DINO features, and all five candidate actions, with no DINO encoder latency charged. The prediction interface lowers priced physical decision cost by 0.002549 (state-clustered 95% interval [-0.002867, -0.002238]; one-sided 95% upper bound -0.002286), with negative effects for all three checkpoint pairs. A controlled-PyBullet audit independently supports a composite task-prediction-regime router. The sequential router remains slower than fixed policies, and its advantage is restricted to low compute prices. The evidence supports incremental routing information in the tested prediction interface beyond one deliberately favoured current-DINO control, but not causal sufficiency, compute saving, closed-loop value, or cross-family generality.