A Control Theory of Predictability in Latent World Models

• URL: http://arxiv.org/abs/2607.10362v1
• Date: Sat, 11 Jul 2026 15:42:26 GMT
• Title: A Control Theory of Predictability in Latent World Models
• Abstract summary: Current practice adopts the prediction error, the single- or multi-step rollout loss on held-out data, as the training and model-selection objective.<n>We show that this assumption is unreliable for a structural reason.<n>A planner does not query the model on the training distribution but on the states that its candidate actions reach, which generally leave the data manifold.
• Score: 39.22646915328561
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Latent world models are trained to predict future states in a learned representation and are then deployed inside a planner that selects actions by simulating them forward. Current practice adopts the prediction error, the single- or multi-step rollout loss on held-out data, as the training and model-selection objective, on the assumption that a lower prediction error yields better control. We show that this assumption is unreliable for a structural reason: a planner does not query the model on the training distribution but on the states that its candidate actions reach, which generally leave the data manifold, so an error averaged over the data cannot by itself govern control. We therefore reframe the objective as the discrepancy between the predicted and the true plan-cost at the plan the planner commits to, and prove that the planner's suboptimality is bounded by twice this discrepancy, whereas the data-averaged prediction error neither bounds nor tracks it. Under a linear-control premise the discrepancy separates into two terms. The first is a small on-manifold residual, on which the predicted and true dynamics agree and which a spectral tax prices through the non-normality of the latent transition operator. The second is an off-manifold divergence, on which an action carries the state off the manifold and the two dynamics diverge; this divergence is the binding term and is bounded by no data-averaged error. Synthetic operators confirm the pricing formulas, and latent model-predictive control experiments confirm the decoupling: across seeds, the single-step validation error is essentially uncorrelated with control success, whereas a fidelity score on the planner-reachable measure tracks it.

Related papers

• Robust Ambiguity Detection (RAD) From Model- and Feature-Space Consistency [6.07595146080737]
  Robust Ambiguity Detection (RAD) framework is advanced for quantifying predictive ambiguity using two complementary metrics.<n>RAD is evaluated on synthetic datasets with systematically controlled overlap, as well as several real-world datasets.
  arXiv  Detail & Related papers  (2026-08-12T01:13:33Z)
• MoP-JEPA: Hard-Assigned Predictor Mixtures for Stochastic JEPA World Models [14.76812870029721]
  MoP-JEPA produces a finite candidate set in one pass.<n>On held-out OGBench transitions, graph search with single-output predictors succeeds on $0.02$--$0.09$ of queries, whereas MoP-JEPA reaches $0.85$.
  arXiv  Detail & Related papers  (2026-07-06T15:47:37Z)
• A Path-Space Formulation of Prediction in World Models: From a Single Action to Prediction, Planning, and Irreversibility [0.0]
  We propose a path-space formulation of prediction in AI world models.<n>We decompose latent dynamics into reversible and irreversible components.<n>We find that attention asymmetry is acquired during training in proportion to the irreversibility of the data.
  arXiv  Detail & Related papers  (2026-06-27T06:00:09Z)
• Constrained Auto-Bidding via Generative Response Modeling [4.3095605379591255]
  Auto-bidding systems aim to maximize advertiser value over long horizons under budget constraints and ratio targets such as cost-per-acquisition.<n>Existing approaches face distinct limitations: control-based pacing reacts to deviations but cannot anticipate future conditions, while RL and generative methods fold constraints into reward signals, obscuring violations and degrading under distribution shift.<n>We shift the learning target from actions to responses with the Generative Response Model (GRM), a history-conditioned sequence model that jointly predicts future traffic volume and horizon-aggregate cost/value curves as functions of a single bid multiplier.
  arXiv  Detail & Related papers  (2026-05-27T01:06:54Z)
• Auditing and Fixing Economic Validity in Tabular Foundation Models for Discrete Choice [12.709654635425442]
  Tabular foundation models achieve strong accuracy on choice prediction tasks.<n>But their predictions often violate the economic logic those tasks require.<n>We propose a two-stage adapter that embeds foundation model predictions within a utility-maximization framework.
  arXiv  Detail & Related papers  (2026-05-26T05:13:20Z)
• Information Fidelity in Tool-Using LLM Agents: A Martingale Analysis of the Model Context Protocol [69.11739400975445]
  We introduce the first theoretical framework for analyzing error accumulation in Model Context Protocol (MCP) agents.<n>We show that cumulative distortion exhibits linear growth and high-probability deviations bounded by $O(sqrtT)$.<n>Key findings include: semantic weighting reduces distortion by 80%, and periodic re-grounding approximately every 9 steps suffices for error control.
  arXiv  Detail & Related papers  (2026-02-10T21:08:53Z)
• Robust Planning for Autonomous Driving via Mixed Adversarial Diffusion Predictions [51.88828388720111]
  We describe a robust planning method for autonomous driving that mixes normal and adversarial agent predictions.<n>We generate a distribution of adversarial predictions by biasing the diffusion model at test time to generate predictions that are likely to collide with a candidate plan.<n>We show the effectiveness of our method on single-agent and multi-agent jaywalking scenarios as well as a red light violation scenario.
  arXiv  Detail & Related papers  (2025-05-18T09:44:57Z)
• CUQDS: Conformal Uncertainty Quantification under Distribution Shift for Trajectory Prediction [6.029850098632435]
  Trajectory prediction models can infer both finite future trajectories and their associated uncertainties in an online setting.<n>We propose the Conformal Uncertainty Quantification under Distribution Shift framework, CUQDS, to quantify the uncertainty of the predicted trajectories.
  arXiv  Detail & Related papers  (2024-06-17T21:25:36Z)
• Rejection via Learning Density Ratios [50.91522897152437]
  Classification with rejection emerges as a learning paradigm which allows models to abstain from making predictions.<n>We propose a different distributional perspective, where we seek to find an idealized data distribution which maximizes a pretrained model's performance.<n>Our framework is tested empirically over clean and noisy datasets.
  arXiv  Detail & Related papers  (2024-05-29T01:32:17Z)
• Neuro-Symbolic Entropy Regularization [78.16196949641079]
  In structured prediction, the goal is to jointly predict many output variables that together encode a structured object. One approach -- entropy regularization -- posits that decision boundaries should lie in low-probability regions. We propose a loss, neuro-symbolic entropy regularization, that encourages the model to confidently predict a valid object.
  arXiv  Detail & Related papers  (2022-01-25T06:23:10Z)
• Test-time Collective Prediction [73.74982509510961]
  Multiple parties in machine learning want to jointly make predictions on future test points. Agents wish to benefit from the collective expertise of the full set of agents, but may not be willing to release their data or model parameters. We explore a decentralized mechanism to make collective predictions at test time, leveraging each agent's pre-trained model.
  arXiv  Detail & Related papers  (2021-06-22T18:29:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.