arXiv subjects

Donna Vakalis

Publications and source records attributed to Donna Vakalis.

4 recordsLinked to original sources

arXiv2026

The Intervention Gap in Latent World Models

Planning-time intervention fidelity is a distinct, measurable property of a learned world model: whether the model's own open-loop transitions move task variables the way matched environment interventions do. In the settings we test, it is neither revealed by reward fit nor ensured by task-anchored training. Across released TD-MPC2 checkpoint sizes, episode return falls as an operator-error diagnostic on task observables grows, while reward-prediction error stays small and nearly flat, and a self-supervised world model trained without task signal preserves the same operator substantially better than a task-anchored model on the shared task. A capture-gated matched-intervention audit then localizes what fails. On Cheetah, three LeWorldModel checkpoints capture the current task query and support decodable real intervention effects; however, their imagined five-step effects are worse than predicting no effect and worse than an environment-endpoint oracle. The failure is task-direction rotation with excess gain, not feature collapse. This severe pattern is conditional: five PreJEPA seeds retain an oracle-relative deficit without it, Finger Spin experiments extend the deficit beyond locomotion with heterogeneous severity across seeds, and shared-bank effect geometry is both candidate- and support-dependent. We also test practice-side questions. In DreamerV3 the posterior distribution, not its sample, carries the current query; ensemble disagreement ranks error only near training support; and a frozen support-aware score degrades held-out error ranking in both tested transfer directions while native disagreement remains informative in both. We conclude that intervention fidelity must be audited directly, capture-first, on the model's native interface.

cs.LG↗

arXiv2026

What a World Model Represents Is Three Questions

World models learn task-relevant information through many routes: observation reconstruction, recurrent state, temporal filtering, and explicit task supervision. Different routes can make different variables available. The same variable can also be available through several routes at once. When it is, looking at which route would increase the training loss most if removed does not tell you which route the model actually uses. The questions are reachability, whether a training signal can identify a task-relevant direction; admission, whether that direction is recoverable from the latent; and assignment, which eligible route carries it. We test them in environments with a known set of required coordinates. A direction cannot enter the latent unless some training signal can identify it. Reconstruction, recurrence, or filtering may already recover some of those coordinates; a reward or value head then has no residual direction to admit. For what remains, how many independent predictions the target supplies is how many coordinates install: one through four independent predictions admit one through four directions, including through the value head. Reachability is not admission: a temporal second-moment coefficient can remain absent under next-token prediction when accumulating it is a fraction of a percent of that loss, and a head that predicts the coefficient restores it. Assignment is a different test. Two routes that each carry the same variable when trained alone do not swap the carrier when we reverse which is more costly to remove. A recurrent model trained on a transformer's recorded sequences shows the same pattern. Near the point where the competing route is beginning to clear the probe threshold, independent training runs disagree. What a world model represents is therefore three questions: what information is reachable, what supervision admits, and which competing route carries it.

cs.LG↗

arXiv2026

Operator-on-F complements value-equivalence: a planning-time diagnostic for latent world models

World-model evaluation for model-based reinforcement learning typically asks whether the learned model predicts reward and value well, which can leave planning-relevant errors in the model's latent rollouts unmeasured. We introduce a complementary diagnostic, operator-on-F, that compares a model's k-step latent pushforward to the environment's on an observable subset F, using the model's own predictor. On a TD-MPC2 size sweep over cheetah-run, reward-prediction error stays within [0.028, 0.091] for every model size - only about 3x variation - so an unnormalized reward-fit check has narrow resolution to distinguish them; the (unnormalized) Bellman residual and reward error themselves have weak relationships with return (Spearman -0.10 and -0.30). Operator error spans 0.28 to 2.62 over the same sizes. At 317M the operator error is 2.62 - an order of magnitude above the 0.28-0.36 cluster - and the planning return collapses to 0.9, while reward-prediction error (0.091) is the highest of the five but stays within the same small [0.028, 0.091] range as the rest of the sweep. The rank correlation between operator error and return loss is -0.90 (anchor-bootstrap 95% CI [-0.90, -0.70] at n=5 sizes; leave-one-out removal of any single size leaves it at -0.80 or stronger). The operator also returns informative, architecture-discriminating estimates in a cross-architecture comparison between TD-MPC2 and a pure-SSL latent world model. The operator diagnostic complements value-equivalence rather than replacing it.

cs.LG↗

arXiv2026

In-Context Reinforcement Learning through Bayesian Fusion of Context and Value Prior

In-context reinforcement learning (ICRL) promises fast adaptation to unseen environments without parameter updates, but current methods either cannot improve beyond the training distribution or require near-optimal data, limiting practical adoption. We introduce SPICE, a Bayesian ICRL method that learns a prior over Q-values via deep ensemble and updates this prior at test-time using in-context information through Bayesian updates. To recover from poor priors resulting from training on sub-optimal data, our online inference follows an Upper-Confidence Bound rule that favours exploration and adaptation. We prove that SPICE achieves regret-optimal behaviour in both stochastic bandits and finite-horizon MDPs, even when pretrained only on suboptimal trajectories. We validate these findings empirically across bandit and control benchmarks. SPICE achieves near-optimal decisions on unseen tasks, substantially reduces regret compared to prior ICRL and meta-RL approaches while rapidly adapting to unseen tasks and remaining robust under distribution shift.

cs.LG↗