Latent Dynamics · 2026-10-14
Relational Learning as World Modeling
Zheyu Zhang, Gjergji Kasneci
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors introduce TabWM and WorldRel to learn relational dynamics in-context. TabWM predicts the sparse structured edits caused by an action in an unseen relational world using observed executions as context, with no parameter updates. Across 512 held-out worlds, complete-edit F1 rises from 0.016 with no context to 0.685 with 64 executions. Recursive edit predictions enable model-based planning with broader search improving realized return.
Why it matters
The paper frames action-conditioned transition modeling for relational systems as learning to predict explicit, composable state edits rather than hidden dynamics. It shows that, even when identical observable states and actions yield different consequences across worlds, behavioral context can resolve ambiguity. The edit-based predictions can be applied recursively for imagining futures and comparing action sequences.
Method
- WorldRel generates executable relational worlds with hidden transition mechanisms; TabWM is trained across worlds to use observed transitions as behavioral context.
- TabWM encodes each execution using action-centered neighborhood information and predicts the consequence of a query action as an unordered set of explicit structured edits (operation, relational location, and required values/contents).
- Predicted edits are applied recursively to roll candidate actions forward for model-based planning; broader search over predicted futures improves realized return.
Abstract (from arXiv)
Relational learning typically predicts from a database state, but decision making asks a different question: how will that state change under an action? In operational systems, the answer depends on system-specific dynamics that may not be visible in the database itself. We view each such system as a relational "world" and ask whether a model can infer how it behaves from observed executions. We introduce TabWM (Tabular World Model) and WorldRel, a generator of executable relational worlds for pretraining. Given executions from an unseen world, TabWM uses them as context to predict the sparse structured edits caused by a new action without parameter updates. Across 512 held-out worlds, complete-edit F1 rises from 0.016 with no observed executions to 0.685 with 64, where a correct edit must match its operation, relational location, and value. Because these predictions are explicit relational edits, they can be applied recursively to imagine candidate futures and compare actions before they are taken. Broader search over these predicted futures improves the realized return of selected action sequences. Together, these results suggest that relational dynamics can be learned in context, broadening relational learning from modeling system state to modeling how that state changes under action.
Related papers
- Inside-JEPA: From Prediction to Relational Grounding in Latent World Models
- From Single-Actor Environments to Multi-Actor Event Worlds: Can LLMs Be Event World Models?
- AnyGoal: On World Model Generalization Beyond Expert Rollouts
- What Must a World Model Distinguish for Planning?
- ALDER: Discovering the Laws of a World by Acting in It
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-12 · Indexed by Pub-AI: 2026-10-14 UTC
Relational Learning as World Modeling
Preprint posted: 2026-10-12 · Original record ↗
- arxiv
- 2610.16763
Explicit source versions (1)
- v1 · 2026-10-12
- Preprint available
Source date: 2026-10-12 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Relational learning typically predicts from a database state, but decision making asks a different question: how will that state change under an action? In operational systems, the answer depends on system-specific dynamics that may not be visible in the database itself. We view each such system as a relational "world" and ask whether a model can infer how it behaves from observed executions. We introduce TabWM (Tabular World Model) and WorldRel, a generator of executable relational worlds for pretraining. Given executions from an unseen world, TabWM uses them as context to predict the sparse structured edits caused by a new action without parameter updates. Across 512 held-out worlds, complete-edit F1 rises from 0.016 with no observed executions to 0.685 with 64, where a correct edit must match its operation, relational location, and value. Because these predictions are explicit relational edits, they can be applied recursively to imagine candidate futures and compare actions before they are taken. Broader search over these predicted futures improves the realized return of selected action sequences. Together, these results suggest that relational dynamics can be learned in context, broadening relational learning from modeling system state to modeling how that state changes under action.
