mlsanityData to 14 Oct 2026

World Models in Pieces: Structural Certification for General Agents

arXiv 2606.24842 ↗First posted 23 Jun 2026cs.AIcited by 0

On the map: World models → World models for language agents

Abstract (arXiv)

In the big-world regime, agents cannot be universally capable and their ability is inevitably specialized across a world model in pieces. Consequently, standard uniform guarantees fail to distinguish between the understanding of critical bottlenecks and irrelevant failures. We first formalize this limitation by proving that general agents are not universal, rendering standard worst-case analysis uninformative. To overcome this, we introduce structural certification, a transition-local framework that maps bounded goal-conditioned performance to entry-wise guarantees on the agent's internal world model. Our main contribution is constructive. We provide algorithms that filter specific transitions using deep compositional goals and prove that a general agent on these goals has a structural world model with a O(1/n) + O(δ) error bound. Conversely, this bound is tight in the small-δ regime, whose existence is explicitly guaranteed by our certification. These results enable the certifiable deployment of general agents by localizing the specific transitions where long-horizon planning is reliable.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in World models for language agents, one of 11 directions in World models. Builds on 20 of its 86 references that are in the map. First posted on arXiv 23 Jun 2026.

Builds on

20 of 86 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025From Word to World: Can Large Language Models be Implicit Text-based World Models?21 Dec
  2. 2024Understanding World or Predicting Future? A Comprehensive Survey of World Models21 Nov

Other references · 18

  1. 2020Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problemscited by 1,920
  2. 2019When to Trust Your Model: Model-Based Policy Optimizationcited by 773
  3. 2024WebVoyager: Building an End-to-End Web Agent with Large Multimodal Modelscited by 438
  4. 2025Large Language Model Agent: A Survey on Methodology, Applications and Challengescited by 205
  5. 2022Goal-Conditioned Reinforcement Learning: Problems and Solutionscited by 158
  6. 2024AgentGym: Evolving Large Language Model-based Agents across Diverse Environmentscited by 120
  7. 2025OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agentscited by 38
  8. 2025Automatic Failure Attribution and Critical Step Prediction Method for Multi-Agent Systems Based on Causal Inferencecited by 18
  9. 2024Dynamic Model Predictive Shielding for Provably Safe Reinforcement Learningcited by 15
  10. 2025Rethinking the Foundations for Continual Reinforcement Learningcited by 15
  11. 2024A Comprehensive Survey of Reinforcement Learning: From Algorithms to Practical Challengescited by 13
  12. 2024Simplifying Latent Dynamics with Softly State-Invariant World Modelscited by 10
6 more
  1. 2024Probabilistic Subgoal Representations for Hierarchical Reinforcement learningcited by 8
  2. 2023Provably Efficient Offline Goal-Conditioned Reinforcement Learning with General Function Approximation and Single-Policy Concentrabilitycited by 5
  3. 2024Goal-Conditioned Reinforcement Learning from Sub-Optimal Data on Metric Spacescited by 3
  4. 2025Look Before Leap: Look-Ahead Planning with Uncertainty in Reinforcement Learningcited by 3
  5. 2024Subjective Causalitycited by 1
  6. 2024Partial Identifiability in Inverse Reinforcement Learning For Agents With Non-Exponential Discountingcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025General agents contain world models
  2. 2026Dual-Frontier: When Can an Agent Trust Its World Model?
  3. 2026Foundation World Models for Agents that Learn, Verify, and Adapt Reliably Beyond Static Environments
  4. 2026TRW: TRACE-RealWorld---An Auditable Consistency Contract for World Models as Materialized Views
  5. 2024WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment
  6. 2026Containment Verification: AI Safety Guarantees Independent of Alignment
  7. 2026Agent vs. Parametric World Models: Hybrid Planning for Reliable Language Agents
  8. 2026General Agents Contain World Models, even under Partial Observability and Stochasticity