Video World Models · 2026-10-07
How Much Planning Is Enough? Reducing Search and Computation in World-Model Planning
Changbai Li, Sirui Li, Yichen Yang, Tongfei Chen, Zichao Feng, Shuwei Shao, Huobin Tan
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 analyze why visual world-model planning can waste computation: fixed “Full” budgets can evaluate excess candidates after outcome sufficiency, and iterative planners repeatedly encode solve-static observation/goal representations inside the refinement loop. They propose SufficientPlan with PSBC (certify a reduced budget using paired closed-loop outcomes) and SCR (reuse solve-static encodings across CEM iterations) without retraining or planner changes.
Why it matters
Decision-time action search in visual world models can be dominated by unnecessary candidates and redundant per-iteration computation. The proposed deployment approach targets these inefficiencies directly: PSBC selects model–task-specific budgets using paired outcome evidence (not action agreement), and SCR reduces latency by caching invariant representations during planning solves while preserving candidate-dependent planning and selected actions.
Method
- Paired Sequential Budget Certification (PSBC): sequentially proposes smaller budgets and certifies the first budget that stays within a predefined Full-performance tolerance using paired closed-loop outcomes on the same episodes.
- Static-Context Reuse (SCR): caches solve-static observation-history and goal encodings once per planning solve and reuses them across all refinement iterations while keeping candidate-dependent planning unchanged.
- Plug-and-play deployment: applies PSBC for budget selection on a frozen development set and runs the original pretrained world-model planner with the certified budget and SCR at test time.
Limitation
PSBC calibrates the planning budget on a frozen development distribution; substantial changes in the model, task dynamics, or goal distribution may require recalibration. SCR’s speedup depends on the fraction of runtime attributable to solve-static computation; when candidate-dependent rollout and cost evaluation dominate, latency reduction may be limited.
Abstract (from arXiv)
Visual world models enable goal-directed control through decision-time action search, but their deployment efficiency is often limited by conservatively large planning budgets. We show that competitive task performance can be achieved without agreement with the Full-budget action, that sufficient budgets vary across model--task pairs, and that iterative planners repeatedly encode solve-invariant context. To address these inefficiencies, we propose {SufficientPlan}, a simple deployment framework that requires no modification to pretrained world models or planner updates. Its {Paired Sequential Budget Certification (PSBC)} component uses paired closed-loop evidence to search for and certify a reduced model--task-specific budget within a predefined Full-performance tolerance. Its {Static-Context Reuse (SCR)} component caches observation and goal representations across search iterations while preserving candidate-dependent planning and selected actions. Experiments across multiple world-model backbones and visual-control tasks show that SufficientPlan substantially reduces search budgets and planning latency while maintaining competitive control performance.
Related papers
- Sparse Planning in Visual World Models via Cost Gradients
- The Planning Limits of Latent World Models
- What Must a World Model Distinguish for Planning?
- Beyond a single latent space: a dual-latent world model for long-horizon planning
- Plans Are Not Executions: Adapting Frozen World-Model Planners to Controller-Induced Execution Shift
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-06 · Indexed by Pub-AI: 2026-10-07 UTC
How Much Planning Is Enough? Reducing Search and Computation in World-Model Planning
Preprint posted: 2026-10-06 · Original record ↗
- arxiv
- 2610.08350
Explicit source versions (1)
- v1 · 2026-10-06
- Preprint available
Source date: 2026-10-06 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Visual world models enable goal-directed control through decision-time action search, but their deployment efficiency is often limited by conservatively large planning budgets. We show that competitive task performance can be achieved without agreement with the Full-budget action, that sufficient budgets vary across model--task pairs, and that iterative planners repeatedly encode solve-invariant context. To address these inefficiencies, we propose {SufficientPlan}, a simple deployment framework that requires no modification to pretrained world models or planner updates. Its {Paired Sequential Budget Certification (PSBC)} component uses paired closed-loop evidence to search for and certify a reduced model--task-specific budget within a predefined Full-performance tolerance. Its {Static-Context Reuse (SCR)} component caches observation and goal representations across search iterations while preserving candidate-dependent planning and selected actions. Experiments across multiple world-model backbones and visual-control tasks show that SufficientPlan substantially reduces search budgets and planning latency while maintaining competitive control performance.