Driving · 2026-11-02
Agentic World Modeling for Agent Test-time Scaling without Environment Replica
Wendi Li, Huseyin A. Inan, Wei-Ning Chen, Yanzhi Chen, Chetan Bansal, Saravan Rajmohan, Sharon Li
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 propose Agentic World Modeling (AWM) to enable test-time scaling (TTS) in shared, non-replicable environments. AWM reformulates next-observation prediction as multi-turn evidence acquisition using safe read-only queries to the real environment and executable computation to construct the predicted observation (without mutating state). Integrated with trajectory-level TTS, AWM enables parallel exploration and improves task success over single-shot and trained world-model baselines.
Why it matters
The paper addresses a practical mismatch: standard agentic TTS assumes candidate trajectories can run in parallel, but shared workspaces make replication unsafe or infeasible. By grounding simulation via adaptive, read-only interaction and executable reconstruction, AWM aims to make speculative next-observation predictions reliable enough to support safe, scalable test-time scaling under partial observability.
Method
- Reformulates next-observation prediction as an interactive multi-turn evidence-acquisition process using safe read-only interfaces and terminal prediction actions.
- Uses hierarchical subagent delegation to keep context bounded, and a read–compute interface: read evidence via tools, then use restricted Python computation to deterministically transform retrieved content into the predicted observation.
- Claims a theoretical advantage for adaptive evidence acquisition over non-adaptive context construction, yielding an exponential adaptivity gap under a dependent lookup structure.
Abstract (from arXiv)
Test-time scaling (TTS) improves large language model agents by sampling and selecting among multiple candidate trajectories, but common realizations assume that these trajectories can be executed in parallel. This assumption breaks in realistic deployments, where enterprise workspaces and cloud services are shared and persistent, costly or infeasible to replicate, and actions may introduce irreversible side effects or cross-trajectory interference. Existing generative world models avoid direct execution, but typically compress environment state into a single prompt, limiting their ability to reliably simulate transitions under partial observability. Instead, we introduce Agentic World Modeling (AWM), which reformulates next-observation prediction as a multi-turn evidence-acquisition task. For each hypothetical action, an agentic simulator adaptively queries the real environment through safe read-only interfaces, reasons over the returned evidence, and uses executable computation to construct the resulting observation without mutating shared state. Integrated with trajectory-level TTS, AWM enables parallel exploration over a single shared environment. Across $\tau$-bench, AppWorld, and OfficeBench, AWM outperforms a single-shot simulation baseline by 40.07% on average, and surpasses a trained world-model baseline by 12.36% on average. These results establish interactive, read-grounded simulation as a practical foundation for safe and scalable agentic test-time scaling in shared, non-resettable environments.
Related papers
- Agent World Model in the Wild: Computer-Use Agent Environment Synthesis
- Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards
- When Does Test-Time Search Help World Action Models? An Empirical Audit of Candidate Selection and Simulator Integrity
- EvoBelief: Falsifiable World Models for Self-Improving LLM Agents
- Agentic Cognitive Depth: Operational Criteria for Evaluating LLM Agents
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-02 · Indexed by Pub-AI: 2026-11-02 UTC
Agentic World Modeling for Agent Test-time Scaling without Environment Replica
First announced by arXiv: 2026-11-02 · Original record ↗
- arxiv
- 2610.40927
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-02 · First detected: 2026-11-02 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2610.40927v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
Test-time scaling (TTS) improves large language model agents by sampling and selecting among multiple candidate trajectories, but common realizations assume that these trajectories can be executed in parallel. This assumption breaks in realistic deployments, where enterprise workspaces and cloud services are shared and persistent, costly or infeasible to replicate, and actions may introduce irreversible side effects or cross-trajectory interference. Existing generative world models avoid direct execution, but typically compress environment state into a single prompt, limiting their ability to reliably simulate transitions under partial observability. Instead, we introduce Agentic World Modeling (AWM), which reformulates next-observation prediction as a multi-turn evidence-acquisition task. For each hypothetical action, an agentic simulator adaptively queries the real environment through safe read-only interfaces, reasons over the returned evidence, and uses executable computation to construct the resulting observation without mutating shared state. Integrated with trajectory-level TTS, AWM enables parallel exploration over a single shared environment. Across $\tau$-bench, AppWorld, and OfficeBench, AWM outperforms a single-shot simulation baseline by 40.07% on average, and surpasses a trained world-model baseline by 12.36% on average. These results establish interactive, read-grounded simulation as a practical foundation for safe and scalable agentic test-time scaling in shared, non-resettable environments.