WorldTravel: A Realistic Multimodal Travel-Planning Benchmark with Tightly Coupled Constraints

TL;DR

WorldTravel introduces a challenging real-world travel planning benchmark with tightly coupled constraints and a multi-modal environment, exposing significant gaps in current AI models' perception and reasoning capabilities.

Contribution

The paper presents a new benchmark and environment for realistic travel planning, highlighting the limitations of current models in handling complex, multi-modal, tightly coupled constraints.

Findings

State-of-the-art models achieve only 32.67% feasibility in text-only settings.

Performance drops to 19.33% in multi-modal environments.

Identifies a perception-action gap and a planning horizon threshold around 10 constraints.

Abstract

Real-world autonomous planning requires coordinating tightly coupled constraints where a single decision dictates the feasibility of all subsequent actions. However, existing benchmarks predominantly feature loosely coupled constraints solvable through local greedy decisions and rely on idealized data, failing to capture the complexity of extracting parameters from dynamic web environments. We introduce \textbf{WorldTravel}, a benchmark comprising 150 real-world travel scenarios across 5 cities that demand navigating an average of 15+ interdependent temporal and logical constraints. To evaluate agents in realistic deployments, we develop \textbf{WorldTravel-Webscape}, a multi-modal environment featuring over 2,000 rendered webpages where agents must perceive constraint parameters directly from visual layouts to inform their planning. Our evaluation of 10 frontier models reveals a significant performance collapse: even the state-of-the-art GPT-5.2 achieves only 32.67\% feasibility in text-only settings, which plummets to 19.33\% in multi-modal environments. We identify a critical Perception-Action Gap and a Planning Horizon threshold at approximately 10 constraints where model reasoning consistently fails, suggesting that perception and reasoning remain independent bottlenecks. These findings underscore the need for next-generation agents that unify high-fidelity visual perception with long-horizon reasoning to handle brittle real-world logistics.