Zero-Shot Mission-Level Evaluation for Aerial MLLM Agents

arXiv cs.CL Papers

Summary

MissionBench is a new benchmark for evaluating multimodal large language models (MLLMs) on long-horizon embodied tasks in aerial 3D environments, revealing that even the best models succeed on fewer than 35% of missions compared to 84.4% human performance.

arXiv:2607.22014v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) are emerging as core reasoning modules for embodied agents, yet it remains unclear how well general-purpose models can solve long-horizon embodied tasks from a single high-level instruction. We introduce MissionBench, a benchmark for mission-level evaluation of MLLMs in aerial 3D environments. It comprises 120 missions across five simulated 3D environments and four task families. Agents must autonomously plan, navigate, and report outcomes using only egocentric observations and its action history, without aerial-specific fine-tuning. Across 22 open- and closed-source MLLMs, the strongest model succeeds on fewer than 35% of missions compared to 84.4% human performance, highlighting the difficulty of multi-step embodied tasks. Despite large variations between model families, we observe gains from scaling, indicating that larger general-purpose models possess stronger zero-shot embodied capabilities. Our analysis shows that mission-level competence requires coordinating multiple capabilities beyond spatial perception, including multi-step planning and adaptive reasoning. This motivates closed-loop evaluation and highlights both the promise and risk of scaling-driven improvements for embodied AI.
Original Article
View Cached Full Text

Cached at: 07/27/26, 07:42 AM

# Zero-Shot Mission-Level Evaluation for Aerial MLLM Agents
Source: [https://arxiv.org/abs/2607.22014](https://arxiv.org/abs/2607.22014)
[View PDF](https://arxiv.org/pdf/2607.22014)

> Abstract:Multimodal Large Language Models \(MLLMs\) are emerging as core reasoning modules for embodied agents, yet it remains unclear how well general\-purpose models can solve long\-horizon embodied tasks from a single high\-level instruction\. We introduce MissionBench, a benchmark for mission\-level evaluation of MLLMs in aerial 3D environments\. It comprises 120 missions across five simulated 3D environments and four task families\. Agents must autonomously plan, navigate, and report outcomes using only egocentric observations and its action history, without aerial\-specific fine\-tuning\. Across 22 open\- and closed\-source MLLMs, the strongest model succeeds on fewer than 35% of missions compared to 84\.4% human performance, highlighting the difficulty of multi\-step embodied tasks\. Despite large variations between model families, we observe gains from scaling, indicating that larger general\-purpose models possess stronger zero\-shot embodied capabilities\. Our analysis shows that mission\-level competence requires coordinating multiple capabilities beyond spatial perception, including multi\-step planning and adaptive reasoning\. This motivates closed\-loop evaluation and highlights both the promise and risk of scaling\-driven improvements for embodied AI\.

## Submission history

From: Jona Ruthardt \[[view email](https://arxiv.org/show-email/66a745b9/2607.22014)\] **\[v1\]**Fri, 24 Jul 2026 06:22:50 UTC \(17,719 KB\)

Similar Articles

MLUBench: A Benchmark for Lifelong Unlearning Evaluation in MLLMs

arXiv cs.AI

MLUBench is a large-scale benchmark for lifelong unlearning in multimodal large language models (MLLMs), featuring 127 entities across 9 classes. The paper identifies that existing unlearning methods suffer from cumulative degradation and proposes LUMoE to mitigate this, showing significant improvements.

PerceptionBench: Evaluating Atomic Visual Perception in Multimodal Large Language Models

Hugging Face Daily Papers

PerceptionBench is a benchmark designed to evaluate atomic visual perception capabilities of Multimodal Large Language Models (MLLMs), using a bottom-up taxonomy of ten atomic perceptual capabilities. Results across 16 frontier MLLMs show no model reaches 60% accuracy, indicating visual perception remains largely unsolved.