PilotBench: A Benchmark for General Aviation Agents with Safety Constraints
Authors: Yalun Wu, Haotian Liu, Zhoujun Li, Boyang Wang
Organizations: School of Computing, National University of Singapore, China · School of Informatics, Xiamen University, China · CESS, Beihang University, China
As Large Language Models (LLMs) advance toward embodied AI agents operating in physical environments, a fundamental question emerges: can models trained on text corpora reliably reason about complex physics while adhering to safety constraints? We address this through PilotBench, a benchmark evaluating LLMs on safety-critical flight trajectory and attitude prediction. Built from 708 real-world general aviation trajectories spanning nine operationally distinct flight phases with synchronized 34-channel telemetry, PilotBench systematically probes the intersection of semantic understanding and physics-governed prediction through comparative analysis of LLMs and traditional forecasters. We introduce Pilot-Score, a composite metric balancing 60% regression accuracy with 40% instruction adherence and safety compliance. Comparative evaluation across 41 models uncovers a Precision-Controllability Dichotomy: traditional forecasters achieve superior MAE of 7.01 but lack semantic reasoning capabilities, while LLMs gain controllability with 86--89% instruction-following at the cost of 11--14 MAE precision. Phase-stratified analysis further exposes a Dynamic Complexity Gap-LLM performance degrades sharply in high-workload phases such as Climb and Approach, suggesting brittle implicit physics models. These empirical discoveries motivate hybrid architectures combining LLMs' symbolic reasoning with specialized forecasters' numerical precision. PilotBench provides a rigorous foundation for advancing embodied AI in safety-constrained domains.
Figures & tables
Fig. 1: Synchronized flight-state snapshot from PilotBench during cruise.
Fig. 2: Nine-phase segmentation in PilotBench based on standard traffic pattern.
Fig. 3: Eight-stage pipeline for building PilotBench .
Fig. 4: Flight and spatial statistics of PilotBench .
Regression Error
Instruction Following
Pilot-Score
Model
MAE ↓
RMSE ↓
Overall ↑
Field Comp.
Field Valid
Format
Overall
Qwen3-32B
9.5409
14.7005
85.43
100.0
91.5
100.0
97.43
90.23
Qwen2.5-72B-Instruct
11.9144
16.3591
83.20
100.0
88.8
100.0
96.64
88.58
DeepSeek-V3
11.9410
16.3006
83.19
100.0
86.7
100.0
96.00
88.31
GPT-4o-mini
12.1116
16.5832
82.99
99.8
86.3
99.8
95.74
88.09
GPT-o3-mini
10.9403
15.0377
84.24
100.0
76.8
100.0
93.05
87.76
TABLE I: Main results on PilotBench
Fig. 5: Performance radar: traditional models shown in blue dominate MAE/VR; LLMs shown in orange, green, and purple gain IF controllability. Qwen3-32B +Phy-CoT achieves the best balance.
Fig. 6: Decoupling prompting effects: +ICL benefits MAE precision; +Phy-CoT drives VR safety gains, suggesting distinct mechanisms for pattern matching vs. constraint satisfaction.
Fig. 7: MAE heatmap across flight phases. LLM errors spike in Climb and Approach, marked in red, while traditional baselines remain stable in yellow, exposing the Dynamic Complexity Gap.
Evaluating large language models (LLMs) in safety-critical, physics-governed environments requires more than accuracy-based metrics, because predictions that are numerically close to the ground truth can still violate operational constraints, combine fields in physically inconsistent ways, or fail to produce usable structured outputs. Existing evaluation protocols do not measure these failure modes reliably. We propose FLY-EVAL++, an evidence-driven evaluation protocol that combines deterministic verification of protocol compliance, physical feasibility, and safety constraints with fixed rubric-guided aggregation into interpretable multi-dimensional scores. We instantiate FLY-EVAL++ for Flight Trajectory and Attitude Prediction (FTAP) by extending the PilotBench setting with history-conditioned and multi-step prediction tasks. Across 66 LLMs, safety compliance is the most discriminative dimension of model behavior: models with comparable predictive performance differ by more than 28 points in safety score, and we observe recurrent failures including safety violations under physically plausible predictions and instability in multi-step rollouts. These results show that evaluation in safety-critical domains should measure constraint satisfaction and structured validity explicitly rather than rely on accuracy-centric reporting alone.
Yalun Wu, Junfeng Fang, Jiawei Wang +6
NExT++ Lab, School of Computing, National University of Singapore · 3Xiamen University · University of Manchester +4
Aviation knowledge question answering cannot directly assess the operational effectiveness and safety compliance of large language models throughout aircraft emergency procedures. We introduce AeroCopilotBench and its executable cockpit environment, ACOE, which define state-transition rules, task goals, and trajectory-level safety constraints based on aircraft-specific Pilot's Operating Handbooks (POHs). The benchmark comprises 12 scenario templates and 73 tasks across two aircraft types, evaluating task completion, safety compliance, execution discipline, and repeatability. Across repeated evaluations of 12 models, the highest safety-gated success rate is 72.6%. Most failed episodes achieve all critical goals but do not satisfy all remaining terminal goals. Across repeated runs, models still omit steps that they execute in other runs of the same task. Analysis of failed trajectories further reveals that some reasons for actions that conflict with the aircraft's POH recur across models. These results expose shortcomings in complete procedure execution, consistency across runs, and aircraft-specific emergency response.
Yuchen Yuan, Zhenghuang Wu, Yuangan Li +2
School of Aeronautic Science and Engineering, Beihang University
Large language models (LLMs) are increasingly proposed for aviation business operations, from documentation and training generation to customer facing assistants. General purpose benchmarks do not measure whether a model reasons safely and correctly about aviation specific operational knowledge, and the high stakes, regulated nature of the domain makes that gap consequential. We present Pre-Flight, an open source benchmark of 300 multiple choice questions drawn from international standards and airport ground operations material, covering international airport ground operations, ICAO and US FAA regulations, aviation general knowledge and complex operational scenarios. Questions were authored and reviewed by practitioners with experience in air traffic management, ground operations and commercial flying. We evaluate a range of contemporary commercial and open weight models using the Inspect evaluation framework, scoring by accuracy under a standard multiple choice protocol, and we maintain the leaderboard on a rolling basis as new models are released. Against an informal expert reference of around 95%, obtained from a low sample quiz of aviation professionals at a conference, even the strongest model evaluated (released in 2026) reaches 82.7%, having improved only gradually from roughly 75% in early 2025. A substantial and persistent gap below expert level reliability therefore remains. We release the dataset, the evaluation harness and the results, and the benchmark is available within the community evaluations package distributed with inspect_evals. We argue that domain specific evaluation of this kind is a necessary precondition for responsible deployment of generative AI in non safety critical aviation operations.
Alex Brooker, Tim Hughes
Airside Labs, London, United Kingdom · Mahino Research, Christchurch, New Zealand