# PhysAI-Bench: A Benchmark for LLM-Based Agentic Decision-Making in Autonomous UAV-Centric Physical AI

arXiv: https://arxiv.org/abs/2609.23695 · HTML: https://mlsanity.com/paper/2609.23695/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 23 of its 31 references that are in the map.
- First posted on arXiv 20 Sep 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 23 of 23)

- [Cosmos 3: Omnimodal World Models for Physical AI](https://mlsanity.com/paper/2606.02800/) (2026)
- [Do generative video models understand physical principles?](https://mlsanity.com/paper/2501.09038/) (2025)
- [RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies](https://mlsanity.com/paper/2506.18123/) (2025)
- [IntPhys 2: Benchmarking Intuitive Physics Understanding In Complex Synthetic Environments](https://mlsanity.com/paper/2506.09849/) (2025)
- [T2VPhysBench: A First-Principles Benchmark for Physical Consistency in Text-to-Video Generation](https://mlsanity.com/paper/2505.00337/) (2025)
- [CausalVQA: A Physically Grounded Causal Reasoning Benchmark for Video Models](https://mlsanity.com/paper/2506.09943/) (2025)
- [PAI-Bench: A Comprehensive Benchmark For Physical AI](https://mlsanity.com/paper/2512.01989/) (2025)
- [Large Language Models for UAVs: Current State and Pathways to the Future](https://mlsanity.com/paper/2405.01745/) (2024)
- [Rethinking Video Generation Model for the Embodied World](https://mlsanity.com/paper/2601.15282/) (2026)
- [A Shortcut-aware Video-QA Benchmark for Physical Understanding via Minimal Video Pairs](https://mlsanity.com/paper/2506.09987/) (2025)
- [ABot-PhysWorld: Interactive World Foundation Model for Robotic Manipulation with Physics Alignment](https://mlsanity.com/paper/2603.23376/) (2026)
- [NavBench: Probing Multimodal Large Language Models for Embodied Navigation](https://mlsanity.com/paper/2506.01031/) (2025)
- [RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies](https://mlsanity.com/paper/2604.09860/) (2026)
- [AirCopBench: A Benchmark for Multi-drone Collaborative Embodied Perception and Reasoning](https://mlsanity.com/paper/2511.11025/) (2025)
- [RoboWM-Bench: A Benchmark for Evaluating World Models in Robotic Manipulation](https://mlsanity.com/paper/2604.19092/) (2026)
- [SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning](https://mlsanity.com/paper/2606.13673/) (2026)
- [Reasoning Beyond Limits: Advances and Open Problems for LLMs](https://mlsanity.com/paper/2503.22732/) (2025)
- [PhysGym: Benchmarking LLMs in Interactive Physics Discovery with Controlled Priors](https://mlsanity.com/paper/2507.15550/) (2025)
- [DeepPHY: Benchmarking Agentic VLMs on Physical Reasoning](https://mlsanity.com/paper/2508.05405/) (2025)
- [Explore with Long-term Memory: A Benchmark and Multimodal LLM-based Reinforcement Learning Framework for Embodied Exploration](https://mlsanity.com/paper/2601.10744/) (2026)
- [α3-Bench: A Unified Benchmark of Safety, Robustness, and Efficiency for LLM-Based UAV Agents over 6G Networks](https://mlsanity.com/paper/2601.03281/) (2026)
- [ACWM-Phys: Investigating Generalized Physical Interaction in Action-Conditioned Video World Models](https://mlsanity.com/paper/2605.08567/) (2026)
- [A Tutorial on Cognitive Biases in Agentic AI-Driven 6G Autonomous Networks](https://mlsanity.com/paper/2510.19973/) (2025)

## Abstract (arXiv)

Recent advances in Physical AI have accelerated the use of foundation models in autonomous systems such as unmanned aerial vehicles (UAVs), which must perceive, reason, plan, and act in dynamic environments. Existing benchmarks assess physical perception, intuitive physics, embodied navigation, and collaborative reasoning, but rarely evaluate the agentic decision-making required for reliable autonomy. We introduce textitPhysAI-Bench, a benchmark for evaluating this capability. It contains 10,178 standardized decision instances automatically extracted from conversational traces of autonomous UAV missions. Each instance preserves mission context, temporal dependencies, physical constraints, Model Context Protocol (MCP) tool calls, Agent-to-Agent (A2A) interactions, sensor observations, and AI-native 6G network conditions, including latency, packet loss, throughput, edge load, and network slicing. We expose only information preceding each decision, preventing future-event leakage and approximating online decision-making. We evaluate 29 foundation models using a two-stage protocol. We select model-specific configurations from 12 combinations of zero-, three-, and five-shot prompting and four temperatures, tested in three runs on a 35-instance, human-verified development set. We then freeze each selected configuration and evaluate it in three runs on a fixed, episode-disjoint set of 500 instances. GPT-5.3 achieves the highest accuracy (52.00%), followed by GPT-5.2 (49.40%) and Grok~4.5 (49.07%). Few-shot prompting generally improves performance, while temperature has limited influence. The results demonstrate that reliable agentic decision-making in Physical AI remains an open challenge. The dataset is available at https://github.com/maferrag/physai-bench