ORBITER: Conflict-Aware Decision-Making for Agentic Last-Mile Delivery
Authors: Mingzhao Li, Chenxi Liu, Yan Zhao, Hao Miao
Organizations: Centre for Artificial Intelligence and Robotics, Hong Kong Institute of Science & Innovation, Chinese Academy of Sciences, Hong Kong SAR · Shenzhen Institute for Advanced Study, University of Electronic Science and Technology of China, Shenzhen, China · The Hong Kong Polytechnic University, Hong Kong SAR
Last-mile delivery aims to handle dynamically arriving orders with couriers while modeling complex spatial and temporal correlations. Recent learning-based methods model spatiotemporal dependencies among orders to predict courier service sequences, but leave next-order decision making unexplained. Describing the current delivery state in language allows LLMs to reason explicitly about the spatial, temporal, and behavioral cues behind an individual decision. As direct predictors, however, LLMs remain sensitive to task presentation and often produce unreliable decisions. To address these challenges, we introduce ORBITER, an agentic Order Arbiter for next-order decision-making in last-mile delivery. ORBITER models courier service through decision points, each containing the courier's spatiotemporal state and visible orders and exposing local trade-offs for modeling and verification. Fixed proposers rank the candidates, and a structured report identifies where their rankings disagree. The LLM uses task-specific tools to gather evidence on the leading alternatives, while an independent critic checks the resulting decision against that evidence. We conduct extensive evaluations on data in four cities, where ORBITER outperforms existing state-of-the-art baselines by up to 9.2% on average showing its effectiveness.
Figures & tables
Figure 1: Previous mobility modeling compresses local choices into full routes, whereas agentic decision making resolves conflicting cues with auditable evidence.
Figure 2: Overall architecture of ORBITER.
City
Packages
Couriers
AvgPackage
Shanghai
1,450k
4,502
15.0
Chongqing
1,172k
2,982
14.0
Yantai
1,146k
2,593
16.0
Jilin
261k
665
13.8
Table 1: Statistics of the LaDe-P subsets.
Model
Shanghai
Jilin
Chongqing
Yantai
ACC@1
ACC@3
MRR
ACC@1
ACC@3
MRR
ACC@1
ACC@3
MRR
ACC@1
ACC@3
MRR
Distance Greedy
0.250
0.575
0.452
0.255
0.570
0.465
0.280
0.600
0.478
0.280
0.610
0.483
Deadline Greedy
0.205
0.505
0.414
0.185
0.430
0.383
0.220
0.565
0.440
0.210
0.535
0.414
Weighted Rule
0.435
0.755
0.628
0.350
0.655
0.543
0.470
0.745
0.636
0.480
0.760
0.640
LightGBM
0.475
0.735
0.644
0.480
0.755
0.646
0.505
0.785
0.665
0.450
0.790
0.644
XGBoost
0.500
0.760
0.660
0.485
0.765
0.655
0.525
0.790
0.680
0.450
0.795
0.642
Table 2: Main results on the four city subsets of LaDe-P. All LLM-based methods use DeepSeek-V4-Flash as the backbone.
Figure 3: Ablation studies.
Figure 4: Performance across backbone LLMs.
Figure 5: Conflicting cues among candidate orders.
Figure 6: Example of a Structured Disagreement Report.
In recent years, the growing complexity of last-mile pickup operations has increased the need for fast and accurate decision-making on logistics platforms. This challenge is fundamentally driven by two key and tightly coupled decision-making processes: order dispatching and routing. Solving them separately overlooks their interdependence, while fully end-to-end learning can be unstable and costly on large, variable-scale instances due to sparse rewards. To solve this problem, we propose an integrated optimization framework which couples a learned routing oracle with real-time dispatching heuristics. For the routing subproblem, we develop a Dynamic-Residual Graph Attention Network encoder with a Look-Ahead Courier-Personalized decoder. For the dispatching subproblem, we develop a routing-oracle-guided dispatching heuristic with local search, where the oracle provides near-optimal solutions to select candidate couriers while retaining real-time scalability. Extensive experiments on real-world datasets from Cainiao Logistics are used to test the performance of our approach, including an offline evaluation and an online rolling-horizon simulation. The experimental results show that our approach outperforms other benchmarks regarding solution quality and solving time, indicating it can effectively support logistics companies in solving real-time and large-scale last-mile pickup problems.
Yida Xu, Zhaofang Mao, Yuheng Miao +2
College of Management and Economics, Tianjin University, Tianjin 300072, China · Laboratory of Computation and Analytics of Complex Management Systems (CACMS), Tianjin University, Tianjin 300072, China
Large language models (LLMs) are increasingly used in recommender systems, but it is often unclear how much performance can be obtained from strong pre-trained backbones alone when they are placed inside a structured recommendation pipeline. In this paper, we present MARS, a modular multi-agent re-ranking framework for repeat-order food delivery recommendation. MARS serves as a controlled hybrid framework for studying how far pre-trained LLMs can go in this setting when combined with lightweight collaborative retrieval and contextual filtering. MARS performs coarse-to-fine recommendation in two stages: cuisine prediction followed by vendor ranking. The framework combines LightGCN-based global preference signals, Swing-based local peer evidence, geospatial filtering, and prompt-driven LLM reasoning over behavioral, temporal, and geographic context. We evaluate MARS on two real-world Delivery Hero benchmarks, DHRD-SE and DHRD-SG, and compare it against heuristic, sequential, graph-based, and food-delivery-specific baselines. We also provide detailed implementation and evaluation protocols, including prompting and decoding. Our study makes three contributions. First, it presents a modular multi-agent framework for repeat-order food delivery recommendation that integrates collaborative signals and LLM-based re-ranking in a transparent pipeline. Second, it shows that strong pre-trained backbones can already be competitive in repeat-order recommendation when paired with lightweight collaborative retrieval. Third, it establishes a reproducible evaluation setting for hybrid LLM recommenders in food delivery.
Jiahao Tian, Zhenkai Wang
The University of Texas at Austin Austin, Texas, USA
The rapid development of large language models, each with distinct capabilities and inference costs, raises a practical deployment question: given an incoming request, which model should handle it? We present OrcaRouter, a production-oriented LLM router that combines a LinUCB-based contextual bandit over lexical and sentence-embedding features with a hybrid offline-online learning protocol. Offline, OrcaRouter obtains full-information feedback by evaluating each candidate model on a curated set of routing prompts, yielding a reward matrix used to fit one ridge regressor per arm. At deployment time, it initializes from these parameters and can optionally continue learning from bandit feedback, updating only the selected model's arm after observing its reward. At the time of our RouterArena submission (May 20, 2026), OrcaRouter-Adaptive ranked second on the public RouterArena leaderboard with an arena score of 72.08, achieving 75.54% accuracy at a cost of USD 1.00 per 1,000 queries.