arXiv subjects

Fangyu Hong

Publications and source records attributed to Fangyu Hong.

4 recordsLinked to original sources

arXiv2026

Reinforcement Learning with Robust Rubric Rewards

While Reinforcement Learning with Verifiable Rewards (RLVR) is effective for deterministically checkable tasks, many vision-language tasks are partially verifiable, demanding multi-criteria supervision (e.g., perceptual details, reasoning steps, and constraints). Rubrics provide a natural interface for this fine-grained supervision, but their effectiveness depends on the execution accuracy during online RL. We propose Reinforcement Learning with Robust Rubric Rewards ($\text{RLR}^3$), extending RLVR from task-level verification to criterion-level verification. $\text{RLR}^3$ routes instance-specific rubrics through two execution paths: an LLM-as-an-extractor paired with a deterministic verifier, or an LLM-as-a-Judge for non-verifiable criteria. To ensure faithful scoring, $\text{RLR}^3$ introduce a minimal exposure strategy that masks ground truths from extractors and images from judges. Furthermore, $\text{RLR}^3$ employs hierarchical aggregation to prioritize essential criteria over additional criteria, and mitigates score saturation within rollout groups. Evaluated on Qwen3-VL-30B-A3B across 15 benchmarks, $\text{RLR}^3$ consistently outperforms RLVR, yielding a 4.7-point improvement over the base model and exceeding the official instruct-to-thinking model gap. Controlled audits confirm our deterministic verification and minimal exposure significantly reduce exploitable false positives.

cs.CV↗

arXiv2026

Visual Preference Optimization with Rubric Rewards

The effectiveness of Direct Preference Optimization (DPO) depends on preference data that reflect the quality differences that matter in multimodal tasks. Existing pipelines often rely on off-policy perturbations or coarse outcome-based signals, which are not well suited to fine-grained visual reasoning. We propose rDPO, a preference optimization framework based on instance-specific rubrics. For each image-instruction pair, we create a checklist-style rubric of essential and additional criteria to score responses from any possible policies. The instruction-rubric pool is built offline and reused during the construction of on-policy data. On public reward modeling benchmarks, rubric-based prompting massively improves a 30B-A3B judge and brings it close to GPT-5.4. On public downstream benchmarks, rubric-based filtering raises the macro average to 82.69, whereas outcome-based filtering drops it to 75.82 from 81.14. When evaluating scalability on a comprehensive benchmark, rDPO achieves 61.01, markedly outperforming the style-constrained baseline (52.36) and surpassing the 59.48 base model. Together, these results show that visual preference optimization benefits from combining on-policy data construction with instance-specific criterion-level feedback.

cs.CV↗

arXiv2026

SkyLink: A Large Vision-Language Model Driven Re-ranking Framework for Cross-View UAV geolocalization

Cross-view UAV geolocalization is fundamentally a challenging large-scale image retrieval task, aiming to determine the geographic coordinates of Unmanned Aerial Vehicle (UAV) queries by matching them against an extensive geo-tagged satellite image database. Most existing methods learn separate feature representations for each view and determine the final prediction using naive heuristics to assess feature similarity, thereby neglecting to model the crucial cross-view relationships. In this paper, we propose SkyLink, a novel plug-and-play ranking framework that pioneers joint relational modeling of inter-view relationships to enhance cross-view UAV geolocalization. SkyLink leverages a Large Vision-Language Model (LVLM) to model the intricate visual-semantic relationships between UAV and satellite views, facilitating effective cross-view matching. To further refine the learning process, we introduce a relational-aware loss. It leverages soft labels to provide a more nuanced supervision signal, mitigating the harsh penalty on near-positive pairs. This approach enhances both training stability and the model's discriminative capacity. Extensive experiments conducted across multiple base retrieval architectures and benchmark datasets demonstrate that SkyLink significantly boosts the ranking effectiveness of existing models, consistently achieving superior performance in various challenging scenarios.

cs.CV↗

arXiv2022

Logistics in the Sky: A Two-phase Optimization Approach for the Drone Package Pickup and Delivery System

The application of drones in the last-mile distribution is a research hotspot in recent years. Different from the previous urban distribution mode that depends on trucks, this paper proposes a novel package pick-up and delivery mode and system in which multiple drones collaborate with automatic devices. The proposed mode uses free areas on the top of residential buildings to set automatic devices as delivery and pick-up points of packages, and employs drones to transport packages between buildings and depots. Integrated scheduling problem of package drop-pickup considering m-drone, m-depot, m-customer is crucial for the system. We propose a simulated-annealing-based two-phase optimization approach (SATO) to solve this problem. In the first phase, tasks are allocated to depots for serving, such that the initial problem is decomposed into multiple single depot scheduling problems with m-drone. In the second phase, considering the drone capability constraints and task demand constraints, we generate the route planning scheme for drones in each depot. Concurrently, an improved variable neighborhood descent algorithm (IVND) is designed in the first phase to reallocate tasks, and a local search algorithm (LS) are proposed to search the high-quality solution in the second phase. Finally, extensive experiments and comparative studies are conducted to test the effectiveness of the proposed approach. Experiments indicate that the proposed SATO-IVND can reduce the cost by more than 14% in a reasonable time compared with several other peer algorithms.

eess.SY↗