Selected work

Representative Papers

A Multi-Source Ensemble Approach to Candidate Generation for Alternative Vacation Rental Property Recommendations

Sep 04, 2026

Alternative property recommendations play a critical role in vacation rental marketplaces, helping users discover relevant options when viewing a specific listing. However, generating high-quality candidate alternatives presents unique challenges: heterogeneous inventory, geographic constraints, rapid availability changes, and long-tail property distributions. We present a comprehensive study of candidate generation (CG) approaches for vacation rental alternatives, comparing collaborative filtering, shallow embeddings, and graph neural network (GNN) methods. Our experiments on a large-scale vacation rental platform (over 2M active properties) show that a hybrid architecture combining item-based collaborative filtering with GNN-based retrieval improves Recall@300 by 14.8% over the strongest baseline, by leveraging the complementary strengths of the two sources: collaborative filtering excels at early recall for properties with rich interaction history, while GNNs discover diverse, non-obvious alternatives and handle cold-start scenarios more effectively. As a component result, GNN-based embeddings alone substantially outperform shallow Hotel2Vec embeddings (48-68% relative recall improvement across K), motivating their inclusion in the ensemble. Crucially, we examine how CG-stage gains carry through to the downstream ranking stage, and find that a stronger candidate pool yields higher downstream ranking quality, though attributing this effect cleanly is complicated by the coupling between candidate generation and ranker training. This recall-conversion gap is an important consideration for practitioners deploying new retrieval methods in two-stage recommendation systems.

0 citationsRead paper

HALLELUAI: A Hallucination-Aware AI System for Ultra-Realistic Image-to-Video Generation at Scale

Jul 24, 2026

This work addresses critical quality challenges in large-scale AI video generation—namely hallucination, motion distortion, and aesthetic inconsistency—by introducing the first end-to-end automated quality control system. The system integrates a video assessment module that evaluates frame-level aesthetics, temporal motion fidelity, and fine-grained hallucination risk, coupled with an intelligent regeneration agent to iteratively refine outputs. Innovatively aligning fine-grained hallucination awareness with machine-enforceable creative guidelines, it establishes a closed-loop audit-and-regeneration pipeline that preserves input image fidelity, ensures brand safety, and maintains visual realism while enabling scalable production. Experimental results demonstrate that the system consistently generates hyper-realistic, production-grade videos, significantly enhancing both perceptual quality and reliability as validated by human expert evaluations.

0 citationsRead paper

Safe Overtaking for Autonomous Racing Using Hierarchical Optimization and Learning-Based Control

Jul 14, 2026

This work addresses the performance-safety trade-off in autonomous racing overtaking maneuvers under nonlinear dynamics and real-time constraints by proposing a hierarchical framework. At the high level, a mixed-integer quadratic program (MIQP) determines overtaking topology, while at the low level, a nonlinear model predictive controller (MPC) formulated in Frenét coordinates generates safe trajectories using discrete-time control barrier functions (CBFs). The key innovation lies in decoupling combinatorial overtaking decisions from continuous safety-critical control and integrating reinforcement learning to adaptively tune the CBF decay parameter online, eliminating manual tuning and enabling robust adaptation to diverse track conditions. Experimental results demonstrate that the proposed approach significantly outperforms fixed-parameter strategies in both simulation and real-world tests, achieving high success rates and consistently balancing safety with performance across varied racing scenarios.

0 citationsRead paper

Serving the Long Tail: Training-Free LLM Candidate Generation for Vacation Rental Marketplaces

Jul 10, 2026

This work addresses the failure of collaborative filtering in vacation rental platforms due to sparse interactions with long-tail listings. The authors propose a training-free LLM-based candidate generation method that leverages only static listing metadata: it uses a general-purpose large language model to generate semantic queries, which are then encoded by a pretrained text encoder and retrieved via approximate nearest neighbor (ANN) search. These candidates are fused with those from an existing item-based k-nearest neighbors (IBKNN) channel using a union strategy. This approach is the first to effectively integrate LLMs with static metadata without fine-tuning, significantly improving long-tail coverage while preserving recommendation quality for popular listings. It achieves the largest gains on tens of thousands of listings previously unserved by IBKNN, consistently matching or outperforming baselines across all recall@K metrics, with further improvements after Learning-to-Rank reranking, and narrows the recall gap between a 3B open-source model and state-of-the-art API models to within 1%.

0 citationsRead paper

Travel-Oriented Reasoning Large Language Model via Domain-Specific Knowledge Graphs

Jun 28, 2026

This study addresses the challenge that large language models often exhibit inaccurate and overconfident reasoning in specialized domains like tourism due to a lack of internalized domain knowledge. To mitigate this, the authors propose a modular reasoning framework that leverages an expert-constructed tourism knowledge graph. By performing multi-hop graph traversals, the framework generates question-answer pairs to supervise the fine-tuning of Qwen3-4B using LoRA. This approach uniquely integrates explicit knowledge graph structures with multi-hop reasoning trajectories, endowing the model with auditable and calibratable domain-specific reasoning capabilities. Experimental results demonstrate that the fine-tuned model achieves an exact match accuracy of 82.4% on a tourism benchmark, substantially outperforming baseline methods by 60.0 percentage points, while also revealing two predominant error patterns.

0 citationsRead paper

Recent publications

Latest Papers

A Multi-Source Ensemble Approach to Candidate Generation for Alternative Vacation Rental Property Recommendations

Sep 04, 2026

Alternative property recommendations play a critical role in vacation rental marketplaces, helping users discover relevant options when viewing a specific listing. However, generating high-quality candidate alternatives presents unique challenges: heterogeneous inventory, geographic constraints, rapid availability changes, and long-tail property distributions. We present a comprehensive study of candidate generation (CG) approaches for vacation rental alternatives, comparing collaborative filtering, shallow embeddings, and graph neural network (GNN) methods. Our experiments on a large-scale vacation rental platform (over 2M active properties) show that a hybrid architecture combining item-based collaborative filtering with GNN-based retrieval improves Recall@300 by 14.8% over the strongest baseline, by leveraging the complementary strengths of the two sources: collaborative filtering excels at early recall for properties with rich interaction history, while GNNs discover diverse, non-obvious alternatives and handle cold-start scenarios more effectively. As a component result, GNN-based embeddings alone substantially outperform shallow Hotel2Vec embeddings (48-68% relative recall improvement across K), motivating their inclusion in the ensemble. Crucially, we examine how CG-stage gains carry through to the downstream ranking stage, and find that a stronger candidate pool yields higher downstream ranking quality, though attributing this effect cleanly is complicated by the coupling between candidate generation and ranker training. This recall-conversion gap is an important consideration for practitioners deploying new retrieval methods in two-stage recommendation systems.

0 citationsRead paper

HALLELUAI: A Hallucination-Aware AI System for Ultra-Realistic Image-to-Video Generation at Scale

Jul 24, 2026

This work addresses critical quality challenges in large-scale AI video generation—namely hallucination, motion distortion, and aesthetic inconsistency—by introducing the first end-to-end automated quality control system. The system integrates a video assessment module that evaluates frame-level aesthetics, temporal motion fidelity, and fine-grained hallucination risk, coupled with an intelligent regeneration agent to iteratively refine outputs. Innovatively aligning fine-grained hallucination awareness with machine-enforceable creative guidelines, it establishes a closed-loop audit-and-regeneration pipeline that preserves input image fidelity, ensures brand safety, and maintains visual realism while enabling scalable production. Experimental results demonstrate that the system consistently generates hyper-realistic, production-grade videos, significantly enhancing both perceptual quality and reliability as validated by human expert evaluations.

0 citationsRead paper

Safe Overtaking for Autonomous Racing Using Hierarchical Optimization and Learning-Based Control

Jul 14, 2026

This work addresses the performance-safety trade-off in autonomous racing overtaking maneuvers under nonlinear dynamics and real-time constraints by proposing a hierarchical framework. At the high level, a mixed-integer quadratic program (MIQP) determines overtaking topology, while at the low level, a nonlinear model predictive controller (MPC) formulated in Frenét coordinates generates safe trajectories using discrete-time control barrier functions (CBFs). The key innovation lies in decoupling combinatorial overtaking decisions from continuous safety-critical control and integrating reinforcement learning to adaptively tune the CBF decay parameter online, eliminating manual tuning and enabling robust adaptation to diverse track conditions. Experimental results demonstrate that the proposed approach significantly outperforms fixed-parameter strategies in both simulation and real-world tests, achieving high success rates and consistently balancing safety with performance across varied racing scenarios.

0 citationsRead paper

Serving the Long Tail: Training-Free LLM Candidate Generation for Vacation Rental Marketplaces

Jul 10, 2026

This work addresses the failure of collaborative filtering in vacation rental platforms due to sparse interactions with long-tail listings. The authors propose a training-free LLM-based candidate generation method that leverages only static listing metadata: it uses a general-purpose large language model to generate semantic queries, which are then encoded by a pretrained text encoder and retrieved via approximate nearest neighbor (ANN) search. These candidates are fused with those from an existing item-based k-nearest neighbors (IBKNN) channel using a union strategy. This approach is the first to effectively integrate LLMs with static metadata without fine-tuning, significantly improving long-tail coverage while preserving recommendation quality for popular listings. It achieves the largest gains on tens of thousands of listings previously unserved by IBKNN, consistently matching or outperforming baselines across all recall@K metrics, with further improvements after Learning-to-Rank reranking, and narrows the recall gap between a 3B open-source model and state-of-the-art API models to within 1%.

0 citationsRead paper

Travel-Oriented Reasoning Large Language Model via Domain-Specific Knowledge Graphs

Jun 28, 2026

This study addresses the challenge that large language models often exhibit inaccurate and overconfident reasoning in specialized domains like tourism due to a lack of internalized domain knowledge. To mitigate this, the authors propose a modular reasoning framework that leverages an expert-constructed tourism knowledge graph. By performing multi-hop graph traversals, the framework generates question-answer pairs to supervise the fine-tuning of Qwen3-4B using LoRA. This approach uniquely integrates explicit knowledge graph structures with multi-hop reasoning trajectories, endowing the model with auditable and calibratable domain-specific reasoning capabilities. Experimental results demonstrate that the fine-tuned model achieves an exact match accuracy of 82.4% on a tourism benchmark, substantially outperforming baseline methods by 60.0 percentage points, while also revealing two predominant error patterns.

0 citationsRead paper