Papers by Zeyi Wen
Enhancing Talent Search Ranking with Role-Aware Expert Mixtures and LLM-based Fine-Grained Job Descriptions (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Existing talent search approaches fail to capture nuanced job-specific preferences and mitigate noise from subjective human judgments. |
| Approach: | They propose a framework that extracts fine-grained recruitment signals from job descriptions and historical hiring data and employs a role-aware multi-gate MoE network to capture behavioral differences across recruiter roles. |
| Outcome: | The proposed framework improves talent search effectiveness and delivers substantial business value. |
Enhancing Online Recruitment with Category-Aware MoE and LLM-based Data Augmentation (2026.acl-industry)
Copied to clipboard
| Challenge: | Existing methods to measure the matching degree of a job and a candidate face several challenges, such as low-quality job descriptions and similar candidate-job pairs. |
| Approach: | They propose a large language model-based method that polishes and rewrites low-quality job descriptions by leveraging chain-of-thought prompts and category-aware Mixture of Experts (MoE) module incorporates category embeddings to dynamically assign weights to the experts and learns more distinguishable patterns for similar candidate-job pairs. |
| Outcome: | The proposed method surpasses existing methods by 2.40% in AUC and 7.46% in GAUC and boosts click-through conversion rate (CTCVR) by 19.4% in online tests, saving millions of CNY in external headhunting expenses. |
EcoTune: Token-Efficient Multi-Fidelity Hyperparameter Optimization for Large Language Model Inference (2025.emnlp-main)
Copied to clipboard
| Challenge: | Multi-fidelity optimization improves inference performance with low-fidelity evaluations, but its static scheduling strategies ignore token consumption, leading to high costs. |
| Approach: | They propose a token-efficient multi-fidelity optimization method that optimizes for inference hyperparameters . their method reduces token consumption by over 80% while maintaining or surpassing performance . |
| Outcome: | The proposed method improves over the HELM leaderboard by 7.1%, 24.3%, 21.9%, and 4.6% while maintaining or surpassing performance. |
Exploiting Careful Design of SVM Solution for Aspect-term Sentiment Analysis (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Aspect-term sentiment analysis (ATSA) identifies fine-grained sentiments towards specific aspects of text. |
| Approach: | They propose a pipeline to predict fine-grained sentiments for specific aspects of text . it decomposes the learning problem into multiple view subproblems and dynamically selects and constructs features with reinforcement learning. |
| Outcome: | The proposed pipeline surpasses SVM-based methods in predictive accuracy while maintaining a faster inference speed and significantly reducing the number of model parameters. |
Efficient Hyperparameter Optimization for LLM Reinforcement Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing hyperparameter optimization methods are inefficient in reinforcement learning due to model scale and resource-intensive training cycles. |
| Approach: | They propose a hyperparameter optimization method that adapts both model size and training budget as fidelity. |
| Outcome: | The proposed method significantly improves the computational efficiency of each trial (up to 14.9) over existing HPO methods. |
Evaluating Small Language Models for News Summarization: Implications and Factors Influencing Performance (2025.naacl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) provide superior summarization quality, but their high computational resource requirements limit practical use applications. |
| Approach: | They evaluate 19 small language models for news summarization across 2,000 news samples . they find that top-performing models achieve comparable results to those of 70B LLMs . |
| Outcome: | The proposed models achieve comparable results to 70B LLMs while generating more concise summaries. |
SEAL: Structure and Element Aware Learning Improves Long Structured Document Retrieval (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for document retrieval use contrastive learning on datasets lacking explicit structural information. |
| Approach: | They propose a contrastive learning framework that preserves semantic hierarchies and masked element alignment for fine-grained semantic discrimination. |
| Outcome: | The proposed framework preserves semantic hierarchies and masked element alignment for fine-grained semantic discrimination. |