arXiv subjects

Zeyi Wen

Publications and source records attributed to Zeyi Wen.

At least 19 recordsLinked to original sources

arXiv2026

Strong Drafts Need Compact Memories: Long-Context Speculative Decoding with Compressed KV Cache

Long-context LLM applications such as document summarization and multi-turn agents require generation from prefixes spanning tens of thousands of tokens, making decoding latency a major bottleneck. Speculative decoding (SD) reduces latency without changing model outputs, but its speedup depends on both accepted draft tokens and draft-step latency: Lightweight drafts are fast but lack the capacity to capture long-range dependencies, whereas strong independent drafts recover acceptance but incur growing KV-access cost at long prefixes. We introduce memory-augmented drafting for long-context SD, equipping a strong independent draft with compressed draft-side KV memory: A lightweight adaptor constructs and incrementally updates this memory to retain distant information and exact recent context. The target verifier retains its full KV cache and applies the standard accept/reject rule, preserving SD's lossless guarantee. Experiments on Llama~3.1-8B and 70B targets at prefix lengths up to 32K show that our method reduces draft-side memory by over 70%. It achieves speedups of up to 2.08x and 3.33x , respectively, over autoregressive decoding.

cs.LG↗

arXiv2026

EntSQL: A Benchmark for Grounding Text-to-SQL in Long-Context Enterprise Knowledge

Text-to-SQL enables natural language access to databases, and recent LLMs have substantially advanced its capabilities. Existing benchmarks such as Spider, BIRD, and Spider~2.0 evaluate schema generalization, large-scale databases, and realistic workflows, but largely overlook enterprise scenarios where SQL generation depends on private business knowledge, such as internal metrics, reporting conventions, and organizational rules. We introduce EntSQL, an enterprise-oriented Text-to-SQL benchmark for evaluating long-context grounding over proprietary business documents. EntSQL contains 1,066 aligned Chinese-English semantic examples across five business domains, with most examples requiring domain knowledge beyond the question and schema and involving complex SQL structures. On English inputs, the best evaluated system reaches only 15.9\% when long-form documents are provided, highlighting the difficulty of grounding SQL generation in enterprise knowledge.

cs.CL↗

arXiv2026

An Efficient Heterogeneous Co-Design for Fine-Tuning on a Single GPU

Fine-tuning Large Language Models (LLMs) has become essential for domain adaptation, but its memory-intensive property exceeds the capabilities of most GPUs. To address this challenge and democratize LLM fine-tuning, we present SlideFormer, a novel system designed for single-GPU environments. Our innovations are: (1) A lightweight asynchronous engine that treats the GPU as a sliding window and overlaps GPU computation with CPU updates and multi-tier I/O. (2) A highly efficient heterogeneous memory management scheme significantly reduces peak memory usage. (3) Optimized Triton kernels to solve key bottlenecks and integrated advanced I/O. This collaborative design enables fine-tuning of the latest 123B+ models on a single RTX 4090, supporting up to 8x larger batch sizes and 6x larger models. In evaluations, SlideFormer achieves 1.40x to 6.27x higher throughput while roughly halving CPU/GPU memory usage compared to baselines, sustaining >95% peak performance on both NVIDIA and AMD GPUs.The code is available at https://github.com/RegiaYoung/SlideFormer.

cs.DC↗

arXiv2026

Preserving Data Privacy in Learning Causal Structure with Fully Homomorphic Encryption

Preserving data privacy is an important topic in structural data management and data mining. However, the issue of privacy leakage in distributed causal structure learning is a persistent challenge, especially in cases where data transmission and computation are required. In this paper, we propose a method based on fully homomorphic encryption (FHE) that performs calculations on ciphertexts, keeping data encrypted in transition and computation. Nevertheless, adopting FHE to causal structure learning is challenging due to the high computation cost and limited support on division as well as logarithm operations in FHE. To tackle this challenge, we propose a series of novel techniques including (i) circuit simplification for better efficiency, (ii) approximation of division and logarithm through Newton-Raphson Reciprocal and Taylor expansion, and (iii) a batching technique with SIMD-acceleration to enhance the whole learning process. Additionally, our method can be easily extended beyond FHE by demonstration of its portability to support differential privacy. Empirical results show that our method achieves high consistency and comparable causal structure with the plaintext version in the datasets tested. Last, our method is efficient and practical to complete learning causal structures in tens of minutes even under the privacy protection of FHE.

cs.CR↗

arXiv2026

DeInfer: Efficient Parallel Inferencing for Decomposed Large Language Models

Existing works on large language model (LLM) decomposition mainly focus on improving performance on downstream tasks, but they ignore the poor parallel inference performance when trying to scale up the model size. To mitigate this important performance issue, this paper introduces DeInfer, a high-performance inference system dedicated to parallel inference of decomposed LLMs. It consists of multiple optimizations to maximize performance and be compatible with state-of-the-art optimization techniques. Extensive experiments are carried out to evaluate DeInfer's performance, where the results demonstrate its superiority, suggesting it can greatly facilitate the parallel inference of decomposed LLMs.

cs.CL↗

arXiv2026

Efficient Hyperparameter Optimization for LLM Reinforcement Learning

Reinforcement learning (RL) for large language models (LLMs) is highly sensitive to hyperparameter configurations, making hyperparameter optimization (HPO) essential yet computationally expensive. Existing multi-fidelity HPO methods remain inefficient for LLM RL due to the massive model scale and resource-intensive training cycles. In this paper, we propose Joint Fidelity Hyperparameter Optimization (JF-HPO), which simultaneously adapts both model size and training budget as fidelity. JF-HPO is empowered by: (i) it leverages a small proxy model of the target LLM for efficient training and evaluation in each HPO trial; (ii) it integrates carefully designed early-stopping strategies based on training dynamics; (iii) it introduces an efficient checkpointing mechanism to eliminate redundant computations. Compared with existing HPO methods, JF-HPO significantly improves the computational efficiency of each trial (up to 14.9 times), while achieving better or competitive predictive accuracy under the same time budget. Notably, compared with utilizing hyperparameter configurations from the VeRL Recipe, JF-HPO delivers performance improvements ranging from 5.8% to 111.6%.

cs.LG↗

arXiv2026

Taxon: Hierarchical Tax Code Prediction with Semantically Aligned LLM Expert Guidance

Tax code prediction is a crucial yet underexplored task in automating invoicing and compliance management for large-scale e-commerce platforms. Each product must be accurately mapped to a node within a multi-level taxonomic hierarchy defined by national standards, where errors lead to financial inconsistencies and regulatory risks. This paper presents Taxon, a semantically aligned and expert-guided framework for hierarchical tax code prediction. Taxon integrates (i) a feature-gating mixture-of-experts architecture that adaptively routes multi-modal features across taxonomy levels, and (ii) a semantic consistency model distilled from large language models acting as domain experts to verify alignment between product titles and official tax definitions. To address noisy supervision in real business records, we design a multi-source training pipeline that combines curated tax databases, invoice validation logs, and merchant registration data to provide both structural and semantic supervision. Extensive experiments on the proprietary TaxCode dataset and public benchmarks demonstrate that Taxon achieves state-of-the-art performance, outperforming strong baselines. Further, an additional full hierarchical paths reconstruction procedure significantly improves structural consistency, yielding the highest overall F1 scores. Taxon has been deployed in production within Alibaba's tax service system, handling an average of over 500,000 tax code queries per day and reaching peak volumes above five million requests during business event with improved accuracy, interpretability, and robustness.

cs.LG↗

arXiv2026

Enhancing Online Recruitment with Category-Aware MoE and LLM-based Data Augmentation

Person-Job Fit (PJF) is a critical component for online recruitment. Existing approaches face several challenges, particularly in handling low-quality job descriptions and similar candidate-job pairs, which impair model performance. To address these challenges, this paper proposes a large language model (LLM) based method with two novel techniques: (1) LLM-based data augmentation, which polishes and rewrites low-quality job descriptions by leveraging chain-of-thought (COT) prompts, and (2) category-aware Mixture of Experts (MoE) that assists in identifying similar candidate-job pairs. This MoE module incorporates category embeddings to dynamically assign weights to the experts and learns more distinguishable patterns for similar candidate-job pairs. We perform offline evaluations and online A/B tests on our recruitment platform. Our method relatively 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.

cs.AI↗

arXiv2026

DOne: Decoupling Structure and Rendering for High-Fidelity Design-to-Code Generation

While Vision Language Models (VLMs) have shown promise in Design-to-Code generation, they suffer from a "holistic bottleneck-failing to reconcile high-level structural hierarchy with fine-grained visual details, often resulting in layout distortions or generic placeholders. To bridge this gap, we propose DOne, an end-to-end framework that decouples structure understanding from element rendering. DOne introduces (1) a learned layout segmentation module to decompose complex designs, avoiding the limitations of heuristic cropping; (2) a specialized hybrid element retriever to handle the extreme aspect ratios and densities of UI components; and (3) a schema-guided generation paradigm that bridges layout and code. To rigorously assess performance, we introduce HiFi2Code, a benchmark featuring significantly higher layout complexity than existing datasets. Extensive evaluations on the HiFi2Code demonstrate that DOne outperforms exiting methods in both high-level visual similarity (e.g., over 10% in GPT Score) and fine-grained element alignment. Human evaluations confirm a 3 times productivity gain with higher visual fidelity.

cs.CV↗

arXiv2026

SoLA: Leveraging Soft Activation Sparsity and Low-Rank Decomposition for Large Language Model Compression

Large language models (LLMs) have demonstrated impressive capabilities across various tasks, but the billion-scale parameters pose deployment challenges. Although existing methods attempt to reduce the scale of LLMs, they require either special hardware support or expensive post-training to maintain model quality. To facilitate efficient and affordable model slimming, we propose a novel training-free compression method for LLMs, named "SoLA", which leverages \textbf{So}ft activation sparsity and \textbf{L}ow-r\textbf{A}nk decomposition. SoLA can identify and retain a minority of components significantly contributing to inference, while compressing the majority through low-rank decomposition, based on our analysis of the activation pattern in the feed-forward network (FFN) of modern LLMs. To alleviate the decomposition loss, SoLA is equipped with an adaptive component-wise low-rank allocation strategy to assign appropriate truncation positions for different weight matrices. We conduct extensive experiments on LLaMA-2-7B/13B/70B and Mistral-7B models across a variety of benchmarks. SoLA exhibits remarkable improvement in both language modeling and downstream task accuracy without post-training. For example, with a 30\% compression rate on the LLaMA-2-70B model, SoLA surpasses the state-of-the-art method by reducing perplexity from 6.95 to 4.44 and enhancing downstream task accuracy by 10\%.

cs.CL↗

arXiv2026

Quasar: Quantized Self-Speculative Acceleration for Rapid Inference via Memory-Efficient Verification

Speculative Decoding (SD) has emerged as a premier technique for accelerating Large Language Model (LLM) inference by decoupling token generation into rapid drafting and parallel verification. While recent advancements in self-speculation and lookahead decoding have successfully minimized drafting overhead, they have shifted the primary performance bottleneck to the verification phase. Since verification requires a full forward pass of the target model, it remains strictly memory-bandwidth bound, fundamentally limiting the maximum achievable speedup.In this paper, we introduce \textbf{Quasar} (\textbf{Qua}ntized \textbf{S}elf-speculative \textbf{A}cceleration for \textbf{R}apid Inference), a novel, training-free framework designed to overcome this "memory wall" by employing low-bit quantization specifically for the verification stage. Our empirical analysis reveals that while aggressive structural pruning significantly degrades verification accuracy, quantization-based verification preserves the logit distribution with high fidelity while effectively halving memory traffic. Extensive experiments on state-of-the-art models (e.g., OpenPangu and Qwen3) demonstrate that Quasar maintains a speculative acceptance length comparable to full-precision methods while achieving a $1.28\times$ improvement in end-to-end throughput. Being orthogonal to existing drafting strategies, Quasar offers a generic and efficient pathway to accelerate the verification leg of speculative execution. Code is available at https://github.com/Tom-HG/Quasar.

cs.DC↗

arXiv2025

Enhancing Talent Search Ranking with Role-Aware Expert Mixtures and LLM-based Fine-Grained Job Descriptions

Talent search is a cornerstone of modern recruitment systems, yet existing approaches often struggle to capture nuanced job-specific preferences, model recruiter behavior at a fine-grained level, and mitigate noise from subjective human judgments. We present a novel framework that enhances talent search effectiveness and delivers substantial business value through two key innovations: (i) leveraging LLMs to extract fine-grained recruitment signals from job descriptions and historical hiring data, and (ii) employing a role-aware multi-gate MoE network to capture behavioral differences across recruiter roles. To further reduce noise, we introduce a multi-task learning module that jointly optimizes click-through rate (CTR), conversion rate (CVR), and resume matching relevance. Experiments on real-world recruitment data and online A/B testing show relative AUC gains of 1.70% (CTR) and 5.97% (CVR), and a 17.29% lift in click-through conversion rate. These improvements reduce dependence on external sourcing channels, enabling an estimated annual cost saving of millions of CNY.

cs.IR↗

arXiv2025

SEAL: Structure and Element Aware Learning to Improve Long Structured Document Retrieval

In long structured document retrieval, existing methods typically fine-tune pre-trained language models (PLMs) using contrastive learning on datasets lacking explicit structural information. This practice suffers from two critical issues: 1) current methods fail to leverage structural features and element-level semantics effectively, and 2) the lack of datasets containing structural metadata. To bridge these gaps, we propose \our, a novel contrastive learning framework. It leverages structure-aware learning to preserve semantic hierarchies and masked element alignment for fine-grained semantic discrimination. Furthermore, we release \dataset, a long structured document retrieval dataset with rich structural annotations. Extensive experiments on both released and industrial datasets across various modern PLMs, along with online A/B testing, demonstrate consistent performance improvements, boosting NDCG@10 from 73.96\% to 77.84\% on BGE-M3. The resources are available at https://github.com/xinhaoH/SEAL.

cs.IR↗

arXiv2025

RAMer: Reconstruction-based Adversarial Model for Multi-party Multi-modal Multi-label Emotion Recognition

Conventional Multi-modal multi-label emotion recognition (MMER) assumes complete access to visual, textual, and acoustic modalities. However, real-world multi-party settings often violate this assumption, as non-speakers frequently lack acoustic and textual inputs, leading to a significant degradation in model performance. Existing approaches also tend to unify heterogeneous modalities into a single representation, overlooking each modality's unique characteristics. To address these challenges, we propose RAMer (Reconstruction-based Adversarial Model for Emotion Recognition), which refines multi-modal representations by not only exploring modality commonality and specificity but crucially by leveraging reconstructed features, enhanced by contrastive learning, to overcome data incompleteness and enrich feature quality. RAMer also introduces a personality auxiliary task to complement missing modalities using modality-level attention, improving emotion reasoning. To further strengthen the model's ability to capture label and modality interdependency, we propose a stack shuffle strategy to enrich correlations between labels and modality-specific features. Experiments on three benchmarks, i.e., MEmoR, CMU-MOSEI, and $M^3ED$, demonstrate that RAMer achieves state-of-the-art performance in dyadic and multi-party MMER scenarios.

cs.CV↗

arXiv2025

Evaluating Small Language Models for News Summarization: Implications and Factors Influencing Performance

The increasing demand for efficient summarization tools in resource-constrained environments highlights the need for effective solutions. While large language models (LLMs) deliver superior summarization quality, their high computational resource requirements limit practical use applications. In contrast, small language models (SLMs) present a more accessible alternative, capable of real-time summarization on edge devices. However, their summarization capabilities and comparative performance against LLMs remain underexplored. This paper addresses this gap by presenting a comprehensive evaluation of 19 SLMs for news summarization across 2,000 news samples, focusing on relevance, coherence, factual consistency, and summary length. Our findings reveal significant variations in SLM performance, with top-performing models such as Phi3-Mini and Llama3.2-3B-Ins achieving results comparable to those of 70B LLMs while generating more concise summaries. Notably, SLMs are better suited for simple prompts, as overly complex prompts may lead to a decline in summary quality. Additionally, our analysis indicates that instruction tuning does not consistently enhance the news summarization capabilities of SLMs. This research not only contributes to the understanding of SLMs but also provides practical insights for researchers seeking efficient summarization solutions that balance performance and resource use.

cs.CL↗

arXiv2024

Fast-PGM: Fast Probabilistic Graphical Model Learning and Inference

Probabilistic graphical models (PGMs) serve as a powerful framework for modeling complex systems with uncertainty and extracting valuable insights from data. However, users face challenges when applying PGMs to their problems in terms of efficiency and usability. This paper presents Fast-PGM, an efficient and open-source library for PGM learning and inference. Fast-PGM supports comprehensive tasks on PGMs, including structure and parameter learning, as well as exact and approximate inference, and enhances efficiency of the tasks through computational and memory optimizations and parallelization techniques. Concurrently, Fast-PGM furnishes developers with flexible building blocks, furnishes learners with detailed documentation, and affords non-experts user-friendly interfaces, thereby ameliorating the usability of PGMs to users across a spectrum of expertise levels. The source code of Fast-PGM is available at https://github.com/jjiantong/FastPGM.

cs.LG↗

arXiv2023

Interpretable Traffic Event Analysis with Bayesian Networks

Although existing machine learning-based methods for traffic accident analysis can provide good quality results to downstream tasks, they lack interpretability which is crucial for this critical problem. This paper proposes an interpretable framework based on Bayesian Networks for traffic accident prediction. To enable the ease of interpretability, we design a dataset construction pipeline to feed the traffic data into the framework while retaining the essential traffic data information. With a concrete case study, our framework can derive a Bayesian Network from a dataset based on the causal relationships between weather and traffic events across the United States. Consequently, our framework enables the prediction of traffic accidents with competitive accuracy while examining how the probability of these events changes under different conditions, thus illustrating transparent relationships between traffic and weather events. Additionally, the visualization of the network simplifies the analysis of relationships between different variables, revealing the primary causes of traffic accidents and ultimately providing a valuable reference for reducing traffic accidents.

cs.LG↗

arXiv2022

Fast Parallel Bayesian Network Structure Learning

Bayesian networks (BNs) are a widely used graphical model in machine learning for representing knowledge with uncertainty. The mainstream BN structure learning methods require performing a large number of conditional independence (CI) tests. The learning process is very time-consuming, especially for high-dimensional problems, which hinders the adoption of BNs to more applications. Existing works attempt to accelerate the learning process with parallelism, but face issues including load unbalancing, costly atomic operations and dominant parallel overhead. In this paper, we propose a fast solution named Fast-BNS on multi-core CPUs to enhance the efficiency of the BN structure learning. Fast-BNS is powered by a series of efficiency optimizations including (i) designing a dynamic work pool to monitor the processing of edges and to better schedule the workloads among threads, (ii) grouping the CI tests of the edges with the same endpoints to reduce the number of unnecessary CI tests, (iii) using a cache-friendly data storage to improve the memory efficiency, and (iv) generating the conditioning sets on-the-fly to avoid extra memory consumption. A comprehensive experimental study shows that the sequential version of Fast-BNS is up to 50 times faster than its counterpart, and the parallel version of Fast-BNS achieves 4.8 to 24.5 times speedup over the state-of-the-art multi-threaded solution. Moreover, Fast-BNS has a good scalability to the network size as well as sample size. Fast-BNS source code is freely available at https://github.com/jjiantong/FastBN.

cs.LG↗