Researcher profile

Wenjun Wang

Wenjun Wang contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

Machine Learning Artificial Intelligence Computation and Language Information Retrieval Information Theory Neural and Evolutionary Computing Social and Information Networks

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author

7works

0followers

7topics

4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring

Low-resource deployment constraints have made model quantization essential for deploying neural networks while preserving performance. Meanwhile, model merging has become an increasingly practical low-resource strategy for integrating multiple task- or domain-specialized experts into a single model without joint training or multi-model serving. Together, quantization and model merging enable an efficient low-resource deployment pipeline by integrating multiple experts into one low-bit model. We formulate this setting as Post-Merge Quantization (PMQ). We show that directly applying post-training quantization (PTQ) to a merged model is unreliable because two distinct deviations are coupled: the quantization deviation introduced by low-bit reconstruction and the expert-relative merging deviation inherited from model merging. To mitigate these deviations, we propose E-PMQ, an expert-guided PMQ framework that uses source expert weights to provide expert- guided output targets during layer-wise calibration, together with merged-weight anchoring to stabilize the calibration and preserve the integrated behavior of the merged model. On CLIP-ViT-B/32 eight-task merging, E-PMQ improves 4-bit GPTQ from 65.0% to 73.6% under Task Arithmetic and from 69.1% to 74.8% under TIES-Merging. On harder settings, E-PMQ improves GPTQ from 34.8% to 76.7% on 20-task CLIP-ViT-L/14 and from 78.26% to 83.34% on FLAN-T5- base GLUE. These results demonstrate that E-PMQ enables effective post-merge quantization and low-bit deployment.

preprint2026arXiv

FeatCal: Feature Calibration for Post-Merging Models

Model merging combines task experts into one model and avoids joint training, retraining, or deploying many expert models, but the merged model often still underperforms task experts. We study this performance gap through feature drift, the difference between features produced by the merged model and by the expert on the same input. Our theory decomposes this drift into upstream propagation and local mismatch, tracks how it propagates and combines through later layers in forward order, and links final feature drift to output drift. This view motivates FeatCal, which uses a small calibration set to calibrate the merged model weights layer by layer in forward order, reducing feature drift while staying close to merged weights and preserving the benefits of model merging. FeatCal uses an efficient closed-form solution to update model weights, with no gradient descent, iterative optimization, or extra modules. On the main CLIP and GLUE benchmarks, FeatCal beats Surgery and ProbSurgery, the closest post-merging calibration baselines: 85.5% vs. 77.0%/78.8% on CLIP-ViT-B/32 Task Arithmetic (TA) and 85.2% vs. 83.7%/82.2% on FLAN-T5-base GLUE. On CLIP-ViT-B/32, 8 examples per task reach 82.9%, and 256 examples per task take 53 seconds, about 4x faster than both baselines, showing better sample efficiency and lower calibration cost.

preprint2026arXiv

Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training

Continual post-training aims to extend large language models (LLMs) with new knowledge, skills, and behaviors, yet it remains unclear when sequential updates enable capability transfer and when they cause catastrophic forgetting. Existing methods mitigate forgetting through sequential fine-tuning, replay, regularization, or model merging, but offer limited criteria for determining when incorporating new updates is beneficial or harmful. In this work, we study LLM continual post-training through three questions: What drives forgetting? When do sequentially acquired capabilities transfer or interfere? How can compatibility be used to control update integration? We address these questions through task geometry: we represent each post-training task by its parameter update and study the covariance geometry induced by the update. Our central finding is that: forgetting can be considered as a state-relative update-integration failure, it arises when the covariance geometries induced by tasks misalign with the geometry of the evolving model state. Sequential updates transfer when they remain compatible with the model state shaped by previous updates, and interfere when state-relative geometry conflict becomes high. Motivated by this finding, we propose Geometry-Conflict Wasserstein Merging (GCWM), a data-free update-integration method that constructs a shared Wasserstein metric via Gaussian Wasserstein barycenters and uses geometry conflict to gate geometry-aware correction. Across Qwen3 0.6B--14B on domain-continual and capability-continual settings, GCWM consistently outperforms data-free baselines, improving retention and final performance without replay data. These results identify geometry conflict as both an explanatory signal for forgetting and a practical control signal for LLM continual post-training.

preprint2022arXiv

DHEN: A Deep and Hierarchical Ensemble Network for Large-Scale Click-Through Rate Prediction

Learning feature interactions is important to the model performance of online advertising services. As a result, extensive efforts have been devoted to designing effective architectures to learn feature interactions. However, we observe that the practical performance of those designs can vary from dataset to dataset, even when the order of interactions claimed to be captured is the same. That indicates different designs may have different advantages and the interactions captured by them have non-overlapping information. Motivated by this observation, we propose DHEN - a deep and hierarchical ensemble architecture that can leverage strengths of heterogeneous interaction modules and learn a hierarchy of the interactions under different orders. To overcome the challenge brought by DHEN's deeper and multi-layer structure in training, we propose a novel co-designed training system that can further improve the training efficiency of DHEN. Experiments of DHEN on large-scale dataset from CTR prediction tasks attained 0.27\% improvement on the Normalized Entropy (NE) of prediction and 1.2x better training throughput than state-of-the-art baseline, demonstrating their effectiveness in practice.

preprint2022arXiv

FadMan: Federated Anomaly Detection across Multiple Attributed Networks

Anomaly subgraph detection has been widely used in various applications, ranging from cyber attack in computer networks to malicious activities in social networks. Despite an increasing need for federated anomaly detection across multiple attributed networks, only a limited number of approaches are available for this problem. Federated anomaly detection faces two major challenges. One is that isolated data in most industries are restricted share with others for data privacy and security. The other is most of the centralized approaches training based on data integration. The main idea of federated anomaly detection is aligning private anomalies from local data owners on the public anomalies from the attributed network in the server through public anomalies to federate local anomalies. In each private attributed network, the detected anomaly subgraph is aligned with an anomaly subgraph in the public attributed network. The significant public anomaly subgraphs are selected for federated private anomalies while preventing local private data leakage. The proposed algorithm FadMan is a vertical federated learning framework for public node aligned with many private nodes of different features, and is validated on two tasks correlated anomaly detection on multiple attributed networks and anomaly detection on an attributeless network using five real-world datasets. In the first scenario, FadMan outperforms competitive methods by at least 12% accuracy at 10% noise level. In the second scenario, by analyzing the distribution of abnormal nodes, we find that the nodes of traffic anomalies are associated with the event of postgraduate entrance examination on the same day.

preprint2022arXiv

Representation Learning on Heterostructures via Heterogeneous Anonymous Walks

Capturing structural similarity has been a hot topic in the field of network embedding recently due to its great help in understanding the node functions and behaviors. However, existing works have paid very much attention to learning structures on homogeneous networks while the related study on heterogeneous networks is still a void. In this paper, we try to take the first step for representation learning on heterostructures, which is very challenging due to their highly diverse combinations of node types and underlying structures. To effectively distinguish diverse heterostructures, we firstly propose a theoretically guaranteed technique called heterogeneous anonymous walk (HAW) and its variant coarse HAW (CHAW). Then, we devise the heterogeneous anonymous walk embedding (HAWE) and its variant coarse HAWE in a data-driven manner to circumvent using an extremely large number of possible walks and train embeddings by predicting occurring walks in the neighborhood of each node. Finally, we design and apply extensive and illustrative experiments on synthetic and real-world networks to build a benchmark on heterostructure learning and evaluate the effectiveness of our methods. The results demonstrate our methods achieve outstanding performance compared with both homogeneous and heterogeneous classic methods, and can be applied on large-scale networks.

preprint2021arXiv

A Novel Genetic Algorithm with Hierarchical Evaluation Strategy for Hyperparameter Optimisation of Graph Neural Networks

Graph representation of structured data can facilitate the extraction of stereoscopic features, and it has demonstrated excellent ability when working with deep learning systems, the so-called Graph Neural Networks (GNNs). Choosing a promising architecture for constructing GNNs can be transferred to a hyperparameter optimisation problem, a very challenging task due to the size of the underlying search space and high computational cost for evaluating candidate GNNs. To address this issue, this research presents a novel genetic algorithm with a hierarchical evaluation strategy (HESGA), which combines the full evaluation of GNNs with a fast evaluation approach. By using full evaluation, a GNN is represented by a set of hyperparameter values and trained on a specified dataset, and root mean square error (RMSE) will be used to measure the quality of the GNN represented by the set of hyperparameter values (for regression problems). While in the proposed fast evaluation process, the training will be interrupted at an early stage, the difference of RMSE values between the starting and interrupted epochs will be used as a fast score, which implies the potential of the GNN being considered. To coordinate both types of evaluations, the proposed hierarchical strategy uses the fast evaluation in a lower level for recommending candidates to a higher level, where the full evaluation will act as a final assessor to maintain a group of elite individuals. To validate the effectiveness of HESGA, we apply it to optimise two types of deep graph neural networks. The experimental results on three benchmark datasets demonstrate its advantages compared to Bayesian hyperparameter optimization.

Institution

Affiliation not imported yet

This author record came from a source that does not expose affiliation metadata. Once the author claims the profile or we enrich the record from another provider, this section will link to the concrete institution.

Topic footprint