arXiv subjects

Yuhuan Liu

Publications and source records attributed to Yuhuan Liu.

4 recordsLinked to original sources

arXiv2026

SCAN: Sparse Circuit Anchor Interpretable Neuron for Lifelong Knowledge Editing

Large Language Models (LLMs) often suffer from catastrophic forgetting and collapse during sequential knowledge editing. This vulnerability stems from the prevailing dense editing paradigm, which treats models as black boxes and relies on coarse-grained parameter interventions that inevitably disrupt preserved knowledge. To address this, we propose SCAN (a sparse editing framework based on Sparse Circuit Anchored Neuron) which transforms editing into a mechanism-aware manipulation by constructing a knowledge circuit via Sparse Transcoders. Experiments on Gemma2, Qwen3, and Llama3.1 across CounterFact, ZsRE and WikiFactDiff demonstrate that SCAN achieves a superior performance, maintaining model integrity on benchmarks like MMLU and GSM8K even after 3,000 sequential edits, whereas other existing methods deteriorate progressively as editing accumulates, eventually resulting in model collapse.

cs.AI↗

arXiv2025

A general language model for peptide function identification

Accurate identification of bioactive peptides (BPs) and protein post-translational modifications (PTMs) is essential for understanding protein function and advancing therapeutic discovery. However, most computational methods remain limited in their generalizability across diverse peptide functions. Here, we present PDeepPP, a unified deep learning framework that integrates pretrained protein language models with a hybrid transformer-CNN architecture, enabling robust identification across diverse peptide classes and PTM sites. We curated comprehensive benchmark datasets and implemented strategies to address data imbalance, allowing PDeepPP to systematically extract both global and local sequence features. Through extensive analyses including dimensionality reduction and comparison studies, PDeepPP demonstrates strong, interpretable peptide representations and achieves state-of-the-art performance in 25 of the 33 biological identification tasks. Notably, PDeepPP attains high accuracy in antimicrobial (0.9726) and phosphorylation site (0.9984) identification, with 99.5% specificity in glycosylation site prediction and substantial reduction in false negatives in antimalarial tasks. By enabling large-scale, accurate peptide analysis, PDeepPP supports biomedical research and the discovery of novel therapeutic targets for disease treatment. All code, datasets, and pretrained models are publicly available via GitHub (https://github.com/fondress/PDeepPP) and Hugging Face (https://huggingface.co/fondress/PDeppPP)

cs.LG↗

arXiv2025

REACT: Representation Extraction And Controllable Tuning to Overcome Overfitting in LLM Knowledge Editing

Large language model editing methods frequently suffer from overfitting, wherein factual updates can propagate beyond their intended scope, overemphasizing the edited target even when it's contextually inappropriate. To address this challenge, we introduce REACT (Representation Extraction And Controllable Tuning), a unified two-phase framework designed for precise and controllable knowledge editing. In the initial phase, we utilize tailored stimuli to extract latent factual representations and apply Principal Component Analysis with a simple learnbale linear transformation to compute a directional "belief shift" vector for each instance. In the second phase, we apply controllable perturbations to hidden states using the obtained vector with a magnitude scalar, gated by a pre-trained classifier that permits edits only when contextually necessary. Relevant experiments on EVOKE benchmarks demonstrate that REACT significantly reduces overfitting across nearly all evaluation metrics, and experiments on COUNTERFACT and MQuAKE shows that our method preserves balanced basic editing performance (reliability, locality, and generality) under diverse editing scenarios.

cs.AI↗

arXiv2021

Design and Implementation of Real-Time Localization System (RTLS) based on UWB and TDoA Algorithm

Nowadays, accurate localization plays an essential role in many fields, like target tracking and path planning. The challenges of indoor localization include inadequate localization accuracy, unreasonable anchor deployment in complex scenarios, lack of stability, and high cost. So the universal positioning technologies cannot meet the real application requirements scarcely. To overcome these shortcomings, a comprehensive Ultra Wide-Band (UWB) based RTLS is presented in this paper. We first introduce the architecture of the real-time localization system, then propose a new wireless clock synchronization (WCS) scheme, finally discuss the time difference of arrival (TDoA) algorithm. We define the time-base selection strategy for the TDoA algorithm, and analyze the relationship between anchor deployment and positioning accuracy. The Extended Kalman Filter (EKF) method is presented for non-linear dynamic localization estimation, and it performs well in terms of stability and accuracy in moving targets.

eess.SY↗